Advances in Artificial Intelligence and Security

In order to improve the 2D cutting stock under the condition of small batch and multiple batches production, we propose to predict the parts demand of subsequent batches first, and then take advantage of the idea of centralized cutting to integrate the predicted parts demand of multiple batches into a larger scale problem to solve. As a supplement, the shortages of actual parts demand caused by prediction error are settled by compensating cutting as it occurred. A model is built for that and it uses the long short-term memory (LSTM) neural networks to predict parts demand, and solve the integrated cutting problem by the classical method of column generation combined with strip construction. To check the effectiveness of this model, an experiment is exerted on it with some simulated historical parts demand data generated by Monte-Carlo simulation method. The experiment results show that the model predicts the parts demands of subsequent batches accurately, and achieves higher overall material utilization rate than that of cutting for each batch without considering use of surplus materials and inventory-based cutting approach proposed by other researchers.

In the process of construction and operation of mountain roads, slope disasters such as landslide and collapse are often encountered, which seriously affect the transportation infrastructure and safe operation in China. Cracks are the early symptoms of most slope diseases. By monitoring the change trend of cracks, the displacement trajectory of the slope body can be reflected in time, which is of great significance for landslide monitoring and early warning, so the safety detection is concentrated in this stage. In recent years, great progress has been made in deep learning-based computer vision methods, which have the advantages of simple observation method, low cost, wide detection area and sustainable monitoring. In view of this, a pixel level segmentation method of slope cracks based on deep convolutional neural network is proposed in this paper. According to the shape characteristics of slope cracks, a deep convolutional neural network was designed. The network was trained on the self-made slope image data set, and the IOU on the validation set reached 75.26%, which realized the precise segmentation and recognition of cracks. Experimental results show that the model has a good ability to characterize the slope cracks, can accurately extract the slope cracks, and provides a reliable basis for the formulation of slope early warning and disaster relief programs.

Driver fatigue is one of the leading causes of traffic accidents. At present, fatigue driving detection has disadvant ages such as low practical application effect and high equipment requirements. This paper proposes a multi-feature point non-invasive fatigue monitoring system based on a support vector machine with a hybrid kernel function. The system detects feature points through a gradient descent tree algorithm based on a cascaded regression and calculates the eye aspect ratio (EAR) and mouth aspect ratio (MAR). The heart rate is obtained through RGB image analysis combined with Euler’s video magnification algorithm. Classify facial features to get fatigued. This paper is based on the Logistic and Radial Basis Polynomial Kernel (RBPK) function to improve the support vector machine, which has better learning and generalization. Finally, this paper uses the Driver Drowsiness Detection Dataset and the author’s dataset to test. The classification accuracy rate for a single picture is 96.92%. In summary, the system proposed in this paper has a better recognition rate for fatigue driving detection.

AI plays an important role in COVID-19 identification. Computer vision and deep learning techniques can assist in determining COVID-19 infection with Chest X-ray Images. However, for the protection and respect of the privacy of patients, the hospital’s specific medical-related data did not allow leakage and sharing without permission. Collecting such training data was a major challenge. To a certain extent, this has caused a lack of sufficient data samples when performing deep learning approaches to detect COVID-19. Federated Learning is an available way to address this issue. It can effectively address the issue of data silos and get a shared model without obtaining local data. In the work, we propose the use of federated learning for COVID-19 data training and deploy experiments to verify the effectiveness. And we also compare performances of four popular models (MobileNet_v2, ResNet18, ResNeXt, and COVID-Net) with the federated learning framework and without the framework. This work aims to inspire more researches on federated learning about COVID-19.

As information security is constantly challenged, it is very necessary to construct a practical and secure new model of information hiding. The existing cover channel models often has the risk of being easily disturbed and destroyed, and the different characteristics of shared resources are needs to be utilized to guarantee its concealment. Due to the security and reliability, decentralization, robustness and other characteristics, Blockchain has been applied to information hiding field (steganography), however, its concealment and practicality are difficult to meet the actual demand. In order to overcome the defects of the covert storage channel and blockchain-based information hiding scheme, we propose a novel network covert channel model based on Blockchain transaction addresses parity and formalize this network covert channel under the blockchain environment. By modulating the parity of the transaction address, the sender can transmit the secret message to the receiver through adding addresses when processing the business without occupying additional block chain space. Meanwhile, the relevant technologies of Blockchain (cryptography, P2P, etc.) ensure that the scheme has good tamper-resistant, multi-line communication and receiver anonymity, guarantying the concealment of communication and the security of information.

In an image, how to quickly and effectively extract the useful regions named target regions in the scene according to the saliency features such as spatial domain, frequency domain etc. for further analysis of salient object detection is one of the challenging topics in the field of image segmentation. Most of the existing salient target detection methods use convolution network to extract high-order semantic features, combine pyramid pooling model to fuse high-order and low-order semantic features, and use Adam or SGD optimizer to optimize the model to obtain the salient object. However, the traditional convolution network model is not optimized for the model parameters, and finally redundant parameters will appear in the model, which will aggravate the training time and practical application detection time of the model. Although SGD is fast, it will fall into a large number of local suboptimal solutions or saddle points in the process of non-convex error function optimization. Adam has better performance, but the speed is slightly slower then t -> ∞ that will not have a good generalization performance. In order to solve the above problems, a new optimization strategy is proposed to compress the model. At the same time, AdaX, an optimizer with SGD speed and Adam performance, is used to optimize the model. Through the test on the open data set DUTS, ESSCD and etc., the proposed optimization model method reduces the parameters of the original model, and also improves the training speed and application detection speed of the model.

According to the rules and requirements of the handling project of the China Engineering Robot Contest, an intelligent handling robot with Freescale KL25 chip as the core controller and capable of automatically identifying multiple colors was designed. Based on the completion of the hardware circuit and mechanical structure, the robot’s trajectory on the field is planned through software programming. Establishing a handling strategy allows the robot to efficiently grasp, transport, and stack objects. The focus is on the infrared tracking module, the color recognition module software design and the robot path planning strategy research. After the completion of the robot construction, through experiments in various links, the total number of loops compared to the traditional handling scheme has increased by about 11.1%, and the running time has been reduced by about 41.5%. The effect is good.

The fast cache could be used in storage clusters to alleviate load imbalance caused by highly-skewed requests between storage nodes. In a smaller cluster, we can use a single cache node to solve the I/O bottleneck caused by load imbalance. However, in a Large-scale cluster, we may need more than one cache node to afford enough capacity, which brings a new load balance problem in cache nodes. DistCache successfully solved this problem by applying the power-of-two-choices. In the above storage clusters, cache nodes cache the hottest objects while ignoring the size of objects, which leads to poor performance when meeting objects with variable sizes. We present SACache, a size-aware mechanism for large-scale storage clusters, which can improve I/O performance by maximizing the benefit of the unit cache. In this mechanism, we set an object admission filter to filter out objects with lower caching benefit. To adapt to changing request patterns, we record recently requested objects and their size, then replay those requests periodically in a cache simulator to find the best cache admission parameter using a greedy algorithm and apply it to the object admission filter. We apply this mechanism in a prototype distributed storage system. Experimental results show that it can increase the system’s overall bandwidth when the object’s size is different.

With the development of mobile edge computing, more and more services are moved to the edge of the network, and devices there are usually with low computational abilities and little storage resources. To make it lightweight and elastic, containers can be adopted in the edge environment when migrating a certain application. With the host OS kernel shared, applications can be deployed with the least computational resources they need, making it possible to deploy more of them on relatively low-end devices. Migration is also used in scenarios like maintenance or load balance, etc. We noticed that in edge environment, devices and servers are usually built with heterogeneous Instruction Set Architectures (ISAs) processors. X86 processors are widely used in desktop PCs, laptops and servers while smart-phones are built with an ARM processor, which leads to a serious problem that a container cannot be migrated to a heterogeneous machine to continue running directly. In this paper, we firstly give an overview of heterogeneous-ISA migration, and its applications and techniques. Then we discuss the existing heterogeneous execution solution from the perspective of applicable scenarios, latency, power consumption, requirements for computational resources, etc. Next, a comparison study is given on each of the characteristics to depict the details and differences in existing works. At last, challenges and open research issues which are waiting for further studies on container migration are listed.

As an important technology of the quantum detection, the quantum synchronization detection is always used in the detection or measurement of some quantum systems. A detection schemes are key to the study of quantum system. The quantum synchronization detection which is presented between the probe and the quantum system is always used in detection application. A probing model is established to describe the probing a qubit system in the cavity field and to reveal the effect of the environment (cavity) on the quantum synchronization occurrence as well as the interactions among environment, a qubit system and probing equipment. By adjusting the frequency of the probe, the in-phase and out-of-phase synchronization can be achieved. So the information of the probe can be used to obtain the quantum system. Simultaneously, the effect of $${\gamma }_{3}$$ γ 3 which describes the interaction strength between the probe and environments for quantum synchronization is discussed under different the values of Ohmic dissipation index s. Finally, the machine learning method is applied to present an optimization for classification and regression of synchronization transition dependent on s and $${\gamma }_{3}$$ γ 3 . This opens the way for studying the generalized form of quantum synchronization through machine learning algorithms (Artificial neural network) in the future.

Most of the hidden dangers of network system security are caused by group events. Group analysis and data mining for them are of great significance to ensure network security. Although the existing group detection algorithms have achieved a series of results, they can only be divided on one of the network structure and group attributes, but cannot combine them together, which has certain limitations. The comprehensive vector can be constructed by collecting and mining the group data which cause the hidden danger of security, which can analyze the hidden danger of security from the aspects of network structure and node attribute, so as to realize the guidance and control of group behavior. Therefore, in view of the above problems, this paper proposes a group detection algorithm based on synthesis vector, which can finally find a special group which is closely connected in structure and very similar in attribute. Firstly, the comprehensive similarity is calculated based on the fusion vector in the sharing layer of the comprehensive vector computing model. Then, reconstruct the weighted network diagram. Finally, based on Louvain algorithm, the improvement is carried out. The improved algorithm is referred to as the W-Louvain algorithm. The W-Louvain algorithm is used to divide the groups, and the closely connected vectors in the structure and the very similar vectors in the attributes are divided into the same group. Experiments show that on multiple datasets the evaluation indexes of W-Louvain algorithm, such as modularity Q, number k of community, density D of community and similarity degree S of comprehensive vector attribute, are better than the existing methods.

The application and promotion of smart distribution transformer terminal based on software definition has played a positive role in promoting the construction of distribution Internet of Things. This paper presents a new non-negative matrix factorization algorithm, generalized projection non-negative matrix factorization algorithm. Based on this algorithm, the statistical monitoring model is constructed, and the monitoring statistics suitable for the new monitoring model are designed. Then, the monitoring model is deployed in the form of a software APP to the smart distribution transformer terminal to realize the operation state monitoring and fault detection function of the distribution network. Finally, using the Simulink in MATLAB as the simulation plat-form to simulate the single-phase grounding fault of distribution network, the result shows that the fault monitoring model based on generalized projection non-negative matrix decomposition can better complete the detection task of single-phase grounding fault, and which detection effect is to meet the real-time requirements of fault detection in the field.

Nowadays, more and more people realize the importance of patent for innovation activities. Patent citation network analysis is one of the most important methods for patent measurement, patent mining, and core patent identification. In nowadays, finding technology trajectories and analyzing major technologies in patent networks are intensively used in technological competition. Main path analysis (MPA) is a famous directed graph-based method to extract main paths in certain networks, such as a citation network. However, the accuracy of main path identification may be distracted due to a large volume of wrong references when using MPA in patent citation networks solely. To tackle this challenge and extract reasonable main paths from patent citation network, in this paper, we combined the classic MPA with the PageRank algorithm and we tested this new combined method on authorized patent datasets. The results show that the improved method achieved better performance in average cited frequency and other indicators of core patents comparing with traditional MPA.

The blockchain system can provide a trust infrastructure for sharing data among untrusted parties. However, storing the original shared data directly on the blockchain is not suitable for large-scale data sharing scenarios. Therefore, we designed a data sharing system architecture in which data hashing and response records are stored on the blockchain and the original data is stored in the off-chain database. This architecture can alleviate the system overload and protect privacy problems to a certain extent. This paper proposes a three-tier system structure to ensure the function of the network. Subsequently, formulate request rules, deploy smart contracts, and build a platform based on the alliance chain. Finally, the system functions and performance are analyzed and compared through experiments. The results show that the system can realize efficient and transparent information sharing while satisfying on-chain and off-chain collaborative management, and the system has certain advantages in function, overall performance and throughput performance.

System Wide Information Management (SWIM), as an advanced civil aeronautical information management method, aims to solve the problems of global civil aeronautical information systems such as difficulty in obtaining common data in a timely manner and high information exchange costs. SWIM was first proposed by Europe and the United States and was recognized and valued by the International Civil Aviation Organization. It is to ensure the correct information transmission at the correct time. In order for the consistent information in the “virtual information pool” to be efficiently, accurately and securely transmitted, the information and data in the SWIM environment need to be defined in detail and standardized, so the two sides of data interaction can maintain the consistency in syntax and semantics. The Aeronautical Information Exchange Model (AIXM) is a core standard model for data transmission and format conversion, which mainly involves information in the aeronautical intelligence field. It covers multiple thematic elements such as airspace, airports and air routes, and provides a standardized description for data conversion and transmission in this field. The article analyzes the model composition of AIXM in the SWIM information exchange model, the modeling process, and the key technologies involved in model establishment, and preliminary design and implementation of AIXM.

Natural scene images often contain a lot of important texts, which carry the information we need, so it has important practical value to locate text information. This paper proposes a Cascade Classifier Text Localization (CCTL) algorithm. Firstly, a cascade classification algorithm based on Real AdaBoost is proposed to improve the accuracy of text localization. Secondly, a perceptual-based grouping algorithm is proposed to establish a perceptual organization framework. The use of adjacency and similar rules to group texts can improve the effect of text grouping. The proposed algorithm is verified and evaluated on the ICDAR dataset. At the same time, it is compared with other algorithms. The experimental results show that the proposed method has superiority in natural scene text detection.

Big data is built on the foundation of today’s horse racing technology. The rise and development of big data has become one of the most typical features of the industry’s Internetization in the new IT era. The data stored in the world has overflowed every database, and industry big data has become the focus of attention of all walks of life. Governments at all levels and all enterprises and institutions hope to dig out high-quality, high-value-added data from big data. Information, and use it to improve their profit and service model, and enhance their status in the public and industry. In the era of big data, for libraries that provide teaching and scientific research services to the teachers and students of the school, they can use efficient and fast information technology, such as big data analysis, to conduct in-depth analysis and analysis of various data in university libraries. Digging, effectively understanding the reading situation of the whole school, and analyzing their reading hobbies and reading habits, the library can effectively use this information to improve the types of books purchased by the library, which can more effectively increase resource utilization.

In order to reduce the impact of traffic incidents on traffic operation, an Automated Traffic Incidents Detection (SVM-AID) algorithm based on Support Vector Machine (SVM) is proposed. This algorithm is of great significance for improving the efficiency of traffic management and improving the effect of traffic management. This article first introduces the background of the topic selection of the Traffic Incidents Detection algorithm, the research status at home and abroad. Then it focuses on the Optimal Separating Hyperplane, linear separable SVM, linear inseparable SVM, nonlinear separable SVM, and commonly used kernel functions. Then, the design flow chart based on the SVM-AID algorithm is given, and the principle component analysis method, Normalization Method and the selection method of Support Vector Machine parameters are introduced. Finally, using the processed data, 4 experiments were designed to test the classification performance of the SVM-AID algorithm, and the influence of each parameter in the SVM on the classification effect was analyzed. The results of the final experiment also showed us the design The effectiveness of the SVM-AID algorithm.

A lightweight buffer protection material was prepared by compounding crop straw biomass waste with secondary fiber, which was used for transportation and packaging applications. By selecting different kinds of crop straws and changing the proportion, the material performance test results show that the system has good buffer performance. Under the same content, the mechanical properties of straw and secondary fiber composite buffer material are better than that of industrial hemp stalk core material. Under the same filling conditions, crop straw/secondary fiber composites have good cushioning and protection performance. This study is of great significance for exploring the application of crop straw/secondary fiber buffer protection materials in logistics and transportation, and developing new environmental protection packaging materials for replacing foam plastic products with biomass wastes, providing reference for the application of logistics and transportation industry.

As data analysis technology becomes more mature and data applications become more and more extensive, people have accumulated a large amount of data. Data analysis technology has had a positive impact in the education field. In English learning, English vocabulary is the key to learning English and the basis for learning English. But the main obstacle to learning English is the memory of English words. Therefore, this research attempts to combine the ten-day recitation method with Ebbing Haus' forgetting curve to assist the memory of English words. Through the grey relational analysis of data analysis, the feasibility and effectiveness of this method are studied. The data used are recorded data in the memory word test, word familiarity rate, fuzzy rate, forgetting rate, test word score data and final word learning score data. Store English word learning records in a database or data warehouse in the form of data. Using these data to reasonably analyze the ten-day recitation method is very helpful for improving English vocabulary learning performance.

Since the rapid growth of China’s population at the beginning of the 21st century, a phenomenon has emerged in most cities: “It is difficult to take a taxi”. Most citizens go to and from work every day, and it is difficult to get a taxi in time, especially when it is rainy or snowy, or when it is a holiday. “Taxi difficulty” has become an issue that needs urgent attention in a city. Following the sustainable development concept advocated by the state, how to improve the utilization efficiency of cars in cities has become a big problem that the relevant departments need to consider. First of all, we think of using dynamic car sharing to optimize the utilization rate of cars, so as to complete the improvement of urban traffic, improve the efficiency of urban transportation system, reduce traffic congestion, fuel consumption and pollution, etc. Secondly we believe that with the implementation of dynamic car sharing with Hungarian Method the efficiency is maximized.

Due to the complexity of the situation in real life, the network topology may be missing, which will cause great obstacles to troubleshooting persons. Because of the complexity of network topology, there are some relations between the alarms of network elements with topology. In response to this problem, a classification algorithm to predict the topological relationship between network elements by alarms is proposed in this paper. Firstly, extract the data features, then perform correlation item mining on the original alarm data, and then use the alarm correlation items of the mined network element pairs as the input set of the classification algorithm to obtain the classification result of whether there is a topological relationship between the two network elements. The result of the maximum correct rate in the test set is 91.16%, and the maximum AUC is 0.95. It can be seen that the classification model established by this method has excellent performance and provides a good idea for completing the network topology.

With the advent of the IoT era, more and more IoT device has begun to integrate into our daily life. Effective recognition of all kinds of IoT device is the premise to ensure the safety of IoT devices. Furthermore, if a certain device can be identified more finely, such as brand and model information. Then it can be associated with the device vulnerability library to obtain more detailed device security information. According to the existing research, the richer the feature lexicon of IoT devices, the finer the granularity of device recognition will be. The paper is based on word2vec technology to expand the IoT device recognition feature vocabulary. Firstly, the keywords carried by the device information are used as the input of the search engine, and the corpus of keywords is constructed. Secondly, we use word2vec to model the corpus, and then extract the known IoT device feature words related to keywords from the model. Finally, based on the binary correlation judgment algorithm proposed in this paper, we judge the correlation between keywords and feature words in turn. If the correlation is consistent, keywords will be added to the corresponding feature lexicon.

With the continuous development of modern network technology and data informatization, information security has been paid more and more attention by the society, and coverless information hiding has been becoming a new and important research field of information security. At present, the capacity of text coverless information hiding is very limited, generally 3–4 words per text. In order to increase the capacity of text coverless information hiding, this paper proposed a novel method of text coverless information hiding. In this method, the punctuation combination modified by Morse code is combined with the use of starter, list and loop statement to realize the hiding of a number of words in a normal text without modified. The experiments’ results show that this new method can hide more words and express all the words composed of representable characters.

The CPI index is the final price of social products and services. It is an important indicator for economic analysis and decision-making, monitoring and control of the overall price level, and national economic accounting. Because the CPI index is originally a univariate data indicator, in order to better predict the economic impact of the new coronavirus pneumonia by analyzing the CPI index, in this paper we chose the gray prediction model that has an excellent predictive effect on univariate data. Firstly, we got all CPI data from 2015 to 2019 from the Oriental Wealth Data Center. Secondly we predicted the CPI index for January, February, and March 2020 by using the gray prediction model. Thirdly, the trend of the forecast data was visualized with Python. Finally, the predicting result was compared with the real data, and the reason for the difference was analyzed.

The star sensor is a high precise attitude measurement equipment for the spacecraft, and star identification is one of the key techniques of star sensor’s attitude determination. Star tracking mode is the main working mode of the star sensor. However, during star tracking, when the star sensor is in the harsh environment, the star imaging is seriously affected, which leads to the decline of star attitude determination accuracy and even the failure of star identification. The star sensor re-enters the “lost in space” mode, resulting in navigation interruption. In this paper, the pure digital simulation method is adopted. Firstly, the SAO guide star catalogue is properly processed, and then we used it to generate star maps. On this basis, a comprehensive noise model of harsh environment is established by analyzing the generation mechanism of the transient effect, plume effect and the false stars which like star. At the same time, Gaussian noise and salt noise are used to pollute the star image. Finally, the joint denoising algorithms such as threshold segmentation, inter frame star image comparison are used to process the star images. The experimental results show that Gaussian noise and salt noise can simulate the influence of space harsh environment on star sensor imaging well. The proposed joint denoising algorithm can restore the star image to the pure mode as much as possible.

Road intersection is one of the crucial elements in a road network. How to accurately detect road intersection is a vital process in road network construction. The traditional approach finds the turning point in trajectory by computing the trajectory point’s turning angle and uses a clustering algorithm to cluster the turning point to yield the cluster point representing road intersection. However, the approach will detect some fake road intersection points. We propose an approach combining turning point detection with grid transformation for road intersection detection to address this issue. The process of grid transformation can filter the point in low-density area. So, the fake intersection point can be filtered by the process as far as possible. We use a real-world trajectory data set to conduct the experiment and receive a good result. The precision of our approach’s experimental result is 14% more than the traditional approach’s result, and the recall and the F-score are also higher than the traditional approach.

In order to promote the structural reform of agricultural supply sides, speed up agricultural modernization and improve the cultivation of new professional farmers, it is necessary to comprehensively evaluate the effect of this new professional farmer training. In view of the problem that the traditional evaluation method has great artificial influence when determining the weight in the evaluation process, the nonlinear mapping ability of BP neural network is used to reduce the subjective influence of people giving weight in the evaluation process, and the error is distributed to all units of each layer through gradient descent to complete the weight update. A classification model was created for this experiment, and 156 groups of data were used for training and 40 groups were used for testing. The classification accuracy on the test set reached 94.87%, so the training effect can be effectively evaluated by BP network. By averaging the weights of individual evaluation results obtained by the classification model, the comprehensive evaluation of this training are obtained finally. According to the comprehensive evaluation results, we can promote training institutions to improve relevant measures and promote the education and training of new professional farmers.

In order to make a comprehensive evaluation of the training effect of the new professional farmers from the training organization and management, training time, training cost and other aspects, and make an accurate judgment on the current situation and development direction of the new professional farmers training in China. Based on the cloud model theory, this paper realizes the transformation of qualitative index and quantitative evaluation, and organically combines fuzziness, randomness and discretization. After the characteristic data of the normal cloud generator is obtained by the reverse cloud generator and the information weight method, the normal cloud image is generated, and the training effect is evaluated according to the cloud image. The evaluation based on cloud model not only takes into account the fuzziness and randomness in the evaluation, but also takes into account the scientific nature of the data and the intuitiveness of the evaluation results. To a certain extent, it overcomes the shortcomings of the traditional evaluation methods, such as unstable evaluation conclusions and complex calculation. The evaluation conclusion is helpful for researchers to make an accurate judgment on the current situation and development direction of the training of new professional farmers in China.

In 21st century, the development of information technology has led to an overall improvement in people’s material level. Especially the big consumer group of college students. Campus information technology has now become a trend, which means that there is a large amount of consumer data accumulation. Behind the huge amount of behavioral data, there is potentially valuable information. By analyzing the data, we can understand the characteristics and needs of the consumption behavior of college students, so that we can better promote the process of campus informationization and help campus management.

Nuclear reactor computing software need to process and maintain complex and massive data sets. In order to meet the requirement of mass data storage and processing in software, the data storage model and I/O method and programming interface of HDF5 (Hierarchical Data Format v5) are deeply studied. According to the data storage and processing requirement of typical nuclear reactor computing data such as assembly or core neutronics computing data and core thermal-hydraulic computing data, RCDF-H5 (Reactor Compute Data Framework based on HDF5) is proposed. KYMRES (KYlin-2 Main RESults databank) and COMRES (COre Main RESults databank) are designed and implemented based on RCDF-H5. The performance tests show that RCDF-H5 has a higher I/O efficiency than conventional storage solutions. A new data storage and management solution for nuclear reactor computing software is provided.

Code comments can improve the readability of program and help the programmer to efficiently promote the maintenance of the software. Therefore, the code comment generation task has important research significance. However, generating high-quality code comments is very challenging, because the function methods called in the code hide a lot of functional information, which is often related to the functionality of the source code. Therefore, we propose to use the function description information to improve the code comment generation model. In the paper, we have designed two models that combine code sequences and API description information to implement the code comment generation task, and conducted experiments on two open source data sets. The experimental results demonstrate the effectiveness of API function description information for code comment generation task.

Volunteer geographic information (VGI), as a new geographic information source, has gradually become an important complement for authoritative data sources. Yet, due to the non-professionalism and spontaneity of most volunteers, and the lack of effective contribution constraints, the issues of attribute quality of OpenStreetMap (OSM) objects, such as semantic inconsistency and incompleteness, have received many attentions. However, there are few methods for effectively improving OSM’s semantic quality. In this paper, we proposed a tag recommendation method based on the combination of FP-Growth and the Markov process. The aim of our algorithm is to improve the quality of OSM objects by recommending some tags when volunteers contribute to the platform. The results show that the method based on FP-Growth and Improved Markov Process can effectively recommend tag-keys for different feature classes.

At present, data are generated all the time in the world, and these data contain great value. However, these data have problems such as huge scale, complex type, fast flow rate and low value density. Therefore, the research of big data analysis means and tools has increasingly become a research hotspot. On the basis of research at home and abroad, this paper attempts to analyze the definition, framework and typical big data processing systems of big data, especially focuses on the analysis and comparison of batch data processing system, stream data processing system, hybrid processing system and graph processing system, and obtains the characteristics, processing mechanism and applicable occasions of each system. The paper hopes to provide some references for understanding big data systems, solving problems in the process of big data processing and developing big data applications, and provide reference for improving the effectiveness and efficiency of data processing.

PHM technology has played more and more important in the current mechanical engineering and acted as an intelligent solution to improve the availability of manufacturing systems. PHM consists of system health monitoring, feature extraction, fault diagnosis, and fault prognosis through remaining useful life estimation. With the help of various fault models, artificial intelligence algorithms, monitoring diagnosis, prediction techniques, PHM also uses a large number of condition monitoring data and information to improve the safety and working life of mechanical devices, minimizes the impact of device failures, and avoids the major accidents caused by the mechanical device malfunctions. In this article, several classical PHM surveys and reviews central to nowadays mechanical engineering are studied, ranging from typical PHM frameworks, practical PHM solutions, to concrete PHM approaches. And these related insightful literatures are surveyed and introduced, which could help more readers to better understand the PHM technologies in reshaping the modernization of current mechanical engineering

Multi-sensor data fusion is an emerging technology, which has been widely used in medical diagnosis, remote sensing, inertial navigation and many other fields. What’s more, the implementation and application of automatic driving system rely heavily on target detection technology. Due to the high mobility and unpredictability of vehicle-mounted equipment, for automatic vehicles, it is arduous to achieve real-time and accurate vehicle target detection by a single sensor means, thus it is difficult to reliably guarantee the safety and stability. This paper proposes a novel object detection method based on a multi-sensor fusion mechanism, which considers the real-time sensing data from two types of sensors including radar and camera. It collects multi-vehicle speed and position information efficiently and reliably. Then, it filters and integrates data according to Extended Kalman Filter, Data Association Filter and some other methods. Furthermore, vehicle-borne equipment makes intelligent decision based on the data. In addition to theoretical support, the designed simulation results also show that the multi-sensor fusion mechanism can detect target vehicles efficiently and accurately, and it has superiority in the stability and accuracy of perception than single sensor sensing method.

Automatic recognition of crop growth period is one of the core parts of precision agriculture support technology. In order to identify different growth periods in real time and obtain crop growth information, a crop growth period estimation method based on fusion features is proposed. First, the crop images are preprocessed to filter out the noise. Then the HOG features, SILTP features and color features are fused. Finally, XQDA is used to measure the similarity to classify and identify the growing period of crops.

Relying on the big data platform, based on data mining technology and label profile technology, this paper analyzes the distribution network equipment data in two typical areas of Changsha and Shaoyang in the past two years, and establishes equipment with five dimensions of attributes, benefits, operation, cost, and life. Characteristic index library; analyze the three application scenarios of operation status evaluation, operation benefit evaluation and investment decision support, and build a multi-angle and comprehensive intelligent holographic portrait label system for distribution network equipment; finally select a distribution transformer in Changsha City As an application case, use Python programming combined with machine learning algorithms such as cluster analysis and decision trees to display the holographic image of the device.

International students are a special group in Chinese cities, and their tourism image perception has important research significance for urban tourism construction. Based on the destination tourism image model proposed by Baloglu and McCleary, the tourism image of Nanjing is divided into three aspects: cognitive image, emotional image and overall image. Aiming at the international student group in Nanjing, the questionnaire survey was used to obtain the perception data of the group’s tourism image in Nanjing with SPSS. The empirical research on the tourism image of Nanjing was carried out from three aspects: tourism image perception factor, overall image perception and representative attractions. The study found that in the structured part of the dimension, international students are most satisfied with “urban environment” and least concerned about “attractions”. Through the single factor measurement, the most satisfied one is the rich historical and cultural heritage, and the most dissatisfied one is the crowded urban environment. In the unstructured part of the international students on Nanjing tourism image, the overall satisfaction is high, in which the destination image perception of the “delicious food” is the best. In the perception of urban tourism image, the perception of “a beautiful city” is the best. The analysis of the characteristics of social population structure shows obvious individualized difference. Boys tend to Zhongshan mausoleum, and girls are more inclined to Confucius Temple Qinhuai scenery belt.

Anomaly detection plays an important role in big data, which deals with high dimensional data effectively and quickly. In this paper is going to propose an anomaly detection method based on isolated forests. In the proposed method, the original data set is divided into different initial static blocks to calculate the intra-block density and mean density, respectively; after calculating the intra-block density of static blocks, the data set is simplified with the mean density of the original data set as a threshold (delete such blocks with intra-block density which is bigger than the mean density); for blocks with lower intra-block density, we construct isolated forests with the recursive method of nodes; the corresponding features are extracted and normalized to calculate the spatial position distance between the cluster center point and other points; based on density and distance, the anomaly scores are added to compare with the corresponding threshold. This method can effectively improve the accuracy of anomaly detection algorithm, and greatly reduce the actual amount of data in the process of anomaly detection. Furthermore, our method could save a lot of computing resources, and improve the efficiency of anomaly detection, and enhance the robustness of anomaly detection algorithm.

In recent years, the emergence of a large number of venture firms has brought great profits to venture capital firms. However, it is not easy to identify venture firms with investment prospects. Therefore, based on frequent item sets, this paper mainly mines the enterprise text information to identify the venture enterprises with investment prospects. Firstly, we use TF-IDF algorithm to extract keywords from enterprise text; Secondly, the word2VEC model is used to vectorize the text keywords, and cosine similarity is calculated with the word vectors in the keyword database; Finally, we use the Apriori algorithm to find frequent item sets and generate association rules, complete vector weighting calculation of combination keywords, and finally retain the first three words or phrases with the highest weight as the identification keywords of the enterprise, thus determining whether the enterprise is a risk company with potential investment prospects. Experimental results show that the proposed method is effective.

Nowadays, with the popularization of electricity, the status of grid business office in people’s lives has become more and more important. Grid business office often have a large flow of people, so it becomes very necessary for abnormal detection of grid business offices. Traditional video surveillance has a lot of problems. This paper uses various computer vision technologies to improve traditional video surveillance. First, the YOLO v3 algorithm is introduced to detect the number of people in the business office, and then the motion foreground extraction algorithm is used to calculate the contour of the human body. The contour judges whether there is an abnormal situation currently, and finally introduces the concept of image entropy to judge whether there is intense movement currently. Finally, in the end of the paper, the current work is summarized and the future word is prospected.

The existing offline business halls are unable to accurately identify customers for accurate marketing, and lack effective detection mechanisms for some abnormal behaviors, and cannot assist security personnel such as security personnel in the security work. The successful application of face recognition technology in the payment field makes it feasible to apply it in offline systems of business halls, and the development of behavior detection technology based on computer vision also makes us actively explore its application in offline scenarios of business halls. Therefore, we researched the application of intelligent detection technology including dynamic face recognition technology and behavior detection technology in offline systems of business halls, designed and implemented a prototype system based on this technology, and tested it in real scenarios. The experimental results show that it is feasible to apply the intelligent detection technology to the offline service system of the business hall.

Natural plant dye is selected as the pH sensitive agent of the freshness indicators, and the solvent is used to extract the dyes from six kinds of plants-Xinlimei radish, mulberry, red raisin peel, pitaya peel, red carnation and purple onion peel. The color change in solution with different pH value shows that most plant dyes are very sensitive to change in pH value-red in acid solution and blue in alkaline solution. According to the pH change of the spoilage process of chilled meat, low-temperature smoked sausage, fruits and vegetables, appropriate plant dyes are selected as the freshness indicator agent. The test result shows that Xinlimei radish dye is more suitable as a pH indicator agent for chilled meat and high-acid fruits. The purple onion skin pigment indicator agent can be used for fresh fish and neutral vegetables. The mulberry pigment indicator agent can be used for medium-acid fruits. And the theoretical design of the natural color food freshness indicator label is given.

The progression of the global COVID-19 epidemic situation is the main focus of attention of all countries in the world. Due to characteristics, such as multi-origins, huge amount, and inaccessibility, of the existing data, an all-round analyzation of the epidemic situation, which is in dire need, is impeded. The aim of the following study is to provide a multi-dimensional analysis of COVID-19 through visualization and dynamic simulation of data. In order to achieve this goal, the study collected related data though multiple platforms and used tools such as Echarts and Java Swing to visualize the data, and then dynamically simulated the transmission model. Moreover, the data of Wuhan has been applied to the SEIR model to study the effect of quarantine on the transmission of COVID-19. Ultimately, the study hopes to demonstrate an effective method of data analyzation that can be applied to prevent and contain similar outbreak in the future.

With the development of intelligence and data construction of electric power system in China, a large amount of data accumulated by electric power enterprises provide a data base for the fine prediction of electric power consumption. High-precision power forecasting model has far-reaching influence on urban planning and construction, smart grid development and so on. The problem of power big data privacy leads to the phenomenon of Power Data Island, which leads to the shortage of the accuracy of the current power forecasting methods. Federal learning breaks the long-standing Data Island phenomenon in power industry, and satisfies the privacy and security requirements of power data in application. This paper presents a high-precision federal learning method for enterprise electricity consumption forecasting, which takes into account weather conditions and enterprise tax information. Based on the FATE platform, this method combines the enterprise electricity consumption data with the tax data, and uses the third party coordinator to conduct encryption training, then constructs the enterprise electricity consumption forecast model based on the SecureBoost algorithm. The simulation results show that the federated learning method can effectively improve the accuracy of enterprise electricity consumption forecasting model.

Raft’s good performance in the CFT system makes it a mainstream implementation solution for the consistency of the private chain system. However, we found in practice that Raft’s performance degradation is very serious when the network latency is greater than 1 ms. Therefore, we simulated the blockchain network environment and tested the TPS performance of Raft and Multi-raft. The results show that when the TPS is the same, Multi-raft reduces network traffic by 30% compared with Raft’s leader, and the CPU load increases slightly. Based on the experimental results, we discussed the optimization scheme of Raft on the geographically distributed system, and provided ideas for future research.

In a non-cooperative game, the Nash equilibrium is computed by the game players with the payoff matrix. With the obtained Nash equilibrium, game players can analyze equilibrium strategies and make optimum decisions. In cloud computing, the computation of Nash equilibrium may be outsourced to the cloud due to the constrained resources of players. However, the payoff matrix may be sensitive and needs to protect against the cloud server. In this paper, we propose a cloud-based framework to secure outsourcing the task of computing mixed-strategy Nash equilibria. In our framework, the payoff matrix is encrypted with additive homomorphic encryption before uploaded and stored in the cloud. By combining secure multi-party computing techniques, we enable the cloud server to compute Nash Equilibria on the encrypted data without disclosing sensitive users’ data. We conduct experiments to verify the effectiveness, evaluate the precision, and then analyze the computational complexities.

Sharing data through clouds has never been more economical and easier than now. To guarantee the confidentiality of the data stored in the cloud storages, the data owners should encrypt their sensitive data before uploading them to the clouds. But, traditional encryption paradigm makes it difficult for flexibly sharing encrypted data between different users. The paradigm of proxy re-encryption (PRE), which can securely delegate the decryption right from one user to another, offers an effective solution to the encrypted data sharing in the clouds. To share the encrypted data with multiple users efficiently and securely, we extend certificate-based PRE into the multi-receiver setting and put forward the notion of multi-receiver certificate-based PRE (MR-CBPRE). By using MR-CBPRE, a data owner can securely distribute his encrypted data to a group of users though public clouds in an efficient manner. We first formalize the syntax and security definition of MR-CBPRE, and then design a concrete MR-CBPRE scheme. In the random oracle model, the proposed scheme is proven to be chosen-ciphertext secure. To demonstrate the merits of our scheme, we analyze its performance by comparing it with the previous certificate-based PRE schemes which consist of only single receiver. As far as we know, it is the first certificate-based PRE scheme in the multi-receiver setting to date.

In the era of the Internet of Everything, the number of terminal devices connected to the network is increasing day by day, and the amount of data at the edge of the network has increased dramatically. Under this background, network applications have put forward higher requirements on network bandwidth and network delay. The centralized cloud computing model can no longer meet network requirements. Edge computing emerged as a supplement and extension of cloud computing. To ensure the security of edge computing communications, efficient and reliable identity authentication technology is particularly important. To this end, first introduced the generation of edge computing, and at the same time analyzed and elaborated several major identity authentication technologies, combined with the demand characteristics of edge computing, analyzed the development direction of identity authentication technology in the edge computing environment.

With the rapid development of information technology and wireless communication technology, SQLite database has been widely used in various occasions in peacetime. SQLite database, as a relatively common database in the Android operating system, usually contains a series of key information such as call records and short messages. Research on SQLite database has a certain positive effect on public security electronic evidence collection. This article hopes to study the security mechanism and cracking method of SQLite database, the encryption and decryption mechanism and operation method of open source software such as SQL Cipher, and the actual forensic operation of WeChat as an example to conduct the research of SQLite database forensic analysis.

In cloud computing, since the program only runs in the cloud, it can be written in a programming language and compiled in a specific target platform. Due to the heterogeneous nature of the edge node platform, many tasks are migrated from the cloud to the edge terminal. So it is not easy to realize the edge computing programming, and the maintenance cost is also high. In order to solve this problem, this paper designs a set of software definable programming strategy through Hypertext Marked Language (HTML) format (independent of the operating system) file, and realizes the programming of the fireworks node of Internet of things (IoT) gateway under the edge computing, which is applied in the remote automatic monitoring system. Compared with the same system under the cloud computing, it is convenient for the user's programming, enhances the real-time performance, and greatly. It reduces the communication data flow, saves the data transmission cost, relieves the overall pressure of the system, and reduces the power consumption due to data transmission data storage.

Boolean functions have been extensively studied in coding theory, cryptography, sequence design and graph theory. By adding two products of three linear functions to some known bent functions, in this paper, we construct a class of bent functions and obtain their dual functions. In the meantime, a class of semi-bent functions and some classes of five-valued Walsh spectra are given.

Formal verification technology based on theorem proving assistant is the only way to strictly guarantee the correctness of the program system. The theorem proof assistant expresses the theorems and the proof process in high-order logic as high-level strategies, but human experts must manually construct proofs by inputting the strategies into the proof assistant. In this article, we propose the use of machine learning and concurrent search methods to improve the degree of automation of theorem proofs, which can help theorem assistants find suitable proof strategies faster and reduce the workload of constructing proofs. Our solution can generate effective tactics and can be used to prove theorems by automated methods more quickly.

E-government is a gradual system engineering, which needs to integrate new information technology. In e-government application scenarios, a large number of data (i.e. text, audio, video, graphics and animation) are managed by platforms maintained by data providers, which usually adopt centralized architecture design. However, this centralized architecture may lead to a single point of failure and data ownership disputes. It is difficult to ensure the integrity of the data and track the trusted traceability of data usage. In order to solve these problems, this paper proposes a cost-effective blockchain data storage framework for e-government data management. Generally speaking, in the field of e-government data management, the application trend of blockchain can be summarized into two aspects: one is to lay the foundation for the development of public services and government management through the establishment of identity authentication system based on blockchain; the other is to make use of the characteristics of blockchain, which can not be tampered with and has full historical records, so that the information resources of different institutions can be connected, which can achieve the goal Now the definition of data ownership and the traceability of data access.

Mobile Edge Computing (MEC) has the potential to enable computation-intensive applications in 5G networks. MEC can extend the computational capacity at the edge of a wireless network by offloading computation-intensive tasks to the MEC server. This paper considers a multi-mobile equipment (Mobile Equipment, ME) MEC system, where multiple mobiles -equipment can perform computational offloading via a wireless channel to a MEC server. To reduce the total cost during the offloading process, an algorithm based on reinforcement learning, Pre-Sort Q, is proposed. First, the transmission delay and calculation delay that computation jobs may experience, the transmission energy and computation energy that the computing system would consume were modeled. Then, the weighted sum of the delay and energy consumption and use preprocessing to determine the offloading decision to minimize system cost. Pre-Sort Q can reduce the weighted sum of delay and energy consumption through experimental simulation analysis and comparison compared with three benchmarks and one method.

The proposal of the blockchain 3.0 architecture represented by Hyperledger Fabric simplifies the process of each node participating in the blockchain network so that a large number of IoT devices can be quickly integrated into the network. However, the participating institutions in the field of power Internet of Things are complex, the performance of participating nodes is not uniform, and the load imbalance caused by it makes it impossible to meet the transaction throughput requirements of the business. In this paper, we established a Zookeeper-based endorser node performance monitoring module and replaced the original random selection strategy with a dynamic endorser node selection strategy based on the available resources of the node, effectively alleviating the endorsement pressure of a single node in high concurrency scenarios, thereby improving the overall concurrent processing capabilities. It is worth noting that our optimization is pluggable and does not make any changes to the interface of the Hyperledger Fabric, and it does not weaken the usability of the system. By comparing the existing architecture with our design, it is found that this dynamic decision-making endorser node selection strategy can help select more appropriate nodes for endorsement according to the node’s system resources and real-time load conditions, thus improving the concurrency of the system.

The internet of Things (IoT) uses wireless sensor devices to build a huge network transmission center, the data is increasing, and the data types have diverse characteristics. Such data raise a certain challenge for data storage and resource scheduling in cloud platforms. Swift can solve the needs of massive data storage capacity, but it is impossible to effectively achieve load balancing and waste resources. Therefore, according to the resource load status of the proxy node, this paper proposes a node resource load balancing strategy based on Linux Virtual Server (LVS). Through the study of scheduling algorithms, a dynamic feedback load balancing algorithm based on a weighted minimum connection is proposed. Experimental results show that the dynamic feedback load balancing algorithm based on a weighted minimum connection can effectively solve the resource scheduling problem.

Since ancient times, our country has been a collection of multi-ethnic coexistence. The diversified cultural differences, the regionality of educational background, and the very different living environment have caused different ethnic groups to have different emotions and attitudes towards the same things. Dealing with these problems will cause a lot of unnecessary trouble. In order to better understand and understand the emotional changes of various ethnic groups, promote ethnic integration, enhance national identity, and strengthen the sense of community of the Chinese nation. Therefore, the establishment of a library of ethnic minority emotions is very necessary for ethnic research. In this paper, we first developed an ethnic minority emotional picture tagging system, which uses the form of a network questionnaire to collect data on emotional pictures in the database. Then use SPSS18 to perform statistical analysis on the data, label each emotion picture, build a library of ethnic minority emotion pictures, and provide quantitative services for subsequent ethnic research.

As a basic work of network security, network traffic recognition plays an important role in network resource management and abnormal network traffic monitoring. At present, network traffic identification has become one of the hottest issues in academic research. In the past research, network traffic analysis was mainly done by Port Matching, Deep Packet Inspection. However, these methods are not perfect, and they are not suitable for today. This paper implements a traffic recognition method based on deep learning and machine learning. Besides, this paper implements unsupervised clustering of traffic. On the UNB ISCX data set, the experimental results are quite good.