The 10th International Conference on Computer Engineering and Networks

The current widely spread of micro-form video is undeniable. However, neither the music-orienting nor the video-orienting way to make a video can totally express the idea of the video maker because of the fixed music types and the cost of time to find a proper music. Based on the deep learning method, this paper studies the automatic matching algorithm of background music for micro-videos which analyzes the background information and the emotions of characters in micro-videos and establishes the model to select the proper background music according to the video contents. Experiments are carried out on the database obtained from TikTok and the result shows that the current Micro-Video Background Music Automatic Matching model in this paper is effective.

The incidence of pneumonia and lung cancer is high in China. Lung CT images are widely used in the screening and adjuvant treatment of lung diseases due to their advantages of a thin layer, high definition, and low noise. The manual film-reading method has strong subjectivity, depends on the doctor’s experience, and now hospitals produce a large number of lung CT images every day. The manual film-reading method is inefficient. The machine can help doctors to do lesion screening, auxiliary diagnosis, and treatment. Conventional machine learning methods have been applied to the recognition of pulmonary nodules in CT images. Still, these methods need to manually extract features that are not comprehensive or proper, resulting in misdiagnosis and missed diagnosis. These methods enable to rapidly detect lung nodules, which is terrible for patients to receive treatment in a timely manner. With the development of deep learning technology, Convolutional Neural Networks (CNNs) have been widely used in recognizing images, such as facial identification, character recognition, car plate recognition, etc. Its ability to solve computer vision problems has won the approval in some natural scene tasks. In this paper, the YOLO_v3 is used for feature extraction and classification, and residual network (ResNet), one of the classical CNN models, is used to decrease the false-positive rate, which improves the detection accuracy of pulmonary nodules in CT images.

Customer segmentation based on mixed variables is an important research direction of current market segmentation methods. However, using the hybrid clustering method to divide consumer groups will overestimate the clustering contribution of categorical variables and shield numerical variables that have important marketing significance. In this study, we improve the hybrid clustering algorithm and design an indicator of assessing the customer groups’ differences. Firstly, the coefficient of Features Discrepancy between Segments (FDS) and three variable weighting strategies are designed. Then, the optimal optimization scheme is determined based on three hybrid clustering algorithms and evaluation indicators. The results show that the variability weighting method can effectively improve the problem that categorical variables dominate hybrid clustering. The clustering performance and stability of PAM method are the best among the three hybrid clustering algorithms. Finally, clustering the consumer consumption values and the basic population information through the weighted PAM method to verify this method’s effectiveness. This study provides a practical application value for the improvement of existing technologies in customer segmentation methods. It also offers the marketing suggestions for the table grape operators.

The Shanghai and Shenzhen 300 Index covers the epitome of 60% of the Shanghai and Shenzhen stock markets, and is a high-level summary of the Shanghai and Shenzhen stock markets. If we can predict the closing price of the Shanghai and Shenzhen 300 Index, it will guide the investment of the Shanghai and Shenzhen stock markets, promote the sound development of the stock market, and improve the overall economic strength of the country. For the study of nonlinear systems, neural networks have unique advantages, especially BP neural networks, which can train, learn and predict complex data. In this paper, BP neural network is used to establish the BP neural network model of the Shanghai and Shenzhen 300 Index. The opening price, the highest price and the lowest price of the Shanghai and Shenzhen 300 Index are used as input variables, and the closing price is taken as the output variable to study and predict the closing of the Shanghai and Shenzhen 300 Index. Through the model results, it is found that the BP neural network model can better simulate the Shanghai and Shenzhen 300 Index and achieve the effect of predicting the Shanghai and Shenzhen 300 Index. Applying BP neural network model to nonlinear estimation of the Shanghai and Shenzhen 300 Index, solving nonlinear problems, can promote the healthy development of China’s stock market.

Planetary gearboxes of cranes are often in high-speed, heavy-duty operating environments, and critical components such as gears and bearings are prone to failure. Diagnosing the fault information of key components in time can effectively avoid losses and accidents caused by mechanical equipment’s shutdown. Aiming at the problems of strong time-varying and unstable vibration signals of planetary gearbox, a fault diagnosis method based on time-frequency features and PSO-SVM is proposed. First, the STFT is used to process the planetary gearbox’s vibration signal to obtain the time-frequency spectrum characteristics. Then, use a support vector machine to process the time-frequency spectrum features to identify the fault type. Simultaneously, the PSO algorithm is used to optimize the support vector machine’s parameters to improve the classification and recognition ability. The analysis experiment and algorithm test of the planetary gearbox’s vibration signal verify that the method can extract effective characteristic parameters from non-stationary and time-varying signals and can quickly and accurately identify the fault type of the planetary gearbox.

The paper aims to expand a way of applying the fuzzy adaptive control strategy by analyzing the advantages and disadvantages of traditional PID control strategies. In view of the high requirements of the variable-pitch servo system for megawatt-level wind turbines, combined with the control characteristics of series-excited DC servo motors, a wind power variable-pitch DC servo system based on fuzzy adaptation was designed. The superiority and reliability of the fuzzy adaptive PI control method are verified by using simulation tools to carry out simulation experiments, and then the experiments are carried out by DC series servo motor and servo driver. In summary, the performance of the wind power variable pitch DC servo system is excellent.

Motor imagery-based Brain-Computer Interface (MI-BCI) has already become one of the hottest research fields and has made great achievement in stroke rehabilitation. Because it is difficult for stroke patients to complete a series of motor imagery tasks, the phenomenon of insufficient data in signal processing and application is often faced. At present, the most effective way to overcome the challenge is to expand the amount of data for analysis. In this paper, we used CycleGAN for data augmentation on brain activities. First, electroencephalogram data is transformed into brain topographic map through EEG2Image. After that, we train a CycleGAN neural network. The topographic maps of healthy people and stroke patients are put into the CycleGAN model for training at the same time. In the process of training, let healthy people learn the features of the stroke patients topographic maps and generate the artificial stroke patients topographic maps. The generated topographic map can be very intuitive to see the similarity with the original topographic map. Experiments demonstrate that the method effectively generate artificial topographic maps to expand stroke dataset. And these topographic maps can be applied to data processing and analysis of MI-BCI.

Ultrasonic flaw detection is the most commonly used method in rail defect detection, and the defect judgment in flaw detection data is the most important link. The B-scan data of ultrasonic flaw detection vehicle is used for data analysis and recovery. A set of rail defect identification model based on BP neural network is established according to the channel distribution characteristics, digital combination characteristics and the arrangement and combination characteristics in time sequence of the recovered data. This method takes the 1–6 channel detection data as an example, and trains the model by extracting the rail head nuclear defect as the sample and making the training set, Experiments show that the model can effectively identify the defect data by setting a suitable threshold value, and give the location of the rail defect, help the flaw detection workers to improve the efficiency of rail defect detection, greatly reduce the labor intensity, and avoid the missed judgment of the rail defect.

Apple’s online grading technology is an important part of Apple’s commercialization. In order to further improve the efficiency of apple grading, In this paper, we study Fuji apple grading method based on machine vision and support vector machine model optimized by genetic algorithm. An apple image acquisition system was built, and the median filtering method was used to remove the noise of the image, then the target region was obtained using canny edge detection algorithm combined with morphological processing. Finally, serving as input, the roundness of apple, mean value of H component in the HSI color space, and the surface defect as the characteristic parameters was sent into the GA-SVM model. The experimental results show that the classification accuracy rate is 92.3%.

In most cities, the monitoring stations are sparse, and air pollution is affected by various internal and external factors. To promote the performance of regional pollution prediction, a deep spatio-temporal dense network model is proposed in this paper. First, the inverse distance weighted (IDW) space interpolation algorithm is used to construct historical emission data with insufficient station records. Based on the properties of spatial-temporal data, a deep spatial-temporal dense network model is designed to predict air pollution in each region. Finally, several experiments are conducted on the real dataset of Hangzhou. The results show that combing IDW spatial interpolation and deep spatial-temporal dense network model can effectively predict the regional air pollution and achieve superior performance compared with ARIMA, CNN, ST-ResNet, CNN-LSTM.

With the rapid development of the Internet and the continuous improvement of online shopping facilities, online shopping has been accepted by contemporary people. Taking JD e-commerce platform as an example, this study explores the relationship among three variables: electronic word-of-mouth, trust and online consumers’ purchase intention, and puts forward the following research hypotheses: (1) electronic word-of-mouth has a positive impact on the trust of online potential consumers, (2) electronic word-of-mouth has a positive impact on the purchase intention of online potential consumers, (3) trust has a positive impact on the purchase intention of online potential consumers. This study uses the quantitative questionnaire survey method and SPSS software to analyze the data, expecting that the research results can promote the operation managers to pay attention electronic word-of-mouth, develop and improve the consumer feedback system, establish the good reputation of enterprises, and promote the operation performance of e-commerce platform.

With the development of image processing technology and application, the recognition requirements for facial details are getting higher and higher. This paper presents a face color comparison method, which can compare the facial color difference quickly. At the same time, through experiments, this paper determined the formula of color difference used by the method and obtained the good accuracy of the method in accordance with human vision, which shows that the method has certain practicability in facial color comparison.

As deep learning continues to be a hot research issue, analysis of brain activities based on deep neural networks have been in the rapid development. In this paper, the convolution neural network is introduced to decode the electroencephalogram (EEG) from stroke patients to design an effective motor imagery based brain-computer interface system. We develop a novel convolution neural network architecture combined with transfer learning to recognize motor imagery tasks from stroke patients. We also explore intra-subject and inter-subject transfer learning methods to evaluate the performance of the proposed framework and to avoid time-consuming training process to low algorithmic complexities. The fine-tune technology has been employed to transfer model parameters. The proposed framework is a combination of EEGNet model and fine-tune technology. The proposed framework is evaluated on stroke patients, the average classification accuracy is 65.91%. Experimental results validate that the proposed algorithm can obtain satisfactory performance improvement for EEG-based stroke rehabilitation of brain-computer interface system, and show that our model has potential for effective stroke rehabilitation motor imagery based brain-computer interface systems.

In order to try different lipstick products in a reusable, low-cost, and hygienic manner, an intelligent virtual lipstick trial algorithm is proposed with assistance of Dlib and OpenCV. In the proposed method, we perform face detection through the pre-trained face detection model of dlib.get_frontal_face_detector in the Dlib library, and then extract feature points from the face in the video according to the above pre-trained model. When lips and other important face parts are recognized, they are then filled with specific color patterns according to lighting environment, highlighting the import characteristics for target lipstick, making it reliably displayed on a smart phone screen. The process is full of simplicity, quickness and convenience.

The N-queens problem is a classical NP hard problem with many solving methods. In this paper, a hybrid chemical reaction optimization algorithm is proposed to solve the N-queens problem. This algorithm combined with chemical reaction optimization algorithm and greedy algorithm, is presented to solve the problem of mixed chemical molecular structure, and design the appropriate molecular coding and chemical reaction of the four basic reaction process and the objective function, the simulation experimental results show that the design of hybrid optimization algorithm to solve the N-queens problem of the chemical reaction have got improved.

Soluble solids contents (SSC) is the general term of monosaccharide, disaccharide and polysaccharide, and is one of the key evaluation indexes of apple taste and nutritional quality. According to the prediction of soluble solid content in Red Fuji apple, a method based on Genetic Algorithm and Extreme Learning Machine was proposed. Genetic Algorithm (GA) was used to select the best near-infrared spectrum of Fuji apple, and the prediction model of Extreme Learning Machine (ELM) was established. The prediction correlation coefficient of the model was 0.9723, and the prediction root mean square error was 0.1855. The experimental results show that the predicted results are in good agreement with the actual data, which plays a good role in the prediction of soluble solid content.

Road cracks seriously affect road life and driving safety, and road crack segmentation has been a key task in many research fields and industries. Scenes of existing road crack datasets are relatively homogeneous, most datasets only have local crack images taken from a single perspective, and the images lack complex interference pattern other than the cracks and the road itself. However, real-world images of roads are much more complex, and roads may have other dark patterns that interfere with crack segmentation. Perspective and lighting conditions may also be different between every image in the dataset, which makes many road crack segmentation methods extremely low accuracy and severely affects the generalization ability of the semantic segmentation model, while many applications require higher precision rather than recall. In this paper, two methods are proposed to improve the accuracy: augmenting complex patterns in road images and the fusion of classification networks. The methods greatly improve the performance of road crack segmentation.

The traditional local invariant features scene matching algorithms have many redundant points, poor real-time and Low anti-geometric transformation. So we propose an unmanned aerial vehicle scene matching algorithm based on CenSurE features and FREAK descriptor. Steps of the algorithm are as follows: use the CenSurE-star operator to extract the feature points, calculate feature vectors by FREAK descriptors, then complete matching process through KNN ratio method, finally use RANSAC location model to get image transform relations, latitude and longitude. Experiments show that: compared with SIFT, SURF, ORB algorithms, this algorithm not only has better robustness for various transformation, but also greatly shorten the processing time of about 86 ms. The positioning error within 0.6 pixel, scale error within 0.02 times, rotation Angle error within 0.04°.

Due to its special imaging conditions, aerial reconnaissance images are easy to be degraded by noise, atmospheric condition, defocus, light exposure, motion-displacement and so on. And the image quality degradation comes down to noise, low contrast and low definition. In order to choose image automatically from massive image data to use in interpretation, appropriate image quality assessment method fit to aerial images should be chosen. The experiments were designed to compare commonly used methods. And the conclusions were that the Mean square error and Mean gradient methods assess noise more accurately. Edge preserving index method assesses definition more accurately, Standard deviation method assesses contrast more accurately, and Structure Similarity (SSIM) method assesses more accurately in image comprehensive quality.

Based on the three typical patterns of e-commerce poverty alleviation found in the field research in many places in Shandong province—“peasant household + e-commerce platform”, “e-commerce park + enterprise + peasant household”, “peasant household + cooperative + e-commerce platform”, this paper conducts an in-depth analysis of the three models from the two aspects of mode operation mechanism and enabling mechanism, and compares the development stage, applicable conditions, enabling characteristics and poverty alleviation effect among the models, summarizes the advantages of various models, and is committed to providing a certain reference for regions that have not yet been lifted out of poverty.

Emotional analysis has been omnipresent. Grabbing adequate information from online shopping evaluation and analyzing it can provide a basis to choose for businesses or consumers. Taking “Jingdong Mall” mobile e-commerce review as an example, the authors of this paper use web crawler technology to obtain sufficient data of 12 products under 7 mobile phone brands in the sales ranking. The authors then use Python’s Jieba library to segment the reviews and gensim module to train word2vec word vector model and classify the mobile e-commerce reviews by constructing LSTM neural network model. It is concluded that the consumers’ views that there exists some advantages and disadvantages of “Apple” and “Huawei” mobile phones.

With the acceleration of urbanization in China, as an essential part of it, parks play an important role in improving the living environment, to promote city brand, and to enhance citizens’ sense of happiness and gain. Currently, the management of the park mainly adopts the manual inspection method that has many disadvantages. This paper proposes a park inspection system based on artificial intelligence. The system has the following main advantages: using multi-source sensing technology and location algorithm, using artificial intelligence, automatic control, not affected by time, place, and other external factors, can improve the work efficiency, enhance the coverage, make the inspection work accurate and effective. It can also reduce maintenance costs and facilitate the construction of intelligent monitoring networks in public places such as parks during the epidemic prevention and control period.

Artificial Intelligence is booming nowadays and has developed tremendous effect in many areas. For Telecommunication, due to the growing network complexity and business diversity, it has been a challenge for operation and maintenance (O&M) of a wireless network. This paper analyzes the requirement for intelligent network O&M, sorts out the selection of artificial intelligence (AI) algorithm, and discusses the application scenarios of relevant algorithms in the field of network O&M. Based on 4 cases of intelligent application, this paper reveals how efficiency and quality of network O&M will be improved by AI technology.

Recently some accidents brought the safety of autonomous vehicles to the public’s attention. Autonomous vehicles are not only hardware products, but also a system of multi-sensor perception modules and Deep Neural Networks (DNNs) software. There are many methods to test traditional software. However, finding a good enough software testing method for vehicles on-board DNNs models is still an open problem. In this paper, we provide a testing method based on boundary value testing. For DNNs, many studies have shown that adversarial examples can disable DNNs, and they are often closed to the decision boundary. Therefore, adversarial examples can be used as test cases. To accomplish this task, we use two novel algorithms to design boundary value test cases: a white-box method based on gradient information and a black-box method using Deep Reinforcement Learning. Further, we propose a whole testing procedure, including a computer simulation test, sensor unit test and real vehicle test. The experimental results show that it is feasible to use adversarial examples for testing and our method could increase efficiency. Our work may have inspirations for automobile manufacturers and researchers.

With the rapid development of Internet technology and artificial intelligence technology, traditional agriculture has ushered in new development opportunities. Integrating artificial intelligence technology into the construction of intelligent agriculture and organically integrating with IoT technology can promote the vigorous development of intelligent agriculture in China. Through the research and analysis of some major problems existing in the current application of artificial intelligence technology in agriculture in China, this paper puts forward Suggestions that artificial intelligence technology can be applied in the implementation of technological innovation in the construction of intelligent agriculture. The example is the construction of fruit and vegetable epidemic disease accurate diagnosis information system based on artificial intelligence technology. This paper discusses the specific application of artificial intelligence technology to modern agriculture and discusses the positive significance of artificial intelligence technology to the development of intelligent agriculture.

In order to improve the recognition rate of face recognition, a face expression extraction method is designed. Firstly, wavelet transform was used to extract low-frequency components, then PCA was used to map low-frequency images to low-dimensional space, and finally, FDA method was used to extract face features in low-dimensional space. Experimental results show that this method can effectively overcome the influence of facial expression and gesture changes on face recognition, and not only improve the face recognition rate, but also improve the recognition speed.

At present, the emergence of this kind of recruitment website promotes the recommendation calculation method’s rapid progress. However, the old recommendation algorithm is based on the preferences or requirements of individual participants. The current job-hunting situation should give not only excellent employees but also give suitable jobs to job seekers. So there is this kind of Reciprocal Recommendation Method. The purpose of such a way is to recommend customers one-to-one. There is a similarity between the two, and use this similarity to ensure the suitability and stability of the job. In this system, enterprises can confirm the keywords needed within the range of resume selection setting, and the system will select students’ resumes according to the keywords set. In this paper, we combine the characteristics of Chinese resume, analyze the two-way reciprocal recommendation mode of school enterprises and students, and finally use the simulation test to test the accuracy and effect of the content-based recommendation algorithm.

Aiming at the sparsity problem existing in the traditional collaborative filtering algorithm, this paper proposed an improved similarity computing method that integrated user rating behavior and item attributes. The sparse matrix is evaluated and predicted by the similarity calculation method, then the prediction rating was filled in the sparse matrix. At the same time, in the context of big data, the data scale was too large to affect the execution efficiency of the recommendation system. Hadoop platform was adopted to implement collaborative filtering recommendation algorithm based on the improved similarity model. Based on large-scale data segmentation, the distributed parallel processing was carried out. The proposed improved algorithm is verified by Movielens which was an internationally standard data set. The verification results show that the personalized recommendation system based on Hadoop platform and improved recommendation algorithm has better recommendation performance.

With the consumer’s request to agricultural products is more and more strict, a traceability system of agricultural products quality based on hybrid mode is designed and developed in this paper. The system connects the various links of crop planting, acquisition, processing, distribution and sales, and obtains the real-time information from crop seed cost to agricultural product sales. It can make farmland scientifically be managed according to the information of crop growth environment, improve the use efficiency of agricultural resources, and is conducive to the sustainable development of agriculture. In addition, information processing and transmission, and sales of the crops from the seeding to the harvest and to the agricultural products are recorded in real time in the database so that the consumers can understand real-time information on produce. When unqualified agricultural products are detected, the production process of such products can be immediately controlled from the source, and the circulation of unqualified agricultural products can be more effectively controlled. In this way, the rights in the interest of consumers can be effectively guaranteed.

An automatic intelligent implementation scheme is proposed based on the analysis of the characteristics of the information network transmission services of surveying ship by corresponding service strategies designation, which can give high quality services. The key of this scheme is an intelligent bandwidth allocation system that is designed for this scheme. This bandwidth allocation method effectively accomplish the intelligent redistribution of the limited bandwidth and improve the efficiency and reliability of the whole public information network of the surveying ship through the mirror data grabbing and analyzing and accomplishing the remote controlling of the encoder and the switch by remote bandwidth adjustment instructions.

Phrases as a level of linguistic analysis, occupying a very important position in this field. Effective phrase analysis is crucial to reduce difficulty of subsequent syntactic analysis, and reduce search space of syntactic analyzer, as well as improving accuracy of machine translation. At present, research in Tibetan phrases for information processing has just started; it needs to be further developed. Based on previous studies on boundary between Tibetan phrases and Tibetan sentences, with characteristics and requirements of Tibetan information processing, in accordance with grammatical function and the principle of automatic analysis to classification of phrases processing, and specify markup codes for Tibetan phrase units in information processing are discussed in this paper.

In this paper, we present a garbage image classification framework to tackle the waste sorting problem which besets residents around the world every day. The proposed framework consists of two modules, a convolutional neural network backbone transferred from the ImageNet classification task and a customized network header designed for the garbage image classification task. To friendly deploy into mobile devices, the proposed method makes an artful tradeoff between classification accuracy and running efficiency. The proposed framework yields 95.62% online classification accuracy on the test dataset provided by Huawei cloud with 95 ms per image inference time occupying 897 Megabytes GPU memory.

Aiming at the problem of ancient ceramics classification, this paper proposes a method of ancient ceramics classification based on BP neural network. Making appropriate improvements to the shortcomings of the model: using ant colony algorithm to solve the problem that the network depends on the initial value; using the method of dynamically adjusting the learning rate to solve the shortcomings of the network which is easy to overfitting; optimizing the ant colony algorithm by improving the path selection strategy and the pheromone update strategy to solve the problem that the ant colony algorithm is easy to fall into the local optimum. Finally, the ancient ceramic classification model based on the BP neural network optimized by the improved ant colony algorithm is constructed. The trained model is evaluated, and the results show that the model has a good effect on the classification of ancient ceramics.

Imaging radar can obtain excellent images of non-cooperative moving targets, which is an essential means of target detection. According to the scattering point and time-frequency characteristics of the ISAR image, it is possible to determine the physical properties such as the target’s component structure. Combining with other radar electromagnetic scattering characteristics, it is the primary technical approach to carry out the full recognition of the target.Manually captioning of the target components and the main structure is a traditional and straightforward method that lacks a quantitative standard, requires heavy workload, and the captioning result may be affected by the subjective judgment. With the development and application of deep learning technology in recent years, remarkable achievements have been made in the fields of image classification, image detection, and machine translation, and the performance of automatic image captioning has been greatly improved. Through intelligent segmentation of ISAR images and feature extraction of segmented regions, we build a database of image features and generate the corresponding text description according to the characteristics of different parts of the image applying the text generation network. Based on the above theory, this paper brings out an ISAR image annotation system design scheme based on deep learning.

Through the construction of public opinion data center as the research object, a series of data acquisition, monitoring and analysis and mining tasks based on public opinion information can provide natural language processing functions such as web data crawling and monitoring, web data parsing, text data preprocessing, text analysis and mining. Research shows that automatic summarization technology, event recognition technology, event context combing and emotion analysis technology can be based on text mining algorithm to achieve automatic keyword extraction, content simplification, similarity calculation, tracking and monitoring, and subjective text with emotional color analysis has a very good application value and efficiency.

Fire detection is quite necessary because fire is very harmful to both lives and properties of humans. However, the causes of fire are various and complicated, which leads to the difficulty of fire prevention. As the increase of video surveillance systems, fire detection from images has become a research hotspot. Traditional algorithms focused on the contour, color and movement features of fire, which are only efficient in particular scenes and cannot be generalized. With the development of deep learning, convolution neural networks with good generalization performance have been broadly used in object detection. In this paper, a one-stage object detector based on deep neural network, namely Single Shot MultiBox Detector (SSD), is introduced to fire detection. In order to improve the generalization performance of detector and tackle the training difficulty of the lack of training samples, a dataset with a large number of flame and fire images, namely HHFire are constructed. The format of annotation is the same as that of Pascal Visual Object Classes. Images from the dataset covers lots of real scenes, including forest, buildings, fields, indoors, etc. Experimental results have demonstrated the efficiency of SSD on HHFire.

The two-stream deep neural networks learn contour evolution information of human activities from image frames and motion information from information evolution between image frames. Optical flow image is one of the commonly-used methods to depict the motion information between frames. The existing optical flow images are usually obtained by off-line calculations. In this paper, a deep neural network named FlowNet2.0 is used to generate optical flow images, and based on which a compound deep neural network is proposed to merge spatial and temporal information to complete the recognition of human activities from videos. Specifically, the proposed compound deep neural network is composed of two sub-networks, static-data-stream learning network and dynamic-data-stream learning network. The former is applied to extract spatial evolution information from RGB images, and the latter generates optical flow prediction images from RGB image sequence first, and extracts temporal evolution information from optical flow images. Finally, the results of two sub-networks are combined under the fusion algorithm to achieve human activities recognition. Experimental results show that the compound deep neural network proposed in this paper can effectively identify human activities from video sequences, and the average classification accuracy on data set UCF101 can reach 83.35%.

Mobile edge computing technology which includes computing, storage, and communication enhanced the service capability of mobile network. With the development of mobile edge computing technology, caching content in the edge of network has been paid more and more attention. Considering the limitation of caching space, it is very important to design cache replacement algorithm. In this paper, we considered the device-to-device (D2D) communication aided edge cache. And proposed a deep Q-learning based cache replacement algorithm to reduce the delay of user request. Simulation results show that the proposed algorithm can reduce the delay compared to traditional replacement algorithm.

In order to solve the high latency problem caused by traffic overloading on the control plane, there is the new generation architecture with the support of the network function virtualization in 5G core network. In this architecture, the new network function AMF interacts directly with UE and is only responsible for processing access requests, so as to shorten latency of receiving requests. However, the 5G architecture still lacks a strategy to control the deployment scale of AMF instances in order to timely complete the reasonable scheduling of resources under the premise of ensuring the quality of service. We put forward a dynamic scaling strategy for 5G core network based on LSTM model with the instance scalability. First, we predict the total volume of the traffic loading in the future through the LSTM model of recurrent neural network. Then, the deployment of instances would be controlled by dynamic threshold algorithm, the increase or decrease of instances deployment number is controlled by the dynamic threshold with upper and lower limits, which would be changed according to the rate of traffic change, so as to complete the deployment of instances in the shortest time. After that, we make the simulation test using by the actual telecommunication data set, and compare our strategy with the deep neural network and the static threshold algorithm, to demonstrate the feasibility and excellent performance of our strategy.

With the rapid development of modern network technology, business traffic in the network is exploding, and network traffic control technology is facing greater challenges. Traditional network flow control methods often cannot calmly deal with complex and dynamic network states, and it is difficult to avoid network congestion. Therefore, we need an intelligent method that can update flow control decisions according to the current network state.Software defined network (SDN) has brought new ideas for controlling network traffic. In the SDN network architecture, the controller of the control plane can send the network routing policy to manage the traffic forwarding in the data plane layer, so as to implement network management and resource scheduling. The programmability of SDN also provides the possibility of implementing flow control algorithms. By combining SDN and reinforcement learning (RL), a good traffic forwarding policy can be obtained, and the problem of traffic control can be solved effectively. In order to achieve intelligent control of network traffic, this paper introduced the deep deterministic policy gradient (DDPG) algorithm in RL. With the goal of maximizing network performance, we designed and implemented a network flow control algorithm based on DDPG. We trained the model, and finally verified the effectiveness and superiority of the flow control method based on DDPG through simulation experiments.

Forests are one of the most important natural resources in the world. It brings great economic value to people’s production and life. However, the occurrence of forest fires will burn plants and kill animals. Forest fires have even caused incalculable losses. The article aims to study a forest fire prevention early warning image processing system based on machine vision. First, the foreground detection of moving targets is performed. Then use the mixed Gaussian model algorithm to determine whether there are moving objects in the video. Then track and identify the flames in the video. Then determine whether it is flame. Draw its outline and give an alarm to the supervision office when the result is fire. Then the early warning system waits for the relevant staff to deal with it. Finally, it summarizes the shortcomings of the design of a machine vision forest fire prevention early warning system in this paper and looks forward to the research prospects.

In order to solve the heavy and dangerous problems of people’s daily glass cleaning, an intelligent glass cleaning robot is designed using the Arduino MEGA 2560 control board combined with a vacuum compressor, a steering gear, a two-position two-way solenoid valve, a vacuum generator and a vacuum suction cup. The intelligent glass-wiping robot uses the Arduino programming software as the control center, and controls the running angle of 8 servos through the Arduino MEGA 2560 expansion board, realizing the gait of 4 servos bionic legs on the glass. Experiments have shown that during the cleaning process of the intelligent glass-wiping robot, the safety rope at the front end can effectively prevent the robot from falling. The intelligent glass-wiping robot uses four bionic legs of the steering gear to move, which can realize the basic functions of movement, adsorption, obstacle avoidance and cleaning.

Weeds in the corn field are extremely harmful to the growth of corn seedlings and should be controlled in time. The main domestic weeding method is still based on large-scale spraying of chemical agents. This method not only causes waste of resources, but also pollutes the environment. In order to accurately identify corn seedlings and weeds in the corn field, we should do selective spraying of chemicals. This article aims to study a ROS-based rapid identification robot system. First, this article introduces the design of the robot system. Then this article introduces the model generation subsystem. The system can identify the collected images and classify and locate the corn and weeds in the images. Next, this article introduces the communication subsystem. The communication subsystem realizes the remote control of the Spark robot, and can transfer the collected images to the model for detection. Moreover, this article introduces the experimental process of the system. Finally, the shortcomings of the design of the ROS-based seedling rapid recognition robot system are summarized and the research prospect is prospected.

With the rapid increase in the number of cars, driving safety has been paid more and more attention. The lane line recognition algorithm has also become a hot research area. In order to realize the unmanned driving of smart cars, this article aims to study a lane recognition system for machine vision. First, the realization of the host computer is introduced. The host computer is convenient to control the trolley. The host computer can also change the model, making the car adapt to driving in different environments. Next, this article introduces lane recognition. In this paper, a fully convolutional network is used to semantically segment the image to obtain the feasible region of the road. Then this article introduces the communication and control part of the system. The emphasis is on the serial communication and control implementation of the system. The system is controlled by computer and Arduino is used as the control board of the car. In addition, this article introduced the experimental process. Finally, the shortcomings of the design and implementation of a machine vision lane recognition system in this paper are summarized and the research prospect is prospected.

Intelligent transportation system (ITS) contributes to allocate transportation resources, from citywide ones to nationwide, more efficiently with the help of algorithms. Due to the fact that the ITS is a quite comprehensive field, it is necessary for researchers to have a better understanding of dominant methods and which one is proper for the targeted subjects of ITS. More importantly, with a knowledge of the remained challenges in developing ITS and relevant techniques, researchers may have a clearer direction to work on. To provide researchers with dedicated information on specific machine learning (ML) techniques used in object recognition and traffic prediction, two essential study subjects in ITS, this paper mainly focuses on deep learning and neural network (NN), one of widely-used ML algorithms, and aims to conduct a brief review on its recent applications in ITS, as well as to mine its potential usage. As a result, this review introduces some popular NN, convolutional neural network (CNN), long short-term memory (LSTM) network, gated recurrent unit (GRU) network, and their hybrid mechanism, first. Then their applications and performance in ITS are described. Finally, this paper discusses constraints on some of them and suggests some promising research directions.

Glaciers are the source of fresh water that have obvious response on climate change. It is crucial and meaningful to monitor glacier area. In this paper, a glacier detection method based on features derived from low-resolution optical data, thermal data and TerraSAR-X images is proposed. The Land Surface Temperature (LST) was firstly obtained by Convolutional Neural Network (CNN). Combined with the velocity information, a low-resolution binary mask was derived for the supervised classification of SAR imagery. Afterwards, a set of suitable features was derived from the SAR intensity image, such as texture information generated based on the gray level co-occurrence matrix (GLCM), and intensity values. With these features above, the glaciers were classified by Random Forest (RF) to distinguish the glacier from the non-glacier areas. Compared to the unsupervised classification only using SAR data, the glacier detection method proposed in this paper achieved a better performance with the overall classification accuracy of 90.88%.

Image fusion can extract redundant information of multiple images into one image, and the goal of image fusion is to better apply to classification tasks. Convolutional neural networks have proved to be an effective way for accuracy of classification. However, fusion and classification usually considered separately. In this paper, we design a ‘fusion-classification networks’, and introduce image fusion technology and 3D convolutional neural networks (3D CNNs) into HSI classification. In the proposed method, the fusion process is guided by the classification result, and the classification accuracy is improved by the fusion process. Image fusion technology is performed on spectral bands to exploit the redundancy information of HSI, and 3D CNNs are applied on the fused image to extract more robust spectral-spatial features. The proposed method is tested on two datasets. Its outstanding performance is validated in comparison with other state-of-the-art approaches.

Traffic flow has the characteristics of randomness, nonlinearity and periodicity. Using deep learning as the method of traffic flow prediction can make full use of traffic flow characteristics for prediction. In this paper, a recurrent neural network-LSTM (Long Short-Term Memory) model and its variant GRU (Gated Recurrent Units) are used to solve the traffic flow prediction problem. According to the road traffic information collected by the automatic measurement station (LAM) of the Dutch traffic management department, the vehicle traffic flow data has been predicted every five minutes. It’s verified that the GRU traffic flow prediction model performs well by comparing the prediction accuracy of LSTM and GRU.

Feature selection (FS) plays an important role in the machine learning (ML) field. Since FS solves the problem of dimensional explosion in ML very well, more and more people are paying attention to FS. Not only that, but this technique also takes advantage of the computational complexities and time reductions. Inspired by the points mentioned above, more and more FS algorithms solved by deep learning framework are appearing. Due to the importance of FS, it is necessary to conduct further research. However, FS is wide coverage, and the algorithms involved are numerous, which makes researchers need to spend a lot of time searching and reading the literature. In order to provide researchers with dedicated information and enable them to quickly have an overall understanding of the FS field, this article will from three aspects, including the main functions and framework of FS, search strategies of FS, and the evaluation strategy and algorithms in related fields to introduce FS from whole to part. Finally, this article discusses some existing problems and points out some promising research directions.

In this paper, a handwritten digit recognition system based on K-nearest neighbor algorithm is designed and implemented. By analyzing the application principle and steps of k-nearest neighbor algorithm in machine learning classifier, we design handwritten digit recognition in Python language. In order to improve the accuracy and accuracy of handwritten character recognition, the method of smooth denoising and character segmentation methods are used to preprocess the data, extract the structure and statistical features. The experiments had shown that the accuracy of the algorithm for handwritten character recognition can reach over 97%.

This paper proposes a method to color black & white portrait images with gender recognition model and condition generation network (cGan). As a cGan model essentially, a method named pix2pix interprets image style transformation as a translation process from input pixels to target pixels. Lots of input to target image pairs should be prepared for model training for different application scenarios. When pix2pix is applied to color black & white portraits, there are obvious differences in color and saturation between portraits of different genders while the ability of gender recognition isn’t owned by pix2pix in essence. Therefore, a gender recognition model is added to the network to produce more realistic and gender specific color portraits. The experiments show that our work has achieved better results than some other common methods.

At first, this paper reviews the present situation and some developing trends of the telemetry system in flight test. Then the paper introduces the new requirements of the telemetry system brought by the development of flight test assessment. At last, this paper shows some considerations in developing the telemetry system, including those on the architecture of network telemetry system, radio frequency resource, the control of telemetry data stream, band efficiency and coding and error control, and RF environment and classification.

Most visitor systems’ information management is at a low level, and the processing method is still semi-manual registration. To improve the level of automation of the visitor system, this paper designs a remote visitor system based on NFC technology deployed on the Android smart-phone platform. It reads and emulates card data through NFC technology and adopts SM series algorithm to encrypt and decrypt the transmitted data. The system effectively solves the problems about security and automation in the traditional guest system.

In this study, we propose a bit slotted ALOHA algorithm based on a polling mechanism (PBSA) to address the tag collision problem of dynamic radio frequency identification (RFID) systems. This algorithm is based on a combination of the dynamic framed slotted ALOHA (DFSA) algorithm and the free framed slotted ALOHA (FSA) algorithm. The polling mechanism is introduced to change the tag removal protocol. While removing the idle slots, the collision slot is converted into a single slot, which improves the tag recognition efficiency per unit frame. Simulation analysis shows that compared with the dynamic framed slotted ALOHA (DFSA) algorithm, the grouped dynamic framed slotted ALOHA (GDFSA) algorithm, and the cancel idle framed slotted ALOHA (CIFSA) algorithm, the PBSA algorithm exhibits lower slot overhead, higher tag throughput, and lower lag loss rate, which improves the efficiency of tag identification in dynamic RFID systems.

With the development of the Internet and the change of consumption concept, e-commerce platform has become the primary way of shopping. In this case, personalized recommendation system has become the main way to find favorite businesses, and also become the main means to sell goods on the platform. n the recommendation system of e-commerce platform, collaborative filtering recommendation algorithm is the most widely used. The main idea of the algorithm is to find the similar neighbor set of the target user, and recommend the user preference of the neighbor set to the target user as a recommendation.In this paper, a collaborative filtering recommendation algorithm based on context information is proposed, which focuses on the influence of user context information on the similarity calculation between users in the recommendation algorithm in the mobile Internet environment. The algorithm takes the user location information and user trust relationship as the key factors. First, the algorithm preprocesses the original score data through SVD to alleviate the scarcity of data, then generates a list of merchant candidate recommendations based on the score similarity of users Using collaborative filtering. And then calculates the travel cost of each merchant in the candidate recommendation list to the target user according to the location context, and adjusts the recommendation weight of the merchant in the list, Finally, the final Top-N merchants are generated to recommend to the target users.In the end of this paper, through the experimental simulation of the proposed algorithm, and compared with the traditional collaborative filtering recommendation algorithm, it is proved that the algorithm further improves the effectiveness and accuracy of the recommendation algorithm.

The development of optimized miniaturize and powerful electronic components keep the strain on engineers to produce optimal cooling designs for better results. One method for cooling these electronic components is with heat sinks which effectively increase the surface area available for extracting the heat from the electronic components. In this paper, a methodology is developed where plane fitting methods are used to make boundary planes that make the boundary for the design of heat sink. This paper helps engineers and researchers with less knowledge of thermal analysis to design a proper heat sink.

Dissolved gas analysis (DGA) is an effective method for oil-immersed transformer fault diagnosis. This paper proposes a transformer fault diagnosis method based on Stacked Contractive Auto-Encoder Network (SCAEN), which can detect the transformer’s internal fault by using DGA data, including H2, CH4, C2H2, C2H4, C2H6. The network consists of a three-layer stacked contractive auto-encoder (SCAE) and a backpropagation neural network (BPNN) with three hidden layers. A large amount of unlabeled data is used to train to obtain initialization parameters, and then a limited labeled dataset is used to fine-tune and classify the faults of trans-formers. The proposed method is suitable for transformer fault diagnosis scenarios, which contains very limited labeled data. when tested on real DGA dataset, the fault diagnosis accuracy is up to 95.31% by SCAEN, which performs better than other commonly used models such as support vector machine (SVM), BPNN, auto-encoder (AE), contractive auto-encoder (CAE) and SCAE.

At present, electric vehicles have a general problem of short battery life and short travel distance. Therefore, introducing supercapacitors and DC/DC converters to form a hybrid energy storage system (HESS) with the battery to make up for the lack of pure battery energy. In this paper, a control strategy for an on-board hybrid energy storage system based on PI control is proposed. First of all, This strategy is through energy management strategy to distribute the required power, and then by changing the duty cycle of the IGBT switch in the DC-DC converter, to control the battery and supercapacitor output currents and to keep the DC bus voltage stable. The simulation results on the build simulation model, Simulink, shows that the proposed control method can accurately track the reference values of battery current and supercapacitor current and stabilize the DC bus voltage, which proves the effectiveness of the control strategy.

Network threats caused by system vulnerabilities is increasing gradually, the distributed vulnerability scanning system can scan large-scale and complex networks and report the vulnerability information. Task scheduling is one of the core components in a distributed system. In this paper, we use dynamic optimization algorithms to improve the task scheduling efficiency of distributed vulnerability scanning system, we propose a PSO-based task scheduling scheme and improves the search ability of particles by adjusting algorithm parameters. We compared the time consume when using existing ‘Resource Aware Scheduling algorithm’ (RASA) with the basic particle swarm optimization (PSO) algorithm and the improved particle swarm optimization (IPSO) algorithm. Our results show that IPSO has better performance than other scheduling methods.

The survey found that during table tennis training there are often many scattered balls that need to be picked up manually, affecting the efficiency of the players’ training. Currently, table tennis is picked up manually using a table tennis picker. In this paper, delta omnidirectional wheeled table tennis automatic pickup robot based on the vision servo achieves fully automatic table tennis pickup by combining the characteristics and requirements of table tennis sports. In this paper, we design a vision algorithm for a table tennis automatic pickup robot system, give a kinematic inverse solution for the delta omnidirectional wheel, complete the design of the robot control system, and build the platform for the experiment. A pickup experiment is conducted on scattered ping-pong balls of different colors. Experiments have shown that the robot can safely clean up scattered table tennis balls on the table tennis field with high positioning accuracy and top pick up rate. It is proved that the robot designed in this paper has extensive application value and prospects.

Radiosonde has always played a very important role in meteorological detection, so how to properly schedule the radiosonde to reach the sensitive region is an urgent problem. In this paper, deep learning is applied to this field for the first time to provide a basis for the reasonable scheduling of radiosonde by predicting the motion trajectory of radiosonde. Based on the radiosonde data from February 2019 to October 2019, this paper uses the radiosonde trajectory prediction model based on GRU (RTP-GRU) to predict the radiosonde trajectory in a period of time in the future. The experimental results show that this model has better performance than baseline methods such as RNN and LSTM. The results show that it is feasible and valuable to explore this field with deep learning method.

Proton exchange membrane fuel cell (PEMFC) has the advantages of stability and high efficiency, but it has little life problems. Therefore, through life prediction and other work, we can know the fuel cell’s health status in time and promote its development. In this paper, aiming at the problem of life prediction, a method based on MLP and locally weighted scatterplot smoothing are proposed, which can reduce the data and improve the prediction efficiency by reconstructing the data, can not only retain the original trend of the data but also remove the noise and peak. Finally, the processed data is applied to the MLP model for life prediction. Experiments show that accuracy can reach more than 96%.

Aiming at the classification of Red Fuji apples based on the content of soluble solids, a multi-classifier model fusion method based on Dempster and Shagfer (DS) evidence theory is proposed. Multi-model fusion is carried out by integrating different modeling methods such as Partial Least Squares (PLS) and Extreme Learning Machine (ELM). This method uses the prediction results of each model to construct the DS probability distribution function, and outputs the final fusion result through the DS fusion rule. Experimental results show that, compared with a single model, the designed multi-model fusion can improve the accuracy of apple classification to a certain extent, which is conducive to improving the efficiency of apple classification.

Few existing online car-hailing scheduling algorithms simultaneously take into consideration benefits and fairness. In this paper, a weighted sum of platform profit, driver income and passenger waiting time was used as the objective function of the assignment model. The improved Hungarian algorithm was employed to solve the problem. The imbalanced assignment problem was converted into a standard assignment problem by a virtual person (car) method. The interests of the platform, driver, and passenger were balanced through weight coefficients. Three groups of experiments have verified that compared with the global search algorithm and the greedy algorithm, the performance of the results obtained by the presented algorithm has a great improvement, and the solution time complexity is lower than the global search algorithm.

In this paper, a minimum adjustment method is proposed to solve the shortest time assignment problem. The algorithm can meet practical needs, but the complexity is high. The paper improves the minimum adjustment algorithm, the new algorithm can take into account the requirements of the shortest time and the shortest total time simultaneously, and its solving process is efficient and straightforward. An example is used to verify the effectiveness of the algorithm.

To resolve the difficulty of sampling real-time data when debugging algorithm on digital signal processing equipment, a board based on FPGA which can transmit data from PCIe to fiber is designed. Computer sends data to DDR3 on the designed board through PCIe. The ARM in FPGA sends data to the PL of the chip through DMA. The PL sends data to digital signal processing equipment through fiber after ping-pang operation in the PL. The digital signal processing equipment exchanges data through fiber with the developed board. The digital signal processing equipment can use the data receiving from the board developing the processing algorithm. The experiment shows that the designed board can meet the simulation requirement when developing the digital signal processing algorithm and can improve the developing efficiency.

This paper designs and implements a LAN-based IP network performance monitoring and comprehensive analysis system. Aiming at the specific network architecture and environment of LAN, the system has three functional modules: network environment monitoring, service flow data capture analysis, and network failure analysis. When network failure occurs, the system can analyze the network status comprehensively to implement the initial judgment and positioning of the fault, which provides a convenient network management method for network maintenance and management personnel.

Compared with unicast and broadcast, multicast communication technology can save network bandwidth under specific communication conditions. But there are some problems appear with the growing application, the analysis of multicast RP election and packet loss during RPT switching to SPT process are introduced, three solutions for this kind of problem are proposed. Because of the special network environment and transmission mode of communication private network, manual configuration method that disables SPT switching has failed, and the introduction method for static multicast group has disadvantages, finally, the RP point is selected from the dynamic pointing converging switch to the dynamic pointing the core switch. It is proved by experiment that the method is effective.

The cross polarization jamming is an novel jamming technology with good effect on conventional monopulse radars. By theoretical analysis, the angular measurement error can reach the half-beam width. However, the polarimetric array fusion monopulse radars can obtain the angular measurements through two polarization arrays, work out the maximum likelihood estimation of the target angle to achieve the final angular. In this way, the radar can have its anti-jamming capability improved. In this paper, a novel model has been proposed for polarization domain countermeasures. The countermeasure performance of cross polarization jamming and polarization fusion monopulse radars has been analyzed based on computer simulation. The results show that the polarization monopulse radar is insensitive to cross polarization jamming and the jamming effect is not ideal, so it is necessary to design pertinent jamming.

Moodle is based on the theory of constructivism, it is used for construction of internet-based courses and the package of web site’s open source code. Moodle supports diversified ways of teaching and learning. It can create a good network learning environment for learners’ autonomous learning, collaborative learning and personalized learning. It also can bring convenience for teachers and school teaching. In view of the advantages of Moodle system, we select it as a platform for the construction, development and application of the college of computer science and technology subject. At first, this paper used the literature research to understand the development idea and the function characteristics of Moodle, and analyzes the architecture of Moodle and related technologies, especially the required environment, conventions and ways of the second exploitation of Moodle theme. It shows the difficulty of the development of each way. Then redefining professional subdirectory according to different requirements of activities module and function module.We designed a typical of college computer science and technology of Moodle theme taking into account from the overall style, color system, font and layout, fresh graduates courses and graduate thesis submit. Finally summarizes the main work of development theme and the problems still needing to improve, equally looks forward to the future work.

Community correction refers to the correction on the five types of objects. Community correction work is to correct the community correction objects’ psychology and behavioral habits by taking proper correction measures for corresponding correction objects, promote them to return to society. Selection of correction measures is a key component in community correction work. However, this work is still at the manual stage in China, which consumes much manpower and resources. To this end, we propose an algorithm namely FFM-FTRL which is the first time that machine learning is applied to addressing the recommendation of community correction measures, improving the efficiency of community correction work. Our method is specifically designed to solve sparse data, and reduce model training time. To evaluate its effectiveness, we select FM and FFM-SGD as our baselines. Experimental results show that F1 score is improved by 4% and 0.5%, and the training time is shortened.

In recent years, with the rapid development of information technology, artificial intelligence and Internet of things technology, the information construction of colleges and universities in China is also accelerating. This study investigates the current situation of the information management and service of nine colleges and universities in Xianyang City, Western China. From the three aspects of information resource construction, management and service, it understands the construction of campus network in the third-tier city in Western China, the construction and application of multimedia network classroom, network system management, application platform management, teaching system management, user service management, learning resources and basic services etc. In view of the shortcomings in management and service, some suggestions for improvement are put forward.

Aiming at the problem that the current network key node positioning effect is not good, the network key node positioning method is optimized and researched by combining the position weight matrix, the characteristic values of the network key node positioning are collected and analyzed by combining the characteristic collection principle, and the distribution area of the network nodes is selected and accurately positioned according to the characteristic collection result. Finally, the experiment proves that the network key node location method based on the position weight matrix has high accuracy and stability, and fully meets the research requirements.

Nowadays, people are increasingly using mobile terminals. In many cases, they are inclined to use E-notes software in their mobile phones to record some simply information. Thus, I developed this E-Note software which named A-Note to suit their requirements. This software was developed by Android Studio 2.3 from google in Java.Also, I use My Eclipse to manage data in the server. In this version, I just implement a simple UI and two basic ways for note inputting. All what I do on A-Note just to form a simple method for recoding online and to make few contribution for the developing of E-Notes software.

At present, user researchers have problems in UX (User experience) data analysis, such as low efficiency, inaccurate context data identification, and low satisfaction of analysis process. Therefore, in order to solve these problems, this paper proposes a design context UX data analysis system to compensate for the shortcomings in the data analysis process. This paper takes the analysis of UX data collected by the CAUX (Context-Aware User Experience) tool as an example, using the relevant methods in Cognitive Task Analysis (CTA), and on the basis of sensemaking loop model, explore the data analysis process of UX researchers through experiments. And carry out demand research for each stage of the analysis process, design a context UX data analysis system according to the requirements. This thesis summarizes the model of UX data analysis process, completes the design of context UX data analysis system, and evaluation experiment proves that the system can effectively solve the problems in the UX data analysis process, and provides a new idea for the UX research practice in the mobile Internet environment.

Untrustworthy behaviors such as false redundant information flooding and malicious recommendation hinder the healthy development of social e-commerce, and predicting and evaluating the credibility of users is of great significance to social e-commerce marketing promotion. Giving objective and appropriate evaluation weights to decision attributes is the first step to predict the credibility of users. This paper proposes a dynamic trust quantification model based on entropy weight method, which can realize adaptive weight setting. Experiments show that compared with similar models, the model is more adaptable and has a higher transaction success rate.

In order to fetch up the shortages of default consensus mechanism of Ethereum, this paper proposes a practical Byzantine fault-tolerant consensus algorithm based on the scoring mechanism. In view of the above two defects of PBFT, this paper proposes a Scoring algorithm based on PBFT (SPBFT). The proposed algorithm takes the advantages of high expansion and high reliability on the basis of solving the Byzantine general problem. Experiments show that our proposed improvement algorithm can not only meet the dynamic scalability of nodes, but also ensure that elected master node in system is more reliable.

In power communication networks, in order to solve the problem of low user satisfaction caused by network link failures, this paper proposes a network topology-aware power communication network link failure recovery algorithm. The algorithm includes three parts: solving the number of services affected by the faulty link, calculating the recovery value of the faulty link, and performing fault recovery based on the recovery value of the faulty link. In terms of the number of services affected by the faulty link, the shortest route used by the power service is used as a criterion to calculate the number of services related to each faulty link. In terms of calculating the recovery value of the failed link, the value of the recovery of the failed link is calculated from four aspects: the value of the traffic and the value of the service traffic brought by the recovery of the failed link, the centrality of the failed link, and the resource overhead of the recovery of the failed link. In the simulation experiment part, the algorithm in this paper is compared with the traditional algorithm, which verifies that the algorithm in this paper has better performance in terms of user satisfaction.

Network slicing is an important technology for implementing on-demand networking based on the 5G network architecture of SDN/NFV. By analyzing the main scenarios of 5G, a network slice orchestration algorithm based on GA-PSO optimization under the SDN/NFV architecture is proposed. This algorithm uses the particle swarm optimization algorithm to quickly converge on the characteristics of the global optimal solution, and design the evaluation function of network slice performance. Moreover, the ability of genetic algorithm to quickly search randomly is used to update and optimize the network slice, and the particle swarm is used to chase the local optimal solution and the global optimal solution to obtain the optimal network slice. Simulation experiment results show that the algorithm can realize the personalized creation of network slices in multi-service scenarios, give full play to the advantages of SDN’s centralized control mode, and reduce network energy consumption while improving network resource utilization.

In the future heterogeneous integrated wireless network environment, delay limited network has the characteristics of distributed self-organization, multi hop transmission, delay tolerance and intermittent link connection. Although it has great application value in many scenarios, it still needs to provide certain QoS guarantee according to the business requirements. Therefore, how to effectively transmit information to meet the needs of different services in terms of transmission rate and delay is one of the key issues in the research of path allocation in delay constrained networks. In this paper, a path allocation method based on service priority is proposed. Network nodes use local information to evaluate the network state and calculate the initial service priority, and then select relay nodes according to congestion degree and encounter probability. Simulation results show that the algorithm improves the quality of service of high priority services in congestion, and provides high performance services for low priority services when the network resources are sufficient.

The trunked system uses many computer nodes in close collaboration to carry a large number of high-performance computing applications, and guarantees strict QoS (Quality of Service) requirements through distribution and backup mechanism. Because of the cooperative and interactive nature of the trunked system, the failure of one node may lead to the failure of other nodes. The system log of the cluster is a very important failure prediction resource, which can use the detailed information of the system running condition recorded in the log to track the system behavior for a long time. At present, three steps mainly affect the performance of the trunked system failure prediction method: event filtering, event mining and event prediction. In the event prediction stage, the mined causal relationship is used for reasonable reasoning to achieve the purpose of fault prediction. However, the prediction method based on rule reasoning needs to scan the database for many times, which costs a lot of time in the pattern matching stage, and the reasoning efficiency is low. Aiming at the above problems, an abstract Weighted Causal Dependency Graph (WCDG) is designed to represent the event rules, and the forward uncertainty reasoning based on the Weighted Causal Graph can achieve the purpose of fault prediction. Compared with other prediction models, this method can store and update event rules more easily and meet the real-time requirement of the system.

With wide use of internet, the problem of anomaly detection in network security has become an import issue of concern. In order to solve the problem, this paper proposes a new method FEN-TLSAE which uses transfer learning to detect network traffic when plenty of labeled data are unavailable. First, we use feature extraction to get two feature subsets and use two different networks to train them. Then we add BN and dropout to the improved autoencoder for regularization. Finally, we use joint inter-class and inter-domain distributional adaptation (JCDA) in the process of transfer learning. Minimizing the marginal and conditional distribution distance between the source and target domains while maximizing the distribution distance of samples between different classes in the source domain. The experiments on the NSL-KDD data set indicate that the proposed method is more efficient than that based on TLSAE. Also, the method has higher detection accuracy, recall and precision.

Multiple-set membership query problem plays a very important role in many network systems, including packet forwarding architectures in big data centers, routing switching devices and internet firewall systems. Many solutions to multiple-set membership query problem are based on Bloom filters. And each of these solutions has its own merits. Some have high query speed and some have good accuracy. But generally, they cannot have both at the same time. In this paper, we propose a probabilistic data structure named constant-weight group coded Bloom Filter (CWGCBF) for fast and accurate multi-set membership query. The key technique in CWGCBF is encoding each element’s set ID to constant weight code, and directly recording this constant-weight code in the bloom filter vector. Experimental results show that in terms of false positive ratio, CWGCBF, Shifting Bloom Filter and Combinatorial Bloom Filter have great advantages to the other compared Bloom filters, and in terms of time efficiency, CWGCBF and ID Bloom Filter are several times faster than the others.

Digital image has the characteristics of high redundancy and convenient processing, which is the ideal carrier of information hiding application. Therefore, digital image steganography has become a hot research direction in the field of information security. In order to ensure the concealment, reliability and security of image steganographic information, an image adaptive steganographic algorithm combining chaotic encryption and minimum distortion function is designed. The algorithm generates random keys, encrypt the secret information using Logistic and ChebyShev mapping, and then embeds the key and the encrypted secret information into the carrier image using the HILL steganographic algorithm. The algorithm stores the embedded key and the decryption key separately, so the attacker cannot get the correct secret information even if he gets the embedded key, which improves the security of the key. Experiments show that the pixel change rate of the algorithm is 6.76%, the peak signal-to-noise ratio is 59.51 db, and it has a very good anti-steganographic analysis ability.

In this paper, the assets, threats, and vulnerabilities of substation power monitoring systems are identified and assigned according to the relevant requirements of national information security risk assessment, and the assignment is optimized in combination with threat and vulnerability. Finally, a calculation method of asset loss is proposed.

With the development of technology, FTP protocol has been widely used but also brought many threats and hidden dangers, including remote attacks, denial of service and so on. Aiming at the above problems, this study develops a vulnerability mining and evaluation technique for FTP protocol based on Fuzzing. This project use Python programming to implement Fuzzing. Using freefloatftp software to build a simple FTP server in Windows XP SP3 operating system. Ftpfuzzer.py is used to attack the server. Besides, Immunity Debugger is used to preliminarily analyze the vulnerabilities mined, so as to provide direction for resisting attacks.

With the rapid development of Internet technology, various network attack methods come out one after the other. SQL injection has become one of the most severe threats to Web applications and seriously threatens various Web application services and users’ data security. There are both traditional detection methods and emerging methods based on deep learning technology with higher detection accuracy for the detection of SQL injection. However, they are all for detecting a single statement and cannot determine the stage of the attack. To further improve the effect of SQL injection detection, this paper proposes an integrated detection framework for SQL injection behavior based on both text features and traffic features. We propose a SQL-LSTM model based on deep learning technology as the detection model at the text features level. Meanwhile, the features of the data traffic are merged. By this integrated method, the detection effect of SQL injection is further improved.

With the rapid development of mobile Internet, the methods of data storage have been greatly enriched. However, many problems exist in mobile internets' current information storage, such as plain-text storage, weak and nondomestic encryption algorithms. This paper designs a secure cloud storage platform with SM algorithms based on the Android platform. The platform uses a combination of encryption and authentication methods that use multiple SM algorithms (SM2, SM3, SM4), which realizes hybrid encryption and authentication of information storage. Finally, lots of experiments are done, and the results show the security and reliability of our platform, the time about regular operations of our platform is in the range of acceptance given performance and security.

In recent years, with the rapid development of information technology, more and more attention has been paid to information security. Biometrics is an important way to ensure information security, and digital vein recognition is an important branch of biometrics, which has an irreplaceable position in the field of information security. However, the traditional finger vein recognition is vulnerable to attacks such as replay, template tampering, and forging original features, which leads to the destruction of the authentication system. This paper proposes a solution to encrypt the feature points of finger vein with the key generated by the relative difference of fingerprint detail points based on the SM4 algorithm to realize the double encryption authentication of user information, so as to improve the security and accuracy of user identity identification and meet the increasing requirements of high security.

For chaotic encryption algorithm, the algorithm of generating chaotic sequence is complex, and the data is float, which directly affects the speed of encryption. In this paper, an interlaced chaotic encryption algorithm is proposed, In this algorithm, the pixel interval is encrypted by chaotic sequence, and the chaotic sequences of other pixels are obtained by XOR the chaotic values of the front and back points, Experiments show that the algorithm can encrypt and decrypt images quickly and has good effect.

Based on the monotone span program, this paper proposes a verifiable multi-secret sharing scheme on the general access structure. Each participant can choose their own secret share and use the RSA cryptosystem to send the share to the dealer through open channels. In the secret recovery phase, the scheme uses the Hash Function to achieve multi-use property and verifiability. Analysis shows that each participant can recover multiple secrets by keeping only one secret share, and shares do not need to be changed when the multi-secrets are renewed. Furthermore, the distribution of secret shares in the scheme does not require a secure channel, which effectively reduces the computational cost of the system. Compared with the (t, n) threshold secret sharing, the scheme has more vivid access structures.

This paper refers to a hardware implementation for executing modular multiplication in public-key cryptosystems using the Barrett reduction, which is a method of reducing a number modulo another number without the use of any division. Considering the flexibility of hardware, the modular multiplier we proposed is able to work over 3 prime fields $$ {\text{GF}}\left( p \right) $$ , standardized by NIST for use in Elliptic Curve Cryptography (ECC), where the size of primes $$ p $$ are 256, 384, and 521 bits. We designed two methods to optimize the modular multiplier: Firstly, the circuit departed $$ 257 \times 257 $$ multiplier into two $$ 257 \times 129 $$ multipliers and an adder, three parts to optimize for clock frequency. Secondly, we proposed a parallel computing architecture to improve the utilization of multiplier and achieve high throughput. This modular multiplier runs at the clock rate of 300 MHz on 40 nm CMOS and performs a 256-bit modular multiplication in 3 cycles, while 384-bit modular multiplication costs 10 cycles and 521-bit modular multiplication costs 25 cycles. The architecture is very suitable for situations requiring high computing speed, such as online ECC signature verification.

Collision attack is an effective method in the field of side-channel analysis to crack cryptographic algorithms, and masking can be used as a countermeasure. Most collision attacks only utilize the traces that will collide. In this paper, we propose a collision attack method that exploits not only traces tending to collide, but also non-colliding traces. It can bring higher efficiency and reduce the number of needed traces significantly. In addition, our method is a random-plaintext collision attack method instead of a chosen-plaintext attack. The experimental results show that our proposed approach is better than the existing collision-correlation attack proposed by Clavier et al. at CHES 2011 [11]. To achieve a high key recovery success rate at 80%, we use at least 60% less traces than collision-correlation attack.

A novel digital image encryption scheme based on Poker shuffling and fractional order hyperchaotic system is proposed. Poker intercrossing operation has nonlinearity and periodicity, and can form permutation group. Firstly, the plain image is transformed into block image, and each block is confused by Poker shuffling operation with key. Secondly, bit plane of every pixel in each block is confused. Lastly, the scrambled image is encrypted by fractional order hyperchaotic sequence. This encryption scheme provides a secure and efficient key stream, and guarantees a large key space. The experimental results and security analysis imply that the new encryption method has secure encryption effect and property of resisting common attacks.

The computing capability of micro encryption devices has been a concerned problem by the people. This paper improves the implementation of LILLIPUT which was proposed in 2016. In this paper, we adopt the method of compromising area and power consumption. And with the utilization of parallel hardware processing and reduced excess hardware, the execution speed of the algorithm is extended with the minimal power consumption. Tested by experiments, the third-order LILLIPUT algorithm of parallel processing is 40% higher than that of the non-parallel processing, with only a small amount of area and power consumption used to significantly improve the computing performance of the micro-equipment.

Indoor positioning system (IPS) provides fundamental location-based service in indoor environment where GPS signal cannot reach. Different from WiFi, Bluetooth and other wireless signal based IPSs, visible light positioning (VLP) has been studied in recent years due to its high accuracy and robustness, and free from electromagnetic interference. Therefore, VLP is a promising technique in indoor public places such as hospital, airport, and shopping center. To achieve higher performance, deep learning has been widely investigated for VLP for its high tolerance to noise and studies have shown significant improvement in challenging environment compared to traditional algorithm such as least square. Unfortunately, we found that deep learning, or network-based algorithms are vulnerable to device heterogeneity and setup deviations for different users and environments, thus trained model may have poor performance in test stage. Besides, feature-label pair is often unavailable during testing for further model reinforcement. In this paper, an unsupervised learning method based on adversarial training is proposed for visible light network regressor. In this method, a feature extractor works to align feature distributions from different domains, while a domain classifier differentiates data from training set and test set. Results show that when multi-dimensional attack exists, our adversarial network model can achieve significant improvement than directly applying a trained model to the test set.

Since the current Chinese spam messages often contain artificial adversarial perturbation and can bypass the traditional detection system, this paper proposed an integrated model based on “ALBERT-Capsule Network.”On the one hand, the model accepts both raw text and word segmentation text as input data, uses ALBERT and capsule networks to extract character-level features and word-level features respectively, and uses a multi-head output structure to ensure that both the word feature stream and the word feature stream can be meaningfully trained to prevent the neural network from depending on one of its branches. Experiments on the SMS spam dataset and trec06c datasets prove that the model is better than the traditional single-feature models such as Bi-LSTM and TextCNN; On the other hand, the traditional character replacement, pinyin replacement and word splitting and other adversarial methods are used to augment the original training dataset, which further enhances the robustness of the model, so that the model can effectively detect spam messages containing adversarial perturbation.

The LAN is often equipped with web servers. At this time, the LAN domain name resolution should be able to access the internal website and be easy to conFig without relying on the LAN export equipment. Based on the DNS protocol, the conventional method is to use dual-domain names, or use a dedicated DNS server, or use bidirectional NAT. These methods are insufficient. Domain name hijacking is a kind of hacker technology, but it can be used to apply domain name resolution in LAN properly. Hijacking can be realized through the DNAT function of net-work devices such as a firewall. The internal website’s domain name can be resolved to the IP address of the local area network by the dedicated DNS server, and other domain name resolution requests can be forwarded to the external DNS server. The method has the advantages of convenience, reducing a load of network exit equipment and no increase in cost.

This paper takes the electric power information network as the model of research, studies the captured data packets and the characteristics of data flow, analyzes the specific characteristics of the audit data from the angle of anomalynetwork traffic, and combines statistical variance method and regression analysis method to put forward an intrusion detection system on bases of distributed Network flow.

In order to solve the problems of low authentication efficiency and easy tampering of authentication data in the electric IoT system, this paper proposes the electric IoT based on blockchain. First, the authentication model based on blockchain is proposed, and four key issues are analyzed, such as data security storage, user biometrics extraction, power terminal equipment feature extraction, and authentication key generation, then technical solutions are proposed. Secondly, based on the authentication model, the power terminal registration process, user registration process, and authentication process are proposed. The performance analysis of the authentication mechanism is carried out from four aspects: availability, integrity, privacy and efficiency. The analysis results show that the authentication mechanism proposed in this paper has good performance. Finally, the application scenarios and precautions of the authentication mechanism are analyzed.

In order to solve the security problems existing when power users access power services in different domains, this paper proposes a dynamic authorization model for power users based on master-slave chain. In this model, the slave Blockchain conducts identity authentication in each autonomous domain, and the main Blockchain undertakes identity authentication among autonomous domains. Moreover, by constructing an attribute access control model, a trusted identity cross-domain authorization mechanism is proposed, and an attribute-based allocation strategy in both single-domain and cross-domain scenarios is designed in detail. Through application analysis and security analysis, it is verified that the trusted identity cross-domain authorization mechanism proposed in this paper can be conveniently applied to existing systems, and has better performance in confidentiality, integrity, and availability.

In order to solve the problem of impersonation of the electronic identity of power maintenance personnel, this paper proposes an trusted identity authentication model based on Blockchain. This model uses plug-in authentication technology to improve the scalability of the identity authentication system. In order to improve the usability of the authentication model, smart devices are named uniformly, and information such as names and corresponding parameters are registered on the distributed data trusted chain through smart contracts. In order to improve the practicability of the model, based on the model architecture, the registration process, login authentication process, and trusted identity card system architecture were designed. In terms of the architecture of the trusted identity system, it includes a Blockchain server, an authentication server, and a mobile terminal. Through the performance analysis, the authentication mechanism proposed in this paper is verified, which effectively solves the problems of poor reliability and centralized storage that easily lead to information leakage in traditional identity authentication.

In the context of the rapid increase of the number of faults in power data communication networks, in order to ensure the reliability of power data communication services, how to quickly recover the affected power data communication services has become an urgent problem to be solved, when the underlying network fails. In order to solve this problem, this paper first analyzes the two indicators of the affected service from the availability of power service and the return of the power service, and the three indicators of node reliability from the node degree, the node centrality, and the proximity between the nodes. Secondly, the entropy weight method is used to calculate the index weights, and the objective evaluation of each index weight is realized. Finally, the fault recovery algorithm of power data communication network based on node reliability is proposed. In the simulation experiment, compared with the traditional algorithm, the algorithm has achieved good results in terms of fault recovery rate and power data communication network revenue.

In order to solve the problems of high false alarm rate and long inference time of the congestion link inference algorithm in power data networks, this paper proposes a congestion link inference algorithm of power data network based on Bayes theory. First, based on the network topology and detection relationship, a detection matrix is constructed, and the detection matrix is simplified using Gaussian Jordan elimination method. In order to reasonably select the detection, the detection information gain is designed to make the judgment. Secondly, in order to accurately infer congested links, a Bayesian model for detecting link associations is constructed based on the detection results, and based on the model, congested link inference is performed using the maximum posterior probability. Compared with the existing algorithms, it is verified that the algorithm in this paper has achieved good results in three aspects: the accuracy of congested link inference, the rate of false positives, and the length of inference.

With the rapid popularization of mobile devices, power grid system relies on mobile devices more and more. The traditional identity authentication cannot meet the security requirements of power grid system. Although continuous authentication can be achieved through behavioral features, there is still a problem of insufficient authentication accuracy. This paper proposes a method of fusion of gait and touch behavior. The fusion level is match level fusion. Weighted addition fusion is used to fuse gait matching score and touch matching score to improve the accuracy of authentication.

With the wide application of network attached storage (NAS), the security problem is becoming more and more serious, together with the increasingly fierce attacks against NAS devices of different manufacturers. In order to better capture various types of attacks against NAS devices and detect security threats in time, this paper proposes a solution named NAS honeypot, mining the potential security threats through modeling and analyzing the NAS threats. Then a NAS honeypot based on device virtualization technology and depth monitoring and attack induction was designed on the basis of the construction NAS attack chain. Experimental results show that the NAS honeypot can effectively capture, record and analyze network attacks against multi-types NAS devices with a strong guiding effect in mastering popular attack method of NAS devices and alleviating the security threats of NAS.

As the value of the big data is revealed, the problem of security prevention and control of customer service data is highlighted. Once the leakage of sensitive customer information will cause serious legal consequences. At present, State Grid Corporation has effectively resisted the risk of data collection, storage, and transmission through internal and external network isolation and data encryption. However, at the data application level, due to frequent data sharing and business collaboration, new risks of leakage continue to emerge in data mining, data interaction between systems, terminal data access, etc. This article first analyzes the risks faced by data security in terminal data access, cross-domain data transfer, and open data sharing. Then, it proposes terminal data access behavior management and control technology, data cross-domain transfer security interaction technology, and data differentiated privacy protection technology in open sharing. Finally, t on the basis of technical research, the core equipment of data protection was developed and tested, and verified the effectiveness of the technology in practice.

In order to solve the problem of low accuracy of fault diagnosis in network slicing environment, this paper proposes an efficient fault diagnosis algorithm based on active detection. By calculating the business correlation of the underlying network resources, a set of candidate detection nodes is formed. Based on the overlapping characteristics of the detection paths, the set of detection nodes is optimized. Active detection technology is used to actively acquire network characteristics, and a suspected fault set is constructed based on the historical fault probability and detection performance, and the fault is quickly diagnosed through the credibility evaluation of the fault set. By comparing the performance with existing algorithms, it is verified that the algorithm in this paper improves the accuracy of fault diagnosis.

It is very necessary to carry out simulation research on the naval surface ships of the world’s powerful army because of their strong combat effectiveness. At present, it is a scientific and effective method to analyze and study the opponents to build the Blue Force simulation system and carry out the virtual combat between Red Force and Blue Force based on the computer network. On the basis of summarizing the current situation of Blue Force simulation system of naval surface ships, this paper expounds the construction method of Blue Force simulation system from three aspects of system function analysis, structure design and running process. It is considered that: in the aspect of function analysis, it is necessary to face the joint operation simulation, access the real army and real equipment to participate in the military exercise; in the aspect of structure design, it is necessary to highlight the fleet-level computer generated forces system, pay attention to the cooperative operation between ships; in the aspect of running process, it is necessary to use new means such as large data technology to fully analyze the key nodes in the running process. The research conclusion can provide ideas and basis for the construction of Blue Force, and can be extended to other combat simulation system construction.

Resource sharing in a distributed environment can improve the utilization of resources. To promote the benign development of resource sharing behavior, a basic evaluation model of resource sharing is proposed, and an evaluation index system of distributed server resource sharing is constructed based on the model in this paper. The judgment matrices are constructed according to the analytic hierarchy process (AHP), and the simulated annealing algorithm (SAA) and the improved simulated annealing algorithm (ISAA) are respectively used to improve the consistency of each judgment matrix. The results show that the average consistency deviation of the judgment matrices optimized by ISAA is reduced 0.0421 and 0.0106 compared to EM and SAA respectively, and the convergence speed is about 77.8% higher than SAA. The standard deviations of the weight differences before and after optimization using ISAA and SAA are 0.0074 and 0.0259, respectively, so the weight fluctuations after ISAA optimization are relatively smaller. The weight distribution corresponding to each index is obtained when the consistencies of the judgment matrices are close to the optimal state, which provides a necessary technical foundation for the evaluation of distributed server resource sharing.

Empirical formula and computer simulation are the common two ways to estimate the dielectric constant of dielectric with mixed structure. Empirical formula can be calculated very quickly, but it restricts the shape and volume of the filling material. Simulation is more reasonable theoretically, but it is computationally complex. This paper introduced a simplified simulation method for measurement of dielectric constant of complex dielectric with sandwich structure and foam structure. The experiments showed that its results became near to the results of common simulation and Maxwell-Garnett empirical formula.

The power supply continuity is the key factor affecting the normal operation of the load. The complex power quality problems (such as two-frequency ripple superposition, voltage sag and ripple superposition) in the dc power supply system will directly affect the continuous power supply of the load. The internal conversion control technology of the dual power switch can guarantee the continuous power supply of the load. To solve the problem of compound power quality in dc dual power supply system, this paper proposes a dual power conversion control technique based on PSO-MP. In this technique, PSO algorithm is used to perform coarse search and MP algorithm to identify the composite power quality disturbance. The example analysis not only verifies the validity and accuracy of the method proposed in this paper, but also improves the reliability of the double power switch and quickly realizes the conversion between two dc power sources.

At present, the research of the power lithium battery management system which was applied to the underwater robot is still in the initial stage. The lithium battery management has become one of the main contents of the development of the underwater robot technology. When the use time of rechargeable battery is short, various factors will reduce the life of the battery. Due to the problem of materials, the price of the battery is limited. Effective management of lithium battery can improve the use efficiency of the battery. In addition, the safe and effective use of electricity has great significance for the energy of the pool, which can extend the battery life and improve the reliability of the battery. The main functions of the battery management system were introduced in this paper. The hardware management system, temperature sampling, temperature detection circuit and charging module of the lithium battery were mainly designed.

In the multi-component analysis of Fourier infrared spectroscopy, the gas mixture of unknown components can be obtained by infrared spectroscopy. This article focuses on the following problems: Under the library spectrum of pure gas, by using the known Pure spectrum, qualitative and quantitative analysis of the mixed spectrum within a certain error range, based on Fourier infrared theory, the corresponding wavelet packet transform algorithm for mixed gas separation is obtained. The experimental results show that gas separation based on Fourier infrared spectrum, the wavelet packet algorithm has the advantages of high accuracy and wide application range. It has a very broad development prospect in the field of environmental gas detection technology.

The frame saw is not limited by the size of the stone. It can be installed 120 saw blades with a length of 4 m. The thickness of the saw blade is generally less than 3.5 mm, which reduces the generation of waste debris in the stone sawing process. There is no noise and dust pollution in the sawing process, which meets the requirements of high energy, low consumption and environmental protection advocated. However, in the process of sawing hard stone, the flatness of stone surface is poor. The experiment of sawing hard stone with reciprocating frame saw is carried out, and the wear morphology of the saw tooth is analyzed by SEM. The main wear forms of diamond particles are whole crystal, micro-fractured, macro-fractured, flat, and pull out. Among the pulling out diamond grains, the abnormal pulling-out diamond grains are the most, and the depth is more than or close to half of the diameter of the grains. Cracking was found at the interface between diamond grain and matrix. Through the systematic analysis of diamond grain wear, the theoretical basis is provided for the design of new sawing mode.

Part-based representations (part-based representations) are widely used in the field of shape matching and classification. We can significantly improve the robustness of the recognition algorithm by using parts, so that the computer can analyze objects from different perspectives, such as global and local. Therefore, we use the theory of part-based representations in the field of shape matching, combined with multi-scale shape decomposition (MSD), to give a multi-scale decomposition based visual part extraction algorithm (MDVPE). The algorithm can accurately separate the visual parts from the feature curves of Dongba hieroglyphs, lay the foundation for designing an efficient Dongba hieroglyphs recognition method, and also provide technical support for researching radicals and other content.

Dongba hieroglyph is a kind of very primitive picture hieroglyphs; it has a characteristic of pictograph to express meaning by using pictures, but also have some features of pictographic, ideographic, self-explanatory and echoism like hieroglyphs. The order of writing makes Dongba hieroglyphs and images essentially different. Extracting the key structural points of Dongba hieroglyphs is very important for studying the writing order and automatically extracting the strokes. Therefore, we combined the structural features of Dongba hieroglyphs and the multiple-scales decomposition theory of shape in computer vision to give the key structure point extraction algorithm. The algorithm uses a discrete curve evolution algorithm to decompose the feature curve by multiple-scales decomposition layer by layer and finally extract the points that can show the key structure of the character. And the experiments show that the algorithm has high accuracy, good stability, and good scaling, rotation and translation invariance.

According to the photovoltaic cell miniature power supply characteristics, this paper proposed a new type of high step-up Boost converter. Compared with the traditional Boost converter, this circuit has three switch tubes, and two switch capacitors are added. A high step-up ratio can be obtained within a range of duty ratios ranging from 0 to 1. Simultaneously, the voltage stress of the power device is reduced, which is beneficial to the selection of switching devices and reduces the loss of the circuit. If the inductor current is continuous, the working principle of the circuit when the duty ratio changes from 0 to 0.5 and 0.5 to 1 is analyzed in detail, and the mathematical model of the converter is established. Finally, the correctness and feasibility of the theory are proved by the Simulink simulation experiment.

In order to solve some problems occurred when creating mobile personas, especially at data collection and data analysis period, a CAUX(Context-awareness User Experience)-based method is used in this paper. We use CAUX tool to collect objective data such as APP using APP usage information, mobile phone power, Call/SMS information etc., and then doing data visualization and analyze data, display user’s mobile phone usage with Gantt chart. Next, summarize behaviors of different users when using mobile APP, and classify according to a certain dimension. Finally, form a set of personas. We create mobile personas with CAUX tools, which compared with traditional method is more concentrate on users‘real behaviors. The CAUX-based personas focused on data collection and data analysis, which can effectively solve the problems when creating personas and leads to a better performance when using.

With the rapid development and popularity of mobile devices, the aggressive adoption of mobile products poses new challenges to both users and User Experience (UX) researchers: Reducing the disruption to users when collecting subjective data, and recognizing users’ subjective data. The goal of this study is to enable the existing context-awareness tools with the ability to collect and recognize high-level context data sought by UX researchers. In this study, the author selects the idle context of users as the research objective, asks users to mark their own data, and summarizes the mapping relationship between the combination of three groups of reliable low-level context information. The Context Awareness User Experience (CAUX) system, which combines context-aware with remote data collection technology, is used for this research. This study shows that when students are both in both a high-level and idle context, if the low-level context combination is used to judge, the accuracy rate is 88%. This method of judging high-level context based on the combination of low-level context information can effectively solve the defects in the perception of high-level contexts by the current remote context-aware tools, thereby solving one of the challenges with remote user experience evaluation.

At present, the photovoltaic subsidy system is based on the centralized trusted third-party operation, which has the problems of data tampering, loss and non-traceability. Blockchain technology has the characteristics of decentralization. The application based on blockchain technology does not need the intervention of a trusted third party. Any two nodes can directly trade, making the transaction data not easy to be tampered with, lost, and traceable. In this paper, the blockchain technology is applied in the field of photovoltaic subsidy, and a photovoltaic subsidy system based on the alliance chain is developed, which solves the above shortcomings of the original photovoltaic subsidy system and makes the data in the system real and reliable.

Aiming at the problems that the single drone cannot comprehensively check the disaster situation and the single drone suddenly fails and cannot continue the rescue during the execution of the rescue request task. A multi-drone system based on ROS is designed. The hardware part of the system adopts Pixhawk to be responsible for UAV attitude control. Obtain the position of the aircraft on the world coordinate system through GPS. Using the Raspberry Pi 3B as the on-board computer is responsible for dividing each module into nodes. This enables publication and subscription of topics such as sensors, aircraft status and coordinates. This facilitates communication between multiple drones. Eventually, a set of three drones taking off at the same time, a PC capable of changing the target position of multiple drones, and a multi-drones system that can avoid collisions between drones can be realized. Compared with the single drone, the multi-drones system based on ROS can effectively solve the problem that the single drone cannot detect the disaster in all directions.

In order to solve the problems of multi-sensor read data and PID control time interval, data return and flight control mode of quadrotor UAV, a quadrotor UAV control system based on msOS platform was designed. The system uses MPU9250 to obtain UAV attitude data, MS5611 barometer to obtain UAV altitude data, AT7456E as the management chip of the OSD unit, and S-Bus signal as the remote-control signal input. Compared with the traditional four-rotor UAV bare-metal system, the system uses dual-task msOS management sensors to read data and PID control. Finally, an MSOS platform quadrotor aircraft system with OSD data return, attitude, fixed altitude and optical flow hovering flight modes were designed.

In order to solve the problem of multi-task simultaneous operation of sensor reading data, remote control data receiving, OSD (On Screen Display) data sending and PID (Proportion Integration Differentiation) attitude control, a traversing machine system based on msOS platform was designed. The system implements a traversing machine system with attitude control, fixed height control, OSD data return, and fast system response speed. First, the structure and flight principle of the traversing aircraft are introduced. Then introduced the hardware design of the system. The hardware system adopts the idea of holistic design. Centralize the interface between the sensor and the function module on the flight control board. Then introduced the software design of the traversing machine system in this paper. Then read various sensor data through msOS. The software executes the cascade PID algorithm to realize the attitude control of the rider. Then introduced the debugging process. Finally, the shortcomings of the traversing machine system based on msOS platform in this paper are summarized and the research prospect is prospected.

Many image super-resolution algorithms can be formulated in a mapping framework based on the natural image prior. Generally, the mapping function with free parameters is learned by minimizing the reconstruction mapping error. In this paper, we obtain a considerable image super-resolution algorithm which gains in accuracy and speed by combining joint mapping learning with fast approximations of sparse coding. A novel “dictionary” training method for single image super-resolution based on feed-forward neural network is proposed. The training algorithm alternates between solving sparse coding problem and learning joint mapping relation problem. The learning process enforces that the sparse coding of a low-resolution image patch can be regarded as the shared latent coordinates for reconstructing its underlying high-resolution image patch with the image high-resolution dictionary. Experiments validate that our learning method shows excellent results both quantitatively and qualitatively.

E-commerce tourism is a model derived from the emergence of e-commerce. This study is aimed at the consumers who use e-commerce tourism products. Taking the consumers who purchase tourism products through tourism websites within one year as the survey object, the satisfaction of consumers with e-commerce tourism products transactions is discussed. 227 random samples are selected for the questionnaire survey. Based on the data of satisfaction of technology acceptance dimension, transaction cost dimension and service quality dimension, the research model discusses customer satisfaction with e-commerce platform.The results show that transaction cost has a significant positive impact on platform service quality perception, and service quality has a significant positive impact on platform use satisfaction. It is suggested that e-commerce tourism platform make use of the advantages of big data to make tourism products and grasp the needs of consumers. In the aspect of setting contract cost, reduce the cost for consumers to unsubscribe, so as to prevent the loss of platform caused by the pressure of unsubscribe cost that consumers fail to purchase products. Increase and simplify complaint channels, automatically identify customer needs, reasonably refund, and reasonably increase the number of customer service platform, so that consumers can solve problems more quickly and efficiently.

Intelligent technologies can effectively improve the management level and safety of training rooms in vocational colleges. This article first analyzes the challenges encountered in the construction of the smart training rooms, then designs a system architecture based on mobile cloud video surveillance technologies for building smart training rooms, analyzes the main software function modules, and discusses video transmission and cameras control. The realization of smart training rooms has made a useful exploration of applying a new generation of information technology to manage training rooms in colleges.

An IoT (Internet of Things) system is a multi-machine system composed of a master controller and multiple slave controllers. In some automatic control systems such as petroleum and electric power, the main control uses UART (Universal Asynchronous Receiver/Transmitter) to communicate with the slave controller. However, the communication rate of the UART interface is limited and the delay time is long, which affects the system’s response to external events. In order to improve the response speed of the system to external events, this paper designs a multi-machine system based on SPI (Serial Peripheral Interface) interface. First, we use the SPI interface of the microcontroller to build a communication network. The circuit structure of the system is simple, and the hardware resource consumption is low. Then, for the problem that the SPI interface cannot detect communication errors and there is no response mechanism, we designed a communication protocol to make up for these defects from the software level. Finally, we use STM32F429IGT6 as the master controller and STM32F103C8T6 as the slave controller to build the experimental platform. The experimental results show that the system is safe and reliable, and can meet the requirements of real-time applications.

In some construction sites, workers often do not wear helmets and cause safety accidents. In order to prevent the occurrence of safety accidents caused by not wearing safety helmets, we propose a detection method of safety helmet wearing based on CenterNet. Input image into fully convolutional network to obtain a heat map, and the peak of the heat map corresponds to the center of the target. The image features on each peak can predict the width and height of the target frame. The network uses dense supervised learning for training. The inference stage is a single forward propagation network without NMS post-processing. Using video capture from the construction site as a part of the data set. Theoretical analysis and experimental results show that when using detection method of safety helmet wearing based on CenterNet, its recognition accuracy and rate meet the requirements of helmet wearing detection.

Sentiment analysis has attracted a wide range of attentions in the last few years. Supervised-based and lexicon-based methods are two mainly sentiment analysis categories. Supervised-based approaches could get excellent performance with sufficient tagged samples, while the acquisition of sufficient tagged samples is difficult to implement in some cases. Lexicon-based method can be easily applied to variety domains but excellent quality lexicon is needed, otherwise it will get unsatisfactory performance. In this paper, a hybrid supervised review sentiment analysis method which takes advantage of both of the two categories methods is proposed. In training phrase, lexicon-based method is used to learn confidence parameters which used to determine classifier selection from a small-scale labeled dataset. Then training set which is used to train a Naive Bayes sentiment classifier. Finally, a sentiment analysis framework consist of the lexicon-based sentiment polarity classifier and the learned Naive Bayes classifier is constructed. The optimal hybrid classifier is obtained by obtaining the optimal threshold value. Experiments are conducted on four review datasets.

An automated video surveillance system plays a vital role in unattended substations. To minimize the camera cost and maximize the monitoring coverage of the substation, efficiently placing the camera is critical. Two optimization models are developed in this paper to satisfy two kinds of requirements of unattended substations monitoring in practice: minimizing cost with full coverage, optimizing coverage with a fixed budget. The second model is mainly proposed for the first time to get the minimum monitoring fitness function as the optimization goal. By introducing piecewise functions in the second model, it meets different monitor-ing requirements in different regions of substations and reduces the over-lapping monitoring area efficiently. The two models are tested using a genetic algorithm to get the optimal layout of the cameras. The validity and practicability of the two models of monitoring in unattended substations are verified by simulation. The results suggest that this model-based optimization approach to cameras layout can be used to improve the efficiency of substation video surveillance systems.

With the development of electric vehicles (EVs), the random access of large-scale EVs to the grid will become a reality in the future, which will have a significant impact on the security and stability of the power system operation. Therefore, this paper proposes an improved multi-objective particle swarm optimization algorithm to optimize the micro grid load scheduling model based on EV. Microgrid is an autonomous system that can realize self-protection, management and control. It is an effective way to realize active distribution network, which is the transition from traditional power grid to smart grid. At present, micro-grid has become a useful supplement of power grid with its flexible power supply and diversified power generation methods. The model includes operation cost, pollutant treatment cost and load variance. The distributed generation (DG) including diesel engine (DE), micro turbine (MT), photovoltaic (PV) array and wind turbine (WT) is studied in the model. The uncertainty of electric vehicle is modeled by Monte Carlo simulation. The simulation results show the effectiveness of the algorithm.

We consider the problem of robot local path planning using traditional dynamic window approach based ROS. By means of an improved dynamic window approach, we are able to reduce the complexity of the problem and provide a practically efficient procedure for its solution. Improved dynamic window approach based on robot dynamics model is introduced and described in the paper. By abandoning the trajectory that the velocity reduces to zero when it encounters an obstacle to simplify the search speed space. This algorithm improves the real-time performance and reduces the computational complexity of the algorithm.

In wireless sensor and actuator networks (WSANs), actuator nodes require much longer charging time to be fully re-charged compared with sensor nodes. Also, actuators are usually mobile in the network. These features bring new challenges to the charging problem in WSANs. Based on the characteristics of WSANs, a practical mobile charging approach (EMC) for WSANs is proposed. In order to ensure the actuator nodes do not fail and meanwhile minimize the number of failed sensor nodes, the next charging candidate is selected according to both the remaining energy situation of the node and current location. Simulation results show that this proposed approach can guarantee the survival of all actuators. Further, it can effectively reduce the node failure ratio of sensor nodes and achieve the trade-off be-tween the charging delay and the charging cost.

As internet of things (IoT) devices are booming, a huge amount of data is sleeping without being used. At the same time, reliable and accurate time series analysis plays a key role in modern intelligent systems for achieving efficient management. One reason why the data are not being used is that outliers are preventing many algorithms from working effectively. Manual data cleaning is taking the majority time before one solution could really work on data. Thus, data cleaning, especially fully automated outlier detection is the bottleneck which should be resolved as soon as possible. Previous work has investigated this topic but lacks study on overview from outlier and detection categorization aspects at the same time. This works aims to start covering this topic and to find a direction regarding how to make outlier detection and labelling more automated and general to be suitable for most time series data from IoT devices.

In the actual path planning of the mobile robot, the algorithm planning of the robot is almost always to find the shortest path, but if the robot is in a narrow working environment with many corners, it will make the robot too close to the obstacle, and the robot’s edge will contact with the obstacle, resulting in the path planning failure. In order to solve the security problem of mobile robot path planning, we put forward a kind of inflation obstacles, refinement algorithm first determine skeleton and area of the mobile robot on the map, and then the refinement algorithm can get skeleton to improve traffic area, connected to the path of starting point and end point, define the passage area. Then by searching the first and second nodes, we can find the optimal path that accords with the safety of the robot. The new path planning algorithm can realize non-collision planning, improve the safety and stability of the robot, and verify its effectiveness and authenticity through simulation results.

Aiming at the problem that the traditional collaborative filtering algorithm has low recommendation accuracy and unsatisfactory results when the matrix is sparse, weighted SlopeOne-IBCF algorithm based on user interest attenuation and item clustering is proposed. Firstly, use the weighted Slope One algorithm to predict user ratings and fill in matrix, fuzzy cluster the items in the data set; then combine the interest decay function to modify the target user’s historical item score, and determine the category of items that the target user is interested in at the current stage; then integrate the time decay function into the process of calculating the similarity, pay attention to the factors that interest and hobbies change with time; finally, the similarity Top k items that are the highest and are not in the user’s historical item set are recommended to the target user. Experiments show that the algorithm in this paper effectively alleviates the problem of poor recommendation and inaccuracy in the case of sparse data, and further improves the accuracy of recommendation.

Output current is an important parameter of the power electronic transformer. In order to study the control strategy of the output current of the power electronic transformer, this paper, aiming at ac-dc-ac type power electronic transformer, carries out the corresponding control strategy simulation of the power electronic transformer in MATLAB/SIMULINK, including: PWM control, parallel operation, SVPWM control and current tracking hysteresis comparison mode. The simulation results show that the output current of the power electronic transformer can present different output characteristics by using different control strategies, and the output characteristics can be optimized to some extent.

Large scale application of wind power in power system will reduce the system equivalent inertia and primary frequency control ability. This paper proposes a frequency coordination control scheme for DFIG under power-limited operation to increase the amount of system inertia via virtual synchronization control, so DFIG can provide a transient frequency support for system. The released reserve power caused by decrease of pitch angle can compensate the output power sag, and reduce steady state frequency deviation. The characteristics of virtual synchronization control and the pitch angle control of DFIG are combined to effectively suppress the system frequency fluctuation caused by load change. Simulation results show that the proposed coordinated control strategy can make full use of the wind turbine’s own load reduction operation and virtual synchronous generator technology to effectively improve the frequency stability and inertia support effect of the grid-connected system, which is conducive to the safe and stable operation of the power system.

Underwater fishing robots have been more and more widely used in ocean breeding nowadays. They have become the most effective and potential tools in underwater fishing. A small-scale underwater fishing robot has been developed, which has the functions of sensing and detecting. This robot can measure the conductivity, temperature, depth (CTD) and robot’s attitude in real time, and at the same time the controller transmits them to the mother ship through CAN bus. In this paper, firstly, the whole structure of the self-designed robot is introduced, then the control system is designed, finally, a physical experiment is carried out. The experiment shows that the robot can meet the design requirement.

A cleaning robot for underwater hull surface is designed, which mainly includes adsorption mode, track driving mode, driving mechanism and cleaning mechanism. The double track moving mode has the advantages of large contact area with the hull surface, good adaptability to the wall surface, large magnetic adsorption capacity, strong bearing capacity and stable movement. The form of rear drive was adopted. When the track passes through the obstacle, the magnetic track that the robot fits on the hull surface is lengthened, so that the track can better adapt to the fluctuation of the hull surface. Washing tools are divided into soft material cleaning tools and hard material cleaning tools. Only the soft material cutter is close to the hull surface, which can avoid the damage of the hull surface. Electromagnet adsorption mode was adopted, which is easy to realize rapid movement. In addition, the working range can be increased with the rotary mechanism, and the working efficiency can be greatly improved with the rotary cleaning head.

At present, the heavy truck tubeless workshop adopts manual or low automatic way to transport the wheels, it is necessary to design a set of control system of wheel automatic transportation line which can run safely and stably in workshop environment, instead of the existing low automation transportation mode. The paper introduces an automatic transportation line using PLC and WinCC technology, the main content includes the introduction of the overall structure of the automatic transportation line, and the composition and function of automatic production line are described in detail, the selection of PLC, step motor to drive, etc. Use PLC technology to design control programs, apply configuration software technology to design the monitoring system, and connect PLC and configuration software to form a control system through Ethernet communication. Finally, the paper describes the experimental results. The system has the characteristics of high automation, data visualization, simple operation and stable operation. It is a safe, stable, efficient and low labor cost tubeless wheel transportation mode.

In order to meet the requirements of positioning accuracy of indoor mobile robot navigation system, this paper merges vision and inertial navigation system (INS) to improve robot positioning accuracy. Aiming at the low frequency of the visual navigation system (VNS) and the high frequency of the INS, a multi-frequency Kalman filter algorithm is proposed, and two different measurement equations are designed. The measurement Eq. (1) updates the position information of the inertial navigation after sampling in the INS, and the measurement Eq. (2) updates the position deviation of the mobile robot after sampling in the VNS. Finally, the accurate position information of the mobile robot is estimated by the inertial navigation position updated by the measurement equation one minus the optimal error updated by the measurement equation two. Realize the effective fusion of the location information of visual matching estimation and INS location information. The experimental results show that the multi-frequency Kalman filter algorithm is improved on the basis of the traditional filtering method, and the integrated navigation method proposed in this paper further improves the positioning accuracy of the INS.

With the development of the electronic industry, smart meters are more and more required by people. Compared with traditional mechanical meters, smart electricity meters have more functions, such as multi-rate metering, intelligent interaction, data transmission, safe electricity, etc. NILM algorithm, also known as Non-intrusive load monitoring algorithm, can analyze the power consumption data through PC to show more detailed power consumption to users, which can greatly improve the power consumption experience of users. Based on AT89C51 microcontroller, ADE7755 energy measurement chip, LCD1602 display module, etc., this paper designed the overall hardware circuit of smart meter; In the software part, it not only realized basic ability of smart meters, but also realized the load recognition function by using the NILM algorithm.

In this paper, the importance of the coverage, security, transmission delay, service rate and cost in each heterogeneous wireless network to the power business is comprehensively analyzed from the perspective of business preference of power distribution communication and the objective conditions of 5G heterogeneous network. The weight distribution of each network attribute for different business is more accurately obtained by using the entropy enhanced AHP and method. The performance ranking of each network provides a scientific, objective, accurate and optimized selection scheme for the power distribution communication service. Compared with other algorithms, the proposed algorithm shows better performance in term of blocking rate.

Off-line handwritten numeral recognition is a pattern recognition problem of the images of ten numbers. To improve the recognition efficiency, the number’s image’s character dimension should be decreased. In order to improve the recognition veracity, the character mode instability resulting from different writing styles and habits should be considered. The article proposed a numbers recognition method which combined with the statistical characteristics and structural features of numbers. Firstly, the principal component analysis (PCA) method was adopted to extract the numeral image’s statistical characteristics. The numeral recognition will be realized through analysis of the reconstruction error of the model, which reconstructed by the principal components. To further determine the type of numeral, the structural features of width and height rate should be added. Finally, through experiments on the numeral image identification, the reliability and accuracy of this method of digital recognition were verified. The deficiency of this method in real-time recognition was analyzed.

3D object detection is an important research direction in the fields of computer vision and pattern recognition in recent years. This technology can provide important technical support for unmanned driving and intelligent robots. Aiming at the challenges of object detection caused by the sparseness of 3D point clouds in outdoor scenes, this paper designs a 3D point cloud multi-target detection method based on pointnet++. The method first preprocesses the collected original point cloud; after obtaining the point cloud of the region of interest, the point cloud is clustered, and then the 3D target detection is performed by pointnet++ to obtain the object category. Finally, get the size and orientation of the target object through the 3D boundingbox. In order to verify the effectiveness of the method in this paper, the point cloud data of real outdoor scenes were collected using lidar, and a sample set was produced for network training. The final results verify that the method can achieve higher detection accuracy and meet the requirements of real-time performance.

With the development of technology, the bandwidth of public networks continues to increase, and the means of information exchange is constantly enriched. Especially the impact of COVID-19, distance education has gradually become popular, and many problems have also been exposed. The fact that the current distance education users are widely dispersed, the number of users is increasing, the network development is unbalanced, and the access methods are diverse, this paper designs a distance education system suitable for edge network environments. The system uses a B/S architecture, uses cloud computing technology to build a data center, and uses edge computing to build edge servers, enabling educational resources to move forward to users, reducing network latency, shortening response time, and improving user experience. This article outlines the system’s architecture, the composition of the functional modules of each part, the working method, the core technology adopted, and the expected effect.

In recent years, the rapid development of the shipping industry at sea has become an essential part of China’s economic development. Large ships carry more items, and changes in temperature and humidity in the closed cabin are especially crucial for preserving items. To grasp the change of temperature and humidity in the cabin in time, an intelligent distributed temperature and humidity collection system based on narrowband Internet of things is designed. The system combines microprocessor module (STM32), sensor module, and narrowband Internet of things to realize a real-time collection of temperature and humidity in the cabin and alarm function when the change exceeds the threshold. The mobile APP is used to realize remote monitoring and data collection monitoring. narrowband internet of things; temperature and humidity collection; STM32; remote control.

Rainfall plays a vital role in guiding clothes drying. This paper proposes an intelligent clothes hanger system based on the analysis of rainfall data. The system consists of an electric mechanism, weather monitoring module, wireless communication module, central control module, network server, and user APP. The weather data monitored by the local weather monitoring module is uploaded to the network server in real-time, users can monitor and control the clothes hanger through the APP. Besides, the web server constructs a two-dimensional rainfall map based on the weather data provided by the clothes hangers in different geographic locations. Using the weight matrix to perform convolution operation on the rainfall map, the weighted average of the rainfall conditions at any point and its surrounding points is realized. The result is used as the rainfall risk prediction of the area corresponding to this point, used to guide drying. Finally, the feasibility of the algorithm is verified by Labview simulation.

Objective To develop Smart Shoes with obstacle detection system for visually impaired people. Method The obstacle detection system adopts a triple-axis accelerometer to collect feet’s acceleration data, and uses an ultrasonic sensor to detect obstacles. The system is controlled by STM32L432KC (a microcontroller from STMicroelectronics), and powered by a Lithium-ion battery that can be recharged either by a charger or by walking. A gait events recognition algorithm is proposed to detect the motion state of feet. Obstacles are detected only when users are walking with foot in stance (ST) phase. Moreover if a fall is detected, the Smart Shoes will connect to the cellphone and call the emergency contacts. Results The overall recognition ratio of the gait events was 90.9%, the ratio of walking, jiggling and fall (simulated) were 91%, 88.5% and 100% respectively. The detection resolution of Smart Shoes depends on the ultrasonic sensor. User’s average obstacles detection distance are all above 50 mm, and for each user the detection distance is proportional to the obstacles’ dimension. Conclusion Experimental results indicate that the Smart Shoes performs stably in real-time, and has high detection accuracy with low false-alarm rate.

LED algorithm is a new lightweight encryption algorithm proposed in CHES 2011, which is used for IOT to protect the communication security of RFID tags and smart cards. It has been found that it is possible to retrieve secret key of algorithm by appropriate analysis of the devices’s power consumption, and the differential power analysis is the most powerful attack. This paper proposed a method of differential power analysis on LED algorithm. Combined with hardware circuit, it designed and implemented differential power analysis of the LED encryption system. The experimental results show that method which is proposed in this paper was cracked the 64 bits key of LED. Thus, without protection, the cipher LED would be difficult to resist differential power analysis. and according this, We provide a general analysis method of differential power analysis on other lightweight cryptographic algorithms.

In view of the problems of electric power wireless communication and to serve the power consumer better, on the basis of the basic principles of TD-LTE’s network structure, traffic model of power service is analyzed in detail, and in the actual case, through the data processing method and comparison analysis method, the technical parameters, theoretical analysis and field test results of 1.8 G and 230 M LTE systems are compared, and the ability of power wireless private network to support power service is verified. The analysis and testing provide a technology support for electric power communication access network construction and lays the foundations for popularization and application of LTE power wireless private network.

Flood disaster is one of the most serious natural disasters since the 20th century. Flood risk management has become an important means to deal with flood problems. The UK’s flood management system has undergone nearly a century of development, and its evolutionary experience is worth thinking about. This paper focuses on the development of the British flood management system based on the relationship between central and local authorities and between water and planning agencies. The development process of British flood management system is summarized, from “flood disaster prevention system” to “flood disaster management system”, and finally to “flood risk management system”. By comparing the development history of the British flood management system with the current situation of flood management in China, this paper suggests that the concept should be the same, and the organization should be tailored to local conditions, and measures should be broadened.

The collection, bearing, analysis and application of the basic electric power data from the ubiquitous Internet of things have put forward new and higher requirements for the data collection, bearing, analysis and application, and higher requirements for communication technologies and methods is raised. In view of this situation, the integration and development means of 5G technology and power system business are put forward, and the application of typical business scenarios in various power systems under 5G technology is analyzed. A typical accurate load control service is selected for testing, and the test results show that the delay is improved. The application analysis of 5G in electric power system supports the Internet of things application, provides the technical support, and lays a solid foundation for the popularization and application of 5G technology.

In view of the problems of electric power wireless communication and to serve the power consumer better, On the basis of the basic principles of TD-LTE’s network structure, interference in the wireless network optimization of power system are analyzed in detail, and put forward a method of interference remediation for power system. Based on the actual cases, the feasibility of interference remediation is verified by means of data processing and comparative analysis. The interference analysis provides a technology support for electric power communication access network construction and late operation and lays the foundations for popularization and application of LTE power wireless private network, and the security of the power network is guaranteed.

There are many problems in distributed energy access, such as many access points, different access distances and miscellaneous access protocols, which results in low reliability of synchronous operation of the whole system and poor reliability of clock, which further affects the efficiency and quality of centralized regulation and control. Therefore, the technical scheme based on time sensitive network is put forward, and the Ethernet equipment based on time sensitive network technology is developed, all the sampling data in the distributed energy control task are under the same time section, and the time sequence of the control task can be measured and predicted. It provides important technical support for the establishment of efficient and low-cost source network load storage and control center. The results can be popularized and applied to the national source network storage and control center, and has a broad application prospect.

Aimed at the problems existing in the planning of power wireless communication, in order to better serve smart grid and carry out the application of time-sharing long-term evolution technology in power communication, a highly reliable planning method for wireless private network is proposed. It improves the survival of the user’s service by using overlapping coverage of base station signals, and the overlapping cells are merged to avoid the same frequency interference. By optimizing the deployment of the organization of the base station, in the case of the same number of base stations, the average inter station distance of overlapping coverage is increased, and the network reliability is improved on the premise of ensuring network performance. The planning scheme solves the problem of mutual interference caused by the same frequency overlapping coverage between base stations.

Edge computing calculates computing tasks on computing resources close to the data source, which can effectively reduce the latency of the computing system, reduce data transmission bandwidth, and ease the pressure on the cloud computing center. However, with the explosive growth of business terminals, the capacity of a single edge node is limited, and it is difficult to meet all business requirements at the same time. Therefore, a task allocation method for power Internet of Things based on two-point cooperation is proposed. First, a task allocation model based on two-point cooperation was established to minimize the average task completion delay while meeting business resource requirements. Then the ECTA-MPSO (Edge Collaborative Task Allocation based on Modified Particle Swarm Optimization) algorithm is proposed, which solves the problem that the task allocation scheme easily falls into a local optimum. Simulation results show that the average delay decreases by 32.8% and 12% respectively compared with benchmark and GA algorithm.

The power Internet of things (IOT), which integrates edge computing, has become a research hotspot because of its characteristics of edge intelligence, wide interconnection and real-time decision-making. However, with the construction and business application development of the power Internet of things, problems such as insufficient processing capacity of edge devices and insufficient utilization of distributed edge resources are increasingly prominent, affecting the high-quality service provision of the power Internet of things. To this end, this paper uses cloud-side collaboration and service provision ideas to propose a cloud-based collaboration-based power IoT heterogeneous shared resource allocation method, through resource optimization to improve the quality of power IoT service provision. From the perspective of resource service matching and integration, the algorithm distinguishes service types, and allocates different cloud-based collaborative resources according to different service provision types to meet the corresponding real-time processing needs of services and ensure optimal allocation of edge resources. Simulation experiments show that the algorithm can give priority to ensuring the efficient use of edge resources on the basis of ensuring that the resource requirements of the business request are fully met, and achieve the purpose of improving the quality of power Internet of Things service provision.

Aiming at the problem of difference in business busyness between edge nodes caused by obvious uneven distribution of service requests in the edge network, this paper proposes a container migration mechanism for load balancing. First of all, the timing of container migration is determined based on the classical static resource utilization threshold model. Then, a container migration model of load balancing combined migration cost is established to minimize the impact of container migration while balancing the load of edge network. Finally, the priority of container migration is calculated from the perspective of resource correlation, and a migration algorithm based on the improved ant colony algorithm is designed for container migration in the context of the power Internet of things. The simulation results show that the container migration mechanism proposed in this paper can not only improve the load balancing degree of edge network but also reduce the cost of container migration.

With the rapid development of information technology, the Power grid Internet of Things technology has been integrated into all aspects of life. The intelligent network formed by the Power grid Internet of Things that interconnects people, things and things has greatly promoted the intelligent development in various fields. However, due to the large number of IoT devices and the limited resources of many devices, it has brought huge challenges to the protection of IoT information data and device management. This paper proposes a lightweight and secure IoT authentication key agreement protocol for the problems in the IoT environment. The protocol adds identity authentication based on the ECDH algorithm to prevent man-in-the-middle attacks. The identity authentication part is designed based on the dynamic password technology with hash as the main operation, and realizes the mutual authentication of the dynamic password based on the identification. The entire protocol has good security and can effectively resist man-in-the-middle, impersonation, eavesdropping, and replay attacks, ensuring the security of IoT device communications. At the same time, the protocol resource overhead is small, and it can adapt to resource-limited IoT devices. In addition, the authentication mechanism combined with cloud-edge-end collaboration can efficiently manage many IoT devices.

With the promotion of the power Internet of things strategy of State Grid Corporation of China, using wireless communication technology to realize the interconnection of everything in multi business scenarios has become a research focus. However, there are still blind coverage areas such as basement, dense urban area and building shadow area in the existing power wireless private network, which can not achieve full communication coverage. For this reason, this paper designs a relay transceiver for deep coverage of power wireless network, which adds relay nodes in the weak coverage area of the network, and evaluates its coverage ability by defining its importance, which effectively improves the coverage of power wireless communication system. The experimental results show that the relay transceiver designed in this paper can realize the deep coverage of the power wireless network and improve the reliability of the power wireless communication network.

As a key technology to promote the development of network applications, network virtualization decouples traditional network service providers into service providers and infrastructure providers, and separates the physical network into multiple virtual end-to-end networks to meet the different needs of the upper tenants. In application scenarios represented by edge computing, virtual networks usually need to be carried by multiple different infrastructure networks in different geographic locations. In order to ensure the security of the internal network topology, each infrastructure provider has incomplete information sharing between them, and the embedding is also restricted by region and resource type. Therefore, this paper considers the impact of the substrate network topology on virtual network embedding from multi-angle, and designs a multi-domain virtual network embedding method based on particle swarm optimization. In this paper, a calculation index based on embedding cost and link interference is designed, and a particle generation mechanism based on node resource richness in particle swarm optimization is proposed to reduce the computing time of the algorithm. Simulation results prove the superiority of the embedding algorithm proposed in this paper in terms of embedding cost and computing time.

Considering a large number of end-to-end services, in order to orchestrate and manage the deployment requests of edge computing servers to provide better intelligent services, in this paper, we adopt network slice. In the regard of network slice, the most important challenge is how to deploy the node and link resource requirements in the request to the underlying network with limited resources. Current researches ignores the needs of multi-domain services. In addition, the greedy method of orchestration is likely to cause node resource constraints, resulting in problems such as low acceptance rate and low revenue. Therefore, in this paper, we propose an intent-based multi-domain anti-fragmentation request orchestration method, including expression method of service requirements, scheduling of services based on priorities, selection of candidate nodes, and multi-domain orchestration algorithm. The experimental results show that our approach has a good performance in improving the acceptance rate and overall revenue.

In the fog computing environment, in order to solve the problem of large service offload delay in large-scale network, this paper first designs a layered service offload management architecture of the Internet of things in large-scale fog computing environment. Based on the management architecture, four service unloading modes are proposed according to the task characteristics. Secondly, based on the theory of deep deterministic policy gradient (DDPG), a decision model of service unloading based on DDPG is constructed. Finally, aiming at minimizing the service offload delay, a service offload algorithm based on DDPG is proposed. In the experiment, the algorithm is compared with the traditional algorithm, which proves that the algorithm has great advantages in the delay and success rate of service unloading.

In an edge computing network environment, in order to improve the utilization of network resources, this paper proposes a server deployment algorithm for maximizing network resource utilization under edge computing. First, the problem is modeled from four aspects: network topology, user service requests, storage and computing resource allocation, and service node traffic usage matrix. Secondly, an objective function for minimizing network traffic is constructed, and a server deployment algorithm for maximizing network resource utilization under edge computing is proposed to solve. The algorithm uses the idea of slack constraints to solve the optimal server deployment strategy. Use tabu genetic search algorithm to find the optimal storage node and compute node deployment location. Use Lagrange operator to solve the best set of service nodes. In the experimental part, it is verified that the proposed algorithm achieves better results in reducing network traffic.

In order to identify the key nodes in the power data service and assign high reliability network resources, this paper proposes a power data network resource allocation algorithm based on TOPSIS algorithm, which consists of five parts: calculating the importance index of communication network resources, calculating the importance index of the service node, calculating the importance evaluation of the service node based on the TOPSIS algorithm, allocating resources to the service node in turn, and allocating resources to the service link. In the simulation experiment, the success rate of resource allocation, resource utilization, and resource allocation performance of key nodes are compared. It is verified that the algorithm cannot decrease success rate and resource utilization, and key service nodes are allocated better resources and improve the performance of resources.

With the rapid development of power grid, the types of power services are becoming more and more diversified, resulting in different service demands. As one of the important technologies of 5G, network slice is used to accommodate different services on the same physical network. After a brief analysis of smart grid background, this paper makes a deep research on the network slice. Each service in smart grid has its own requirements of bandwidth, reliability and delay tolerance. In order to ensure the QoS of Smart Grid, a dynamic optimization scheme of network slicing resources based on reinforcement learning is proposed. This algorithm adjusts the network slice resource dynamically, we can predict the traffic by considering the change of traffic in future network slice, and then deduce the partition of network resource in the future. The reinforcement learning algorithm is then used to make the state of network resource partitioning at future moments influence the current partitioning policy to get the best current policy. Based on this algorithm, the fast response to the change of network demand can be guaranteed in the process of resource allocation, and it is verified by simulation.

The massive use of Internet of Things (IoT) mobile devices (MDs) and the increasing demand for their computing have created huge challenges for the current development of the IoT. Mobile edge computing (MEC) and cloud computing provide a scheme for these problems. In the process of offloading, MDs and servers are facing difficulties such as high consumption and high latency. So it is necessary to reasonably offload computing tasks to MDs, edge servers, or cloud servers. In view of this situation, the research direction of this article is how to reduce the power consumption of the device and the server while ensuring that the delay requirements of different tasks are met. First we formulate the proposed problem as a nonlinear combinatorial optimization problem, then propose the cloud-edge-terminal collaboration offloading algorithm based on improved genetic algorithm (IGA). Finally, the characteristics of the algorithm are studied by simulation and compared with other algorithms to verify the performance of the algorithm.

With the development of Internet of Things (IoT) technology, the types and the volume of business have been increasing rapidly. The existing centralized cloud processing model is hard to meet the requirements of delay-sensitive and compute-intensive services. So, mobile edge computing and cloud computing are introduced to realize a cloud-edge-terminal collaboration network architecture. However, there still exist problems such as high energy consumption and long delay among devices and servers. To overcome these challenges, a cloud-edge-terminal collaboration offloading scheme based on first fit algorithm (FFA) and genetic algorithm (GA) is proposed, which combines two allocation modes. On one hand, for delay-sensitive tasks, FFA is designed to quickly offload tasks. On the other hand, for dense tasks, GA is designed to accurately offload tasks. To make the best of the advantages and avoid the disadvantages of FFA and GA, we adopt the method of using two algorithms alternately, and restrict the rules of the alternations. At the end, the characteristics of the algorithm are studied by simulation and compared with other algorithms to verify the performance of the algorithm.

IoT (Internet of Things) terminal security monitoring and evaluation encompasses technique and administration. There are several uncertainties in the monitoring and assessment process of various types of IoT terminals, which cannot be fully quantified. Therefore, it is difficult to achieve completely objective safety risk evaluation. For this to happen, the study proposes an IoT terminal security monitoring and assessment model based on grey relational cluster analysis. First, by integrating experts’ knowledge, the conditional probability matrix (CPM) of cluster analysis are explained, which lays the foundation for establishing the security monitoring and assessment model. Then, through the grey relational cluster algorithm, the subjective judgment information of the experts on the threat degree of the target information system is synthesized as prior information. At the same time, through the observation node of objective evaluation information, the safety threat levels are synergized to realize the continuity and accumulation of safety evaluation. Ultimately, simulation examples verify the rationality and effectiveness of the pattern.

Most current researches use packet loss as a prerequisite for congestion to design congestion control algorithms. Wi-Fi, 4G, satellite networks and other wireless networks have been widely used. Given that it is easily influenced by the natural environment, the transmission of wireless network features in large delay jitter and high random bit error rate (BER). In this network environment, the performance of traditional congestion control algorithm is poor. This paper proposes an improved congestion control algorithm NBBR. The algorithm is based on a feedback idea to adjust the sending rate of the sending end. Link capacity is used to adjust the transmission rate and the size of the congestion window to shield the transmission rate reduction caused by high bit error rate in wireless communication environment. Different delay measurement strategies are determined by the degree of delay jitter, which can reflect the network condition more accurately and improve the utilization rate of the network. Finally, by comparing the simulation experiment results, it is proved that the proposed algorithm can maintain higher throughput in the wireless network with higher delay jitter and higher BER.

With the rapid development of wireless network, the amount of heterogeneous services increases, which have significant differences in QoS requirements. However, the traditional service management methods realized by artificial distinction are difficult to satisfy the QoS requirements of services with high bandwidth and burstiness. Therefore, it has great significance to better use the limited network resources and design suitable QoS guarantee mechanisms for wireless network. In this paper, a reinforcement learning based QoS guarantee traffic scheduling algorithm RQTS is proposed to optimize the utilization rate of wireless network. First, the traffic scheduling optimization model is constructed based on Lyapunov theory, which sets the optimal system utilization as the objective and all the restrictions are transformed into queue stability problems. Then, the QoS routing method RQTS is adopted to solve the optimization problem through traffic access control and transmission path control, the effects of different transmission channels, transmission rates and transmission path delays of the system are examined. Simulation results show that, the proposed mechanism can satisfy the QoS requirements and optimize the weighted system utilization.

With the continuous development of computer networks in recent years, the scale and types of services carried by the network are increasing. Accurate traffic prediction results provide the main support and reference basis for network operation and maintenance functions such as network attack detection. Since network traffic has certain dynamics, continuity, and long correlation and self-similar characteristics, the artificial intelligence method is generally used for network traffic prediction. Among them, the recurrent neural network has a short-term memory performance and has a good prediction effect for time series data such as network traffic. However, when the time series span is relatively long, the problem of gradient disappearance or gradient explosion may occur, so further optimization is required. In this paper, we propose a network traffic prediction method based on parameter pre-training of clockwork neural network. This method is first based on CW-RNN, and then introduces the differential evolution algorithm to pre-train the clock parameters. At the same time, the differential evolution algorithm is further improved by changing its crossover factor and mutation factor to improve the accuracy of its convergence. Through computer simulation, the flow prediction method proposed in this paper can obtain accurate prediction results.

Electricity is the most critical input product for most businesses in the world nowadays. The utilization of electricity has assisted in the discovery of breakthrough disclosures and advances and has in time, become the most crucial constraint in the development of economies. With the introduction of distributed energy resources in microgrids, energy users are gradually turning into prosumers i.e., users who simultaneously produce and consume energy. This has brought about peer-to-peer (P2P) energy markets where prosumers can trade their surplus energy. However, there are privacy and security issues associated with such markets. Prosumers find it insecure to trade their power in an untrusted and nontransparent environment. In a P2P energy market, an intermediary oversees the transaction process between parties on the market and this poses a threat to the privacy of the prosumers. Consequently, a unified and robust energy trading market is required for prosumers to trade their energy. To address these security and privacy challenges, we leverage the properties of the blockchain innovation to propose a secure energy trading market where prosumers can trade their energy without compromising their privacy.

With the development of Internet of Things (IoT), the types and the volume of IoT services have been increasing rapidly. Mobile edge computing (MEC) and cloud computing has recently emerged as a promising paradigm for meeting the increasing computational demands of IoT. More and more computation offloading algorithms of MEC and cloud computing have appeared. However, existing computation offloading algorithms cannot have particularly good performance in various scenarios. In this regard, we proposed a cloud-edge-terminal collaborative computation offloading algorithm based on Asynchronous Advantage Actor-Critic. It uses Asynchronous Advantage Actor-Critic to make the task choose one of the two algorithms that has better performance in their respective scenarios, and achieves the complementarity of the advantages and disadvantages of the two algorithms. Finally, the characteristics of the algorithm are investigated by simulation and compared with other algorithms to verify the algorithm’s performance.

In the edge computing environment, in order to reduce the energy consumption of the entire network on the premise of meeting user needs, this paper proposes a low-energy edge computing resource deployment algorithm based on Particle Swarm. First, an edge computing service model based on SDN is designed, which includes three types of devices and three types of management processes. Secondly, the two requirements of user content service request and calculation service request are analyzed, and three energy consumption models of the entire network are constructed: storage energy consumption, calculation energy consumption, and transmission energy consumption. Finally, based on the main idea of the particle swarm optimization algorithm, the solution method is modeled, and a low-energy edge computing resource deployment algorithm based on particle swarm optimization is proposed. In the experimental part, the algorithm in this paper is compared with the traditional algorithm in different service request arrival rate environments. It is verified that the algorithm in this paper reduces the number of storage nodes and computing nodes and saves energy consumption on the entire network.

In the fog computing network, in order to improve user task execution efficiency under resource constraints, this paper proposes an efficient fog node resource allocation algorithm based on Taboo genetic algorithm. First, the resource constraint problem is modeled as a communication resource constraint and a computing resource constraint problem, and the performance constraint problem is modeled as a task execution delay constraint problem. Secondly, the advantages of taboo genetic algorithm are analyzed, and key processes such as chromosome coding, fitness function, selection process, taboo crossing process, and taboo mutation process are designed. Finally, an efficient fog node resource allocation algorithm based on Taboo genetic algorithm is proposed. In the experimental part, from the two aspects of the number of different tasks and the number of different fog nodes, it is verified that the algorithm in this paper significantly reduces the execution time of user tasks.

The rapid growth of the wireless sensor network scale and the number of services presents new challenges to the reliability of the network. To solve this problem, this paper proposes a network reliability optimization algorithm based on service priority and load balancing. The algorithm includes three steps: selecting the virtual network to be migrated, migrating the virtual network, and evaluating whether the life cycle of all physical nodes meets the threshold. The strategy of judging whether the life cycle of all physical nodes meets the threshold is adopted to effectively prevent the problem that some physical node resources are overused due to migration. In the simulation experiment, the algorithm in this paper is compared with the traditional algorithm, and it is verified that the algorithm in this paper has achieved good results in terms of effective node indicators.

In order to solve the problem of low success rate of virtual network resource allocation, this paper proposes a virtual network resource migration algorithm based on resource reservation under 5G network slicing. In order to improve the utilization of network resources through virtual network migration, this paper proposes a prediction method of the underlying network link resource demand to measure the resource demand for a certain link in the future. The urgency calculation method of the underlying link is designed, and a greedy migration algorithm is adopted to realize the resource migration on the underlying link. This paper proposes a virtual network resource migration algorithm based on resource reservation, and compares it with traditional algorithms through simulation experiments. It is verified that the proposed algorithm achieves good results in terms of virtual network mapping success rate and underlying network resource utilization.

In the 5G network slicing environment, in order to solve the problem of low utilization rate when a large-scale underlying network allocates resources to virtual networks, this paper proposes a virtual network resource allocation algorithm based on reliability in large-scale network environment. First, the reliability of the virtual network is modeled from the importance of the virtual network nodes. Second, the reliability of the underlying network is analyzed from the perspective of the community introverted character and the community relationship value, and the underlying network community division algorithm is proposed. Community reliability and node reliability are analyzed in terms of community and node reliability. Finally, a virtual network resource allocation algorithm based on reliability in large-scale network environment is proposed. In the experimental part, from the two aspects of the underlying network revenue and virtual network mapping success rate, it is verified that the proposed algorithm achieves good results in large-scale virtual network resource allocation.

In the 5G network slicing environment, in order to solve the problem of low success rate of virtual network mapping in the existing research, this paper proposes a resource allocation algorithm of power communication network based on reliability and historical data under 5G network slicing. First, the reliability of the virtual network node is analyzed from three aspects: the CPU resources of the virtual node, the connected link resources, and the centrality of the node. The reliability of the underlying network node is analyzed from three aspects: the reliability matrix of the underlying node, the CPU allocation history matrix of the underlying node, and the underlying link allocation history matrix. Secondly, based on the virtual network model and the underlying network model, a resource allocation algorithm of power communication network based on reliability and historical data is proposed. In simulation experiments, it is verified that the algorithm in this paper effectively improves the revenue of the underlying network and the mapping success rate of the virtual network.

In the 5G network slicing environment, in order to improve the utilization of underlying network resources, this paper proposes a resource allocation algorithm of power communication network based on mapping relation under 5G network slicing. The successful virtual network resource allocation case is modeled as a knowledge underlying to provide data support for the new virtual network resource allocation problem. According to the topology characteristics of the virtual network, the importance of the virtual node is analyzed to determine the priority of the virtual node when the virtual network resource is allocated. In the experimental part, it is verified that the algorithm in this paper has achieved good results in terms of virtual network mapping success rate and underlying network resource utilization.

In the context of 5G network slicing, in order to solve the problem of low service reliability, a high-reliability virtual network resource allocation algorithm based on service priority is proposed in this paper. The algorithm includes four steps: priority ranking of virtual network request and virtual nodes, reliability ranking of underlying nodes, resource allocation for virtual nodes, and resource allocation for virtual links. The priority ranking of virtual network request and virtual nodes mainly include two processes of ordering the virtual network based on service priority and ordering the virtual node based on centrality. In the step of allocating resources to virtual nodes, the main work is to allocate resources to virtual nodes based on the resources and reliability of the underlying nodes. In the simulation experiment, the algorithm in this paper is compared with the traditional algorithm, which verifies that the algorithm in this paper has achieved good results in terms of key business reliability indicators.

When using vehicles to hunt down criminal vehicles, the difficulty of chasing criminals is increased due to factors such as vision, surrounding environment and road conditions. In order to solve this problem, we have designed a real-time vehicle tracking system based on ROS. The system uses Pixhawk as the flight control platform in the hardware part. Its internal integrated attitude control module and altitude control module are mainly responsible for controlling the attitude of the aircraft to achieve the effect of stable flight of the aircraft. The GPS module is responsible for acquiring coordinate position data, so that the aircraft can achieve the fixed point effect. This system solves the problem that a processor has insufficient resources when processing relatively large data. Compared with using vehicles to chase criminal vehicles, this system can solve the problem of increasing the difficulty of chasing criminals due to factors such as visual field, surrounding environment and road conditions. This system greatly improves the pursuit efficiency.

This paper deals with the recognition of a relationship between the cold-resistant viruses by fluorescent staining, especially the recognition of Hurst characteristic of cold-resistant viruses based on the time series of differential normalized fluorescence indices, and derived consensus gene position maps. The complex covariance is calculated to find the kin relationship between the different coronavirus clade. Differential Normalized Fluorescence Indices (DNFI) is one of the most commonly used indexes to extract virus information by fluorescent staining medical images, widely used in virus classification and growth evaluation. In this paper, a novel method based on Hurst of time series of differential normalized fluorescence indices derived gene sequence is proposed, which considers the whole time series of differential normalized fluorescence indices generated various gene maps and is simple and practical.

At present, the traditional music therapy for autistic children requires professional music therapists to carry out auxiliary therapy on the side of the patients, judge the emotional state of the children, and manually play the corresponding music for treatment according to the current emotional state. Traditional music therapy not only requires professional therapist to keep an eye on it and rely on professional experience to observe and judge the effect of music therapy on autistic children, but also makes mistakes in judgment because autistic children do not show their inner emotional activities. The system uses the EEG acquisition equipment of EMOTIVEPOC+14 channel to collect the brain waves of autistic children. The median negative emotions of normal people correspond to the impulsive outward emotions of autistic children, calm emotions, and restrained autistic emotions to conduct preliminary experiments. After the self-EEG signal is collected, the detail component threshold is denoised by wavelet decomposition, the SVM algorithm is used to classify the positive, neutral and negative emotions. According to the principle of playing homogeneous music, autistic children finally have a calm state, so as to achieve the purpose of music intervention therapy. The system visually displays the identified EEG signals to the interface, which can feedback the emotional state in real time, so that the effect of music therapy for autistic children can be systematically evaluated. In this paper, the EEG data of normal people are used to verify the feasibility of the system, and the classification accuracy is 88.8%.

In order to improve the EEG recognition accuracy and real-time performance, a classification and recognition method for optimizing the penalty factor C and kernel parameter g of Support Vector Machine (SVM) based on Particle Swarm Optimization (PSO) algorithm is proposed in this paper. Firstly, the Regularization Common Spatial Pattern (R-CSP) was used for EEG feature extraction. Secondly, the penalty factor and the kernel function were optimized by the proposed PSO algorithm. Finally, the constructed SVM classifiers were trained and tested by the two class EEG data of right foot and right hand movements. The experimental results show that the recognition rate for EEG classification of the PSO-SVM is average 2.2% higher than the non-parameter-optimized SVM classifier, and it is significantly higher than the traditional LDA classifier, which proves the feasibility and higher accuracy of the algorithm.

Due to the nonlinear characteristics of EEG signals and the rhythm characteristics of motor imagery, the low recognition rate of using single feature extraction algorithm, a feature extraction method based on wavelet transform and fuzzy entropy is presented in this paper. The EEG signals are decomposed to three levels by the wavelet transform, according to the ERS/ERD phenomena during motor imagery, the alpha rhythm and beta rhythm signal can be extracted by the algorithm of fuzzy entropy. Finally, the motor imagery EEG signals are classified by a support vector machine classifier. BCI Competition IV Datasets1 has been used to conduct the experiment, the experimental results show that the feature extraction method combining wavelet transform and fuzzy entropy is much better than the ways of using single fuzzy entropy, sample entropy, or others, and its highest recognition rate is 90.25%.

Aiming to tackle the shortage of traditional white balance algorithms on face image, an eye white based algorithm is proposed under the assumption of stability of eye white in YCrCb color space. It first compares the eye white parts of input image with standard ones under normal white light source, to produce a gain coefficient, then corrects the color cast to the image. The experiments show that our method overcomes traditional counterparts both in subjective vision viewpoint and objective evaluation while retains good applicability and simple implementation.

With the vigorous development of marathon, running has been favored by people. While running brings people health, joint injury is also a major hidden danger in this sport. In order to study the characteristics of the joints of the lower limbs of the human body including the joint angle and joint torque during running, especially the effect of running speed on the above mechanical characteristics, this paper recruited 20 testers to complete periodic running at a running speed of 6 km/s–14 km/s. The Functional Assessment of Biomechanics System (FAB system) was used to collect kinematics and dynamic parameter data. On basis of the seven-body rigid body model, the Lagrangian method was used to establish the dynamic model of the human lower limbs to obtain the joint moments of the various links of the lower limbs. Through the above model and combined with experimental research, this paper concludes: running speed has different effects on hip, knee, and ankle flexion angles and extension angles. From the perspective of joint torque, the hip and ankle joints of the human lower limbs are significantly affected by the changes in running speed during running and are prone to injury. For runners, running performance can be improved by increasing swing leg torque and reducing swing leg inertia. This research has guiding significance for improving athletes’ level, designing and improving sports equipment and developing exoskeleton robots.

Accurately simulating the biomechanical parameters of human lower extremity movement is very important to analyze the biomechanical characteristics of human lower limbs. In this paper, the joints of the lower limbs of the human body are simplified to a seven-rigid body model in the sagittal plane, and the kinematics of the lower limbs of the human body and the dynamic model of the supporting period are established by the D-H method and the Lagrange method. In order to verify the above model, this paper builds a simulation model of human rigid body based on SimMechanics. Finally, comparing the experimental data of FAB measurement with the simulation data, it is found that the kinematics and dynamics models used in this paper are accurate enough to complete the measurement of the mechanical properties of the human lower limbs.

The medical gas system is a system engineering directly related to the life and property safety of patients and doctors. In order to ensure the safety of hospital patients’ gas use and the regular operation of related medical equipment, this paper introduces the analysis model (Convolution Weight Analysis Model, CWAM for short) of medical oxygen operation state in the medical gas system. It obtains the operation state regulation of the tank through CWAM model. Through the analysis of the operation status of the medical oxygen tank, we can adequately grasp the abnormal operation status, timely prevent, and ensure the regular operation of the hospital oxygen supply system.

The hospital logistics management is an integral part of hospital management, and safety management is the most important part of hospital logistics management. Intelligent and efficient standardized management mode is an essential guarantee for the safety production of hospital logistics. Therefore, intelligent security management of hospital logistics has gradually become the core of hospital management. The realization of this platform, on the one hand, strengthens the hospital to the logistics jurisdiction power equipment security monitoring management, enhances the Early Warning and prevention ability, improves the level of safety management in the hospital.

Tibetan medicine entity recognition is a primary task of medical entity recognition. Building a Tibetan medicine knowledge graph is the preliminary work of medical big data research. In this paper, Bi-directional Long Short-Term Memory (BiLSTM-CRF) is used for automatic recognition of Tibetan medicine entities. Then we construct a Tibetan medicine knowledge graph based on Nodejs. Entity recognition in medical documents and electronic medical lay a solid foundation for the research of medical big data.

Based on the POI data of medical facilities and residential quarters in Kunming in 2018, using GIS spatial analysis methods such as Kernel Density Estimation, Average Nearest Distance, and Direction Distribution were used to study the spatial distribution characteristics of Kunming medical facilities and propose optimization strategies. The results show that: (1) From the perspective of spatial structure, the spatial distribution of medical facilities in Kunming is relatively uneven, with the main urban in the city as a “single-center” cluster, there are four characteristics have emerged: 1) Urban crowding 2) The suburbs are empty 3) Old city crowded 4) New city zone is empty. (2) From the degree of regional agglomeration, the concentration of medical facilities in Kunming is high. The concentration of Special Hospitals are higher than that of Grade A Class Three Hospitals, General Hospitals and Community Hospitals. (3) From the perspective of spatial development direction, medical facilities develop along the “northwest southeast” direction, which is consistent with the direction of spatial expansion of residential quarters and the direction of urban expansion.

With the continuous progress of biotechnology and gene technology, scientists have found a large number of microRNA-related data through experiments. Among them, analyzing the characteristics, structure and function of microRNA by various methods is one of the most important applications in bioinformatics. To survey the mechanisms involving gene alteration and miRNAs in pancreatic cancer, we focused on transcriptional factors as a point of penetration to build an abnormal expression regulatory network. In this network, we found that some pathways with differentially expressed elements (genetic and miRNA) showed some self-adaption relations, such as SMAD4. A total of 32 genes and 88 microRNAs with 199 directed edges were identified. The evidence have shown that the expression of microRNAs varies with the severity of pancreatic cancer deterioration.

This paper presents a new method for modeling RNA secondary structures based on stochastic tree adjoining grammars. In order to get better predict results, we shows using Stochastic Tree Adjoining Grammars to inference RNA secondary structures. firstly, we expound the tree adjoining grammars and stochasitc model and the operations of TAG tree. Secondly, discusses the key problems for RNA secondary structure prediction using tree adjoining grammars modeling, finally we design a tree adjoining grammars model to predict RNA structure, The experiment choose 8 species sequence from EMBL database, and experimental verification of the validity of the model, the experimental results shows that the tree adjoining grammars in the prediction of RNA sequence structure has long-range correlation, can improve the prediction accuracy.