Skip to main content

2022 | Buch

Advances in Smart Vehicular Technology, Transportation, Communication and Applications

Proceedings of VTCA 2021

herausgegeben von: Prof. Tsu-Yang Wu, Dr. Shaoquan Ni, Dr. Shu-Chuan Chu, Prof. Chi-Hua Chen, Prof. Margarita Favorskaya

Verlag: Springer Singapore

Buchreihe : Smart Innovation, Systems and Technologies

insite
SUCHEN

Über dieses Buch

This book includes selected papers from the fourth International Conference on Smart Vehicular Technology, Transportation, Communication and Applications (VTCA 2021), held in Chengdu City, Sichuan Province, China, during May 22–24, 2021. The conference is technically co-sponsored by Southwest Jiaotong University, Shandong University of Science and Technology, Fujian University of Technology, and Minjiang University. The book includes research works from engineers, researchers, and practitioners interested in the advances and applications in the field of vehicle technology and communication. The book covers four tracks, namely (1) vehicular networking security, (2) vehicular electronics, (3) intelligent transportation systems, and (4) smart vehicular communication networks and telematics.

Inhaltsverzeichnis

Frontmatter

Intelligent Transportation Systems

Frontmatter
Chapter 1. Deep Learning Short-Time Interval Passenger Flow Prediction Based on Isomap Algorithm

With the increasing complexity of subway lines, people’s demand for subway travel is also increasing. Reasonable regulation of vehicles on different zones can not only improve the efficiency of people’s travel but also lay the foundation for future short-time zone passenger flow prediction. The Isomap algorithm is used to represent the high-dimensional data by the low-dimensional method after transformation, and then the low-dimensional data are sorted from small to large, which results in the ordered OD data pairs. The ordered OD data pairs are then sorted in the database one by one for the last month, the corresponding data sets are constructed, and then the data are trained using the recurrent neural network model GRU to derive the passenger flow prediction results for the following week.

Junxi Chen, Kaihan Yu, Kangjie Wu, Jinshan Pan
Chapter 2. Prediction of Subway Interchange Passenger Flow Based on Recurrent Neural Network Time Characteristic Model

With the rapid development of social economy, urban rail transit has become the mainstream of people's travel mode. For the reasonable dispatch of vehicles, rail transit operation departments have an increasing demand for short-term passenger flow forecasting. Through the analysis of the passenger flow data of Chongqing Metro from October 1, 2018, to October 15, 2018, the GRU neural network is used to analyze the Chongqing Metro stations from October 16, 2018, to October 30, 2018, to predict the arrival passenger flow data, and use the logit model-based transfer method selection combination model to obtain the transfer sharing rate, and finally obtain the predicted passenger flow of each transfer method. The comparison with other models and optimization algorithms shows that this model has better prediction accuracy, stability, and robustness. GRU neural network has associative memory function, simple structure, small amount of calculation, and anti-noise ability, so it is worth popularizing in subway passenger flow prediction applications.

Xiao Chen, Junxi Chen, Yi Liu, Zichen Zhan, Jinglan Lei
Chapter 3. Research on Dynamic Passenger Flow Distribution of Rail Transit Based on Multi-dimensional Euclidean Distance

With the rapid development of urban rail transit, choosing subway for short-distance transportation has become the primary plan for many residents. The passenger flow distribution of urban rail transit has an important reference for improving ride comfort and urban rail transit operation management. This paper starts from the historical passenger flow data in and out of the station, uses the original data of the urban rail transit automatic fare collection system, and is based on the multi-dimensional Euclidean distance formula. Dynamic passenger flow distribution is performed on OD (Origin-Destination) data to obtain section passenger flow data. The result of dynamic passenger flow allocation can provide basic data for the passenger flow forecasting module on the one hand, and provide certain guidance for the planning and designing of rail transit on the other hand.

Qianqian Wu, Lin Sun, Furong Jia, Yi Liu, Zichen Zhan
Chapter 4. Study on Valid Travel Route Selection of Integrated Passenger Transportation

With the vigorous development of the integrated transportation network, the choice of passenger travel route is becoming more and more diversified. Therefore, it is important to build the integrated transportation network by establishing the cost function of valid travel route to measure the route that passengers may choose. Based on passenger travel demand, an optimization model of valid route selection for integrated transportation network was established based on generalized travel cost. According to the characteristics of the model, an improved BFS algorithm based on dynamic updating generalized cost function and determining valid travel route by key nodes was designed to solve the valid route. Finally, a simulation was given to verify the effectiveness of the model and algorithm.

Yao Yang, Chenyan Kan, Dingjun Chen
Chapter 5. Loading and Reinforcing Safety Evaluation of Railway Out-Of-Gauge Freight Considering the Dynamic Transportation Process

In order to evaluate the safety of loading and reinforcing scheme of railway out-of-gauge freight from the perspective of dynamic transportation process, the SIMPACK simulation model is established to collect the dynamic data during the transportation process, and safety factors are summarized into four indexes: wagon derail coefficient, wagon capsize coefficient, freight slip coefficient and freight capsize coefficient. On this basis, combine with the principle of neural network evaluation, this paper constructs a safety evaluation model of loading and reinforcing of railway out-of-gauge freights with the above-mentioned transportation process safety factors as the input layer and the comprehensive safety degree as the corresponding output layer. The results show that the lateral shift of gravity center for car loaded the gravity center height for car loaded and the longitudinal shift of gravity center for car loaded have the greatest impact on transportation process safety, and the influence degree is 0.52, 0.32, 0.35, respectively. And the classification of out-of-gauge freight and the length of protruding end have less impact. Other factors have relatively little impact.

Hao Chen, Wenxian Wang, Min Zhou, Xueqin Li
Chapter 6. Design of Rail Freight Products in the Arduous Mountainous Area on the Plateau Under the Concept of Modern Logistics

The railroad construction in the arduous mountainous area on the plateau is gradually improved. As the main channel of passenger and cargo transportation in and out of the difficult and dangerous mountainous areas on the plateau, how to develop cargo transportation on the basis of completing passenger transportation tasks and meet the transportation needs of resident’s living and production materials is an urgent problem to be solved. Based on the demand for railway freight transportation in the arduous dangerous mountainous areas on the plateau, this paper combines modern logistics concepts and accurately locates customer needs on the premise of fully adapting to the characteristics of freight transportation in the arduous dangerous mountainous areas on the plateau. Based on the product genealogy, the railway freight system in the arduous dangerous mountainous areas is designed with refined product design, to promote the improvement of railway freight capacity.

Shiwei Yang, Qingchen Yao, Can Yang, Miaomiao Lv, Mengyuan Yue

Smart Vehicular Technology

Frontmatter
Chapter 7. An Estimation of Vehicle Vertical Dynamics Using Inverse Method

The well-defined input to the vehicle or simulation model is one of the essential requirements to test driving ride comfort typically. This work suggests an approach to determining vehicle vertical dynamics, wheel vertical forces, and pitch torques based on an inverse direction methodology. The simple Kalman filter is used to form a filter that generates the residual innovation sequences with a recursive estimator. A least-squares algorithm is applied to compute the load’s magnitudes from the car systems measured dynamic response data. In numerical simulations, we analyze the current strategy’s feasibility and precision with a model driving estimation of wheel loads of a half-car over deterministic and random road profiles. The results from the simulation show that the proposed approach correctly measures the vertical dynamics of the vehicle.

Dong-Cherng Lin, Trong-The Nguyen, Jeng-Shyang Pan, Chang-Der Lee
Chapter 8. An Enhanced Flower Pollination Algorithm for Power System Economic Load Dispatch

This research proposes a solution to the economic load dispatch (ELD) problem using the enhanced flower pollination algorithm (EFPA). The EFPA has captured advanced features, e.g., simple structure, quick search, and implementation, due to introducing a random jump perturbation in the global pollination phase and updating the switching probability according to the optimal global value of each iteration. The mathematically expressed ELD is described as a typical problem of multi-constraint nonlinear optimization. Two case calculation experiments will be used to assess and evaluate the proposed system’s optimization efficiency from multi-dimensional economic perspectives with its feasibility and efficacy solution. The validation results show that the proposed scheme has more convergence speed and robustness than the other comparative methods.

Hung-Peng Lee, Trong-The Nguyen, Thi-Kien Dao, Van-Dinh Vu, Truong-Giang Ngo
Chapter 9. Distribution Vehicle Routing Optimization Based on 3D Loading

The rapid development of e-commerce leads to increasingly fierce competition in the logistics industry. Cost reduction and efficiency has become the key issue for logistics enterprises to survive. Loading optimization and path optimization are the two core issues in distribution activities, and they are inseparable and related. The joint optimization of the two is more in line with the actual distribution optimization needs. Based on the analysis of the interaction between routing and loading, this paper proposes a multi-objective optimization model with the minimum distribution cost, the minimum number of vehicles and the maximum customer satisfaction as the objectives. A hybrid nesting algorithm based on NSGAII algorithm is designed, in which the packing module is designed by tree search algorithm combined with the deepest and leftmost algorithm. Finally, the real data of China Railway Express Chengdu branch are taken as a case to verify the feasibility of the model and algorithm. Combined with the results of the algorithm, some suggestions are put forward.

Xuan Luo, Jin Zhang, Tingyu Yin, Hongxing Zhu, Mingyue Qiu
Chapter 10. Robust and Fast Registration for Lidar Odometry and Mapping

Outliers, such as sensor noise, abnormal measurements, or dynamic objects, can damage the overall accuracy of a Simultaneous Localization and Mapping (SLAM) system. Aiming at to improve the performance of Lidar SLAM systems in urban scenes containing a large number of outliers, we propose a real-time, feature-based, and outliers-rejection Lidar SLAM system. By embedding an outlier elimination method based on 4-points congruent sets into a state-of-the-art SLAM framework and further optimizing the traditional single-step registration to coarse-to-fine registration, we can solve the problem of time-consuming, high motion drift, and wrong mapping caused by the current Lidar SLAM systems which cannot effectively detect and eliminate the outliers in surrounding environment.

Wenbo Liu, Wei Sun
Chapter 11. Collaborative Control of Unmanned Vehicle Matrix Formation Based on Autonomous Neighborhood Negotiation

In large-scale activities or performances, the formation of unmanned vehicles on the large platform has become a shocking means of artistic expression. How to achieve low-cost and reliable cooperative control and highly reliable formation in a limited observation area with viewing Angle constraints is a very valuable area of research. This paper puts forward a kind of driving control method of unmanned vehicle formation with matrix arrangement based on a small-scale region driving consultation. An observation and neighborhood negotiation topology is designed, which can realize the automatic maintenance of unmanned vehicle formation in the leader-slave mode with a single leader. This method can provide a low-cost and high-reliability cooperative control scheme for large-scale unmanned vehicle formation performance.

KaiXuan Wang, YuTing Shen, FuQuan Zhang, Jianglong Yu, Zhang Ren, Liang Zhuo
Chapter 12. Unmanned Vehicle Task Scheduling Method Based on Iterative Cognitive Interaction

At present, in the field of autonomous driving, the application of unmanned vehicles has made considerable progress. However, there are still a large number of engineering problems to be solved in practical application. For example, the ability of multi-vehicle interactive autonomous operation in high-dynamic and time-sensitive environment is not sufficient for large-scale application. These kinds of problems restrict the effective integration of unmanned intelligent transportation system and current traffic system. Facing the problems described above, this paper focuses on introducing knowledge-driven and experiential memory decision-making process for agents, which is based on the research on the process characteristics of human brain cognitive reasoning under the condition of dynamic changes in the observation results of multi-agent clusters. Thus, the method of unmanned vehicles tasks scheduling method based on iterative cognitive interaction is proposed in this paper. In view of the behavior and decision-making process of multi-vehicle in aspects of cognition, interaction, and association, the analysis and research on the time scale will provide a research route for solving the continuous autonomous operation of multi-vehicle system in a high dynamic environment. Contribution of this paper can provide theoretical basis and practical value for multi-application fields.

Yuting Shen, Xin Meng, Kaixuan Wang, Fuquan Zhang, Yueqing Gao, Lulu Chen
13. Event Sequence T-Way Test Strategy for Events Driven System

Liu, Yuqi Pei, Daming Fang, ShiyuanEvent-driven softwares (EDS) are now widely developed and used. Common example of Event-driven software span multiple domains from embedded systems to web and GUI applications. Testing methods based on event executing permutation is common in software testing field. Combinatorial method has been applied to generate sequence coverage array (SCA) such as t-seq algorithm developed by Kuhn et al. The SCA generated was aimed at n distinct events which occurs exactly once in sequence. However, event may be repeatable in test sequence, because, for most reactive or event driven systems, events occur multiple times in the course of practical use. Existing sequence coverage array generated by combinatorial method may exist redundancy for repeatable events. Therefore, we propose a reduction algorithm for removing the unwanted subsequence of SCA. Furthermore, we introduced a direct construction method to generate one-row test sequence, which satisfy t-way permutations coverage of n events.

Yuqi Liu, Daming Pei, Shiyuan Fang

Artificial Intelligence

Frontmatter
14. QUasi-Affine TRansformation Evolutionary Algorithm for Feature Selection

Du, Zhi-Gang Pan, Tien-Szu Pan, Jeng-Shyang Chu, Shu-ChuanQUasi-Affine TRansformation Evolutionary Algorithm (QUATRE) is a currently emerging meta-heuristic evolutionary algorithm. QUATRE has the ability to balance exploitation and exploration in the optimization process, and the algorithm optimization uses matrix operations to greatly reduce the time complexity for solving the same problem. This series of advantages makes this algorithm adopted by a large number of researchers. In this paper, QUATRE is used to optimize the Feature Selection (FS) of the wrapper method. K-Fold Cross-Validation (KFCV) method is also used to divide the test set and training set of the sample, and then use the K Nearest Neighbor (KNN) algorithm for feature classification. In the optimization process, we use a threshold (choice) for feature identification to select useful features. Finally, the 9 standard test data sets in UCI are used to verify the effectiveness of the QUATRE algorithm.

Zhi-Gang Du, Tien-Szu Pan, Jeng-Shyang Pan, Shu-Chuan Chu
15. Advanced QUasi-Affine TRansformation Evolutionary (QUATRE) Algorithm and Its Application for Neural Network

Hu, Pei Pan, Jeng-Shyang Chu, Shu-ChuanThe QUasi-Affine TRansformation Evolution (QUATRE) was first proposed by Meng et al. in 2016. It has the characteristics of few parameters and fast convergence. This paper brings two methods to improve its solution quality. The opposite position and comprehensive learning greatly advance the ability of jumping out of local traps when the QUATRE falls into stagnation. Their performance is verified by 23 benchmark functions. In the end, they succeed to train the parameters of neural network and predict the long-term traffic flow in Qingdao.

Pei Hu, Jeng-Shyang Pan, Shu-Chuan Chu
16. An QUasi-Affine TRansformation Evolution (QUATRE) Algorithm for Job-Shop Scheduling Problem by Mixing Different Strategies

Yang, Qing-Yong Chu, Shu-Chuan Chen, Chien-Ming Pan, Jeng-ShyangHow to solve the Job-Shop Scheduling problem (JSP) effectively and make the most efficient use of resources has always been the focus of academic and engineering circles. Aiming at the traditional JSP problem, this paper proposes a new QUasi-Affine Transformation Evolution algorithm (QUATRE) to solve it, called QUATRE-SAO for short. The QUATRE-SAO algorithm combines Simulated Annealing (SA) strategy and Opposition-based Learning (OBL) strategy to enhance the algorithm to jump out of local optimum and further improve the optimization performance of the algorithm. Through the comparative experiment of FT and LA series standard test examples, the results show that the QUATRE-SAO algorithm can solve the JSP problem better and can get a better solution.

Qing-Yong Yang, Shu-Chuan Chu, Chien-Ming Chen, Jeng-Shyang Pan
17. QUasi-Affine TRansformation Evolution Algorithm for Optimal Power Flow of Integrated Electrical Network Combining Thermal Power with Wind Power

Li, Jianpo Gao, Min Chu, Shu-Chuan Li, Geng-Chen Pan, Jeng-ShyangWith the development of electric power industry, it is of great technical and economic significance to introduce the optimal power flow (OPF) calculation into the economic analysis of electric power market, which can not be realized by the traditional power flow calculation. At present, many researchers have applied evolutionary algorithms to the calculation of optimal power flow. QUasi-Affine TRansformation Evolutionary (QUATRE) algorithm is a new evolutionary algorithm, combined with the OPF model of power market, which has high efficiency and significance for the economic analysis of power market. In this paper, the minimization of generation cost and active power loss is taken as the objective function, and QUATRE is selected as the optimization tool to study the OPF. In this paper, the simulation experiment of this problem is carried out, and compared with Particle Swarm Optimization (PSO), Artificial Bee Colony (ABC) and Moth Swarm Algorithm (MSA). Through the analysis of the experimental results, it can be seen that QUATRE is superior to other algorithms in terms of power generation cost, active power loss and CPU time.

Jianpo Li, Min Gao, Shu-Chuan Chu, Geng-Chen Li, Jeng-Shyang Pan
Chapter 18. Hybrid Optimization Algorithm Based on QUATRE and ABC Algorithms

Artificial bee colony optimization algorithm (ABC) is an optimization algorithm based on swarm intelligence which is obtained by observing the behavior of bees looking for nectar and sharing food information with bees in the hive. QUasi-Affine TRansformation Evolutionary (QUATRE) is an algorithm that uses quasi-affine transformation as an evolution method because ABC has the shortcoming of weak ability to develop new nectar sources, and QUATRE has weak search ability but strong development ability, so this paper combines these two algorithms to a certain extent and proposes an improved artificial bee colony optimization algorithm (QUA-ABC). QUA-ABC is inspired by the location update formula in QUATRE and proposes a new location update formula suitable for ABC. In this study, experiments were conducted using the internationally used CEC2013 data set. The optimization accuracy and convergence speed of QUA-ABC were compared with the original ABC. The results show that the QUA-ABC algorithm has stronger capabilities and better performance.

Xin Zhang, Linlin Tang, Shu-Chuan Chu, Shaowei Weng, Jeng-Shyang Pan
Chapter 19. Joint Adaptive Reception Algorithm with Ant Colony Optimization for Asynchronous Cooperation Transmission Systems

A joint adaptive reception algorithm is proposed for asynchronous cooperation systems, which is aiming at reducing the interference caused by the asynchronous relays, where ant colony optimization algorithm is adopted to enhance the adaptive training process. The coordination node is usually needed for cooperative communication systems, which is employed to apply relays timing synchronization as well as relay selections, thus achieving the diversity order. However, it could be difficult to choose the center, when the requirement is too rigor for some certain system, such as the complexity-limited ones. For this kind of non-centered systems, the relaying nodes usually work asynchronously, so it is necessary to remove the inter-symbol interference (ISI) in the received signal, in order not to degrade the whole transmission performance. As an effective means, equalizer is usually utilized to carry out ISI cancelations, where zero forcing and minimum mean square error are known as the criterions. When channel state information is unavailable, adaptive equalizer can be considered, with which to train the weightings of equalizer. In this paper, to improve the training process of conventional adaptive algorithms, we investigate and modify ant colony optimization to propose a hybrid and combined adaptive architecture, and the converging property can be guaranteed. Computer simulations show that under Rayleigh fading cooperation communication channels, the proposed algorithm has faster convergence speed and can achieve better detecting performance than conventional adaptive equalizer.

Aiyong Zhang, Changjie Liu, Pengfei Qin
20. Multiple Data-Dependent Kernel Learning for Circuit Fault Diagnosis

Jianfeng, Wang Meixi, Wu Hanzhi, LiAn analog circuit fault diagnosis method based on multi- data correlation kernel is proposed, and the UCI data set is used to verify the effectiveness of the proposed method. Then, a fault diagnosis method structure of tolerance circuit based on SVM is proposed. Taking Sallen key filter circuit as an example, the specific steps of establishing an analog circuit fault diagnosis model, including fault injection, are introduced: circuit simulation, fault feature extraction, and design of SVM fault classifier based on multi-data correlation kernel. Then, the Sallen key filter circuit and leap frog filter circuit are selected as the diagnosis objects. The HSPICE software is used to inject the fault into the circuit under test and establish the fault simulation model, so as to obtain the circuit data under different circuit states, and the circuit samples are used to establish the fault classifier based on SVM. Finally, the effects of SVM + MK, SVM + DK, and SVM + MDK on the fault classifier diagnosis are compared. The experimental results show that the three methods used in this paper are better than the analog circuit fault diagnosis method based on standard SVM, and the proposed analog circuit fault diagnosis method based on multi-data correlation kernel is the best in terms of diagnosis effect. On this basis, the SVM + MDK algorithm is more effective The establishment time and diagnosis efficiency of the model are relatively good.

Wang Jianfeng, Wu Meixi, Li Hanzhi
Chapter 21. Calculation of Spacecraft Transfer Trajectory Based on Modified Differential Evolution Algorithm

Aiming at the optimization problem of spacecraft transfer trajectory, based on differential evolution algorithm, a fast optimization method for transfer trajectory with modified differential algorithm is proposed. The method can quickly find the best transfer trajectory that satisfies the specified constraints through the intelligent optimization based on the population individuals. It avoids the problems of large calculation and low efficiency caused by traditional iteration methods. The actual application of the calculation of the transfer trajectory under certain initial conditions shows that the method can quickly determine the transfer orbit parameters.

Gui-bo Zheng, Qing Yin, Guan-qun Wu, Ke-yan Huang
22. Channel Pruning and Quantization-Based Learning for Object Detection with Computing Source Limited Application

Zhao, Fei Liu, Huanyu Hu, Moufa Deng, YingjieWith the rise of convolutional neural network (CNN) in the field of computer vision, more and more practical applications need to deploy CNN on mobile devices. However, due to the large amount of CNN computing operations and the large number of parameters, it is difficult to deploy on ordinary edge devices. The neural network model compression method has become a popular technology to reduce the computational cost and has attracted more and more attention. We specifically design a small target detection network for hardware platforms with limited computing resources, use pruning and quantization methods to compress, and demonstrate in VOC dataset and RSOD dataset on the actual hardware platform. Experiments show that the proposed method can maintain a fairly accurate rate while greatly speeding up the inference speed. The proposed model designed in this paper achieves 76.74% mAP on the VOC dataset, which is 4.76 times faster than the original model.

Fei Zhao, Huanyu Liu, Moufa Hu, Yingjie Deng
Chapter 23. Retinal Vessels Segmentation Based on Multi-scale Hybrid Convolutional Network

Retinal fundus image can reveal the information on the early symptoms of diabetes, hypertension, hyperlipidemia and other diseases. Accurate segmentation of retinal vessels can assist the detection and diagnosis of the related diseases. Due to the intricate characteristic information of retinal vessels images, traditional segmentation methods lead to inaccurate segmentation for small vessels and pathological segmentation errors. In this paper, we propose a new multi-scale hybrid convolution U-Net. Firstly, we take the hybrid convolution module by combining dilated convolution and standard convolution as the core structure for feature extraction to obtain more abundant semantic feature information, while expanding the receptive field. Then, we add a multi-scale fusion module to the network encoding and decoding connection part, which fuses the feature information of different layers to reduce the loss of information and enhance the representation ability of the network. We evaluate the performance of the proposed method on two public retinal datasets (DRIVE and CHASE_DB1). The results of quantitative and qualitative experiments show that the proposed model can improve good accuracy in retinal vessels segmentation.

Rui Li, Zuoyong Li, Xinrong Cao, Shenghua Teng
Chapter 24. Location Optimization of Service Centers for Seniors Based on an Improved Particle Swarm Optimization Algorithm

The world's population is gradually aging, and the construction of Service Centers for Seniors (SCS) has become an important issue worthy of concern. In this paper, a particle swarm optimization algorithm with random weight and synchronous learning factor (RSPSO) is proposed to optimize the location and compared with three improved PSO algorithms. Experimental results show that RSPSO bears a faster convergence with better improvements on global searching. Furthermore, it can effectively avoid falling into the local optimal solution. The results also demonstrate the superiority of RSPSO over PSO in location optimization of SCS.

Wei-Feng Wang, Ruo-Bin Wang, Shuo Yin, Zhi-Wei An, Lin Xu
25. IBPO: Solving 3D Strategy Game with the Intrinsic Reward

Li, Huale Cao, Rui Hou, Xiaohan Wang, Xuan Tang, Linlin Zhang, Jiajia Qi, ShuhanIn recent years, deep reinforcement learning achieves great success in many fields, especially in the field of games, such as AlphaGo, AlphaZero and AlphaStar. However, reward sparsity is still a problem in the 3D strategy games with a higher dimension of state space and more complex game scenarios. To solve this problem, in this paper, we propose an intrinsic-based policy optimization algorithm (IBPO) for reward sparsity. The IBPO incorporates the intrinsic reward into the traditional policy, which composed by the differential fusion mechanism and the modified value network. The experimental results show our method can obtain better performance than the previous methods on the VizDoom.

Huale Li, Rui Cao, Xiaohan Hou, Xuan Wang, Linlin Tang, Jiajia Zhang, Shuhan Qi
26. An Operation with Crossover and Mutation of MPSO Algorithm

Zhong, Yuxin Chen, Yuxin Yang, Chen Meng, ZhenyuAs an efficient and simple optimization algorithm, particle swarm optimization (PSO) has been widely applied to solve various real optimization problems in expert systems. However, avoiding premature convergence and balancing the global exploration and local exploitation capabilities of the PSO remains an open issue. To overcome these drawbacks and strengthen the ability of PSO in solving complex optimization problems, a modified PSO using adaptive strategy called MPSO is proposed, although MPSO has achieved excellent performance, and its convergence and stability are still some defects. In this paper, we presented a new variant of MPSO algorithm which can explore the search space deeper than the previous method, and better performance can be achieved under CEC2013 test suite.

Yuxin Zhong, Yuxin Chen, Chen Yang, Zhenyu Meng
Chapter 27. Deep Learning on 3D Point Cloud for Semantic Segmentation

In this paper, a method by applying deep learning method onto the point clouds data for semantic segmentation is proposed. Three convolutional neural networks, PointNet, PointNet++, and DGCNN, are replicated, designed, and analyzed. In order to avoid problems introduced by some other methods due to the preprocessing step, here, PointNet, PointNet++, and DGCNN are directly used onto the 3D point cloud. Experiments verified the effect of these neural networks on point clouds for semantic segmentation. Methods based on PointNet and PointNet++ show good results, while DGCNN-based reached state-of-the-art performance.

Zhihan Ning, Linlin Tang, Shuhan Qi, Yang Liu
Chapter 28. An Improved Arithmetic Optimization Algorithm with a Strategy Balancing Exploration and Exploitation

With the increasing complexity and difficulty of practical problems, higher requirements are put forward to optimization techniques, especially the improvement on reliability and performance of meta-heuristic algorithm. In this paper, an improved arithmetic optimization algorithm (IAOA) is proposed, and it is compared with two algorithms—particle swarm optimization (PSO) and arithmetic optimization algorithm (AOA) on 13 benchmark functions. Experimental results show that the proposed algorithm performed better than the compared algorithms in solving particle problems in most cases.

Ruo-Bin Wang, Shuo Yin, Wei-Feng Wang, Zhi-Wei An, Lin Xu

Networks and Security

Frontmatter
29. MPSiam: A Fast Multiplexing Siamese Tracking Network

Li, Donghao Shen, Ce Hu, Jinxing Yuan, DipingSiamese trackers have achieved remarkable performance in accuracy. However, the high memory cost and inference speed have restricted the deployment of the state-of-the-art trackers in mobile applications. To address this issue, this paper presents a backbone consisting of multiplexing convolution blocks that newly proposed by us, which combine the spatial multiplexing operation and channel multiplexing operation. The spatial multiplexing operation is inspired by the subpixel convolution in super-resolution tasks. The channel multiplexing operation is inspired by the channel shuffle in ShuffleNet. These two modules can be used to effectively optimize the multiply–accumulate (MACC) operation, by multiplying the number of operations and then adding it to a network. We employ this new module to build a novel lightweight backbone for the SiamRPN++ tracker. We trained this model and evaluated its performances on the VOT2018 and OTB2015 datasets. Our model is compressed to 43 MB, the inference time was 83 FPS, and the experiments were carried out in a single NVIDIA 2080Ti GPU. Our model is superior to MobileNetv2-SiamRPN++, which has a model size of 58 MB and the inference time of 55 FPS, and our method also managed to reduce the MACC from 1.2 to 0.5 B. Compared with SiamRPN++ with Resnet50 backbone, our model achieved a compression rate of 4.8 $$\times $$ × and speedup of 3.3 $$\times $$ × , just losing 3% EAO.

Donghao Li, Ce Shen, Jinxing Hu, Diping Yuan
Chapter 30. A Single-Phase-to-Ground Fault Location Method Based on Deep Belief Network

When a single-phase-to-ground (SPG) fault occurs in a resonant grounding distribution system, the amplitude of the transient zero-sequence current waveform at the upstream detection node of the fault point is greater than the amplitude of the transient zero-sequence current waveform at the downstream detection node, and the two polarities are opposite. The transient zero-sequence currents at the detection nodes on the same side of the fault point are very similar. Based on this, the paper proposes a new method of SPG fault location based on a deep belief network (DBN). Firstly, this method uses the fault transient zero-sequence current waveform obtained from each detection node in the simulation model as the input of DBN, and the deep features of the fault signals are extracted. Secondly, the deep features are divided into upstream detection nodes category and downstream detection nodes category by a supervised classifier. And then, the fault location is implemented by analyzing the network structure of fault detection nodes. Finally, the testing results of the simulation data prove that the algorithm has high recognition accuracy under different fault grounding points, different initial phase angles of faults, different grounding resistances, and different types of faults, and has certain practical engineering application value.

Jia-Min Li, Shi-Jian Liu, Xiang Shao, Jeng-Shyang Pan
Chapter 31. A Data Fusion Scheme in Wireless Sensor Network Based on Optimizing Parameters of Neural Network

This study suggests a scheme of data fusion strategy in wireless sensor networks (WSNs) based on optimizing the neural network (NN) to decrease data redundancy, increase data transmission, and save communication energy consumption in WSN. The optimal parameters are optimized by applying the bat algorithm (BA). The optimized neural network (NNBA) is used to fuse captured data in cluster head (CH) and then forwards the combined data to the base station (BS) of a WSN. The simulation experiment is implemented in several scenarios to test the proposed scheme performance. The proposed scheme's results show that the proposed algorithm can save sensor node energy consumption, extend the lifetime, and increase the data fusion accuracy of WSN.

Thi-Kien Dao, Trong-The Nguyen, Van-Dinh Vu, Truong-Giang Ngo
Chapter 32. Cluster-Based Two-Level Mesh Routing Protocol for Wireless Sensor Network

Wireless sensor networks (WSNs) have broad application scenarios in military and civilian fields. How to reduce the energy consumption of network communication and extend the life cycle of the network has always been a research focus of WSN. Clustering is one of the important methods to extend the network life cycle. However, the existing clustering routing protocol has room for further optimization of the connection mode between the common node and the cluster head node, the network topology structure and the communication mode between nodes. In this paper, a cluster-based two-level mesh (CTM) routing protocol for WSN is proposed, which supports common nodes to connect to the cluster head and communicate with any node in a multi-hop manner. It also specifies the topological structure of the two-level mesh and two communication modes. Finally, simulations on the OMNeT++ platform verify that the protocol can effectively reduce network and energy costs.

Qi-yuan Zhang, Bo Sun, Jian-ming Xu, Jian He, Ji-liang Mu
Chapter 33. PD Detection and Analysis of Cross-Bonded Cable Based on Broadband Sensor and Three-Phase Amplitude Relation Analysis Method

The power cable is one of the most important transmission equipment. The long cable common use the cross-bonded grounding method to reduce the loss of shielding layer and decrease the voltage rise at the end of cable. Therefore, it is difficult to diagnosis the PD detection result of cross-bonded cable in field because the PD signals will be influenced by both the environment electromagnet interference and the interference among different phases. The broad band sensor is used to detect the PD of cross-bonded cable in field in this paper. The waveform of the PD pulse, PRPD, and 3PARD diagrams are analyzed to diagnose the results. The results have shown that the broadband sensor can anti the environment electromagnet interference because it can obtain more information of PD signal. The 3PARD diagrams can division the interference among different phases.

Liwei Wang, Bingwei Liu, Bin Wang, Xutao Han, Zhentao Liu, Wei Sun
Chapter 34. A Dynamic and Fair Timeout Heartbeat Detection Technique for Server Clusters Using Nginx Reverse Proxy

In recent years, more and more SMEs use their existing equipment enterprises to form server clusters for the internal welfare platform of enterprises through Nginx reverse proxy. For this kind of platform, the health detection mechanism of nodes in the server cluster plays a very important role in ensuring the high availability of the system. Nginx’s native health detection mechanism is very weak. The widely used Tengine provides Nginx with a separate health detection module, but it still fails to consider the problem of different node performance caused by heterogeneous hardware and uneven load on software, and uses a unified heartbeat detection strategy, which leads to the problem that Nginx cannot detect the faulty node in time or mistakenly determines the “Fake dead” node as the faulted node. In this paper, node performance parameters such as CPU utilization and IO utilization are selected to construct a weight calculation model by entropy method to quantify node performance, and then a node fault misjudgment loss model is constructed to optimize the accuracy of node fault judgment. Finally, a dynamic fair timeout algorithm based on Nginx is proposed to make Nginx’s heartbeat timeout strategy more fair and real-time. Experiments show that compared with the heartbeat detection technology before improvement, the algorithm proposed in this paper improves the accuracy of node fault detection by 18%, and this algorithm is an effective technology to improve the high availability of cluster.

Beiping Ma, Wei Zhang
Chapter 35. Visually Meaningful Image Encryption Algorithm Based on Parallel Compressive Sensing and Cellular Neural Network

At present, most image encryption algorithms protect the image by converting the original image into a visually meaningless noise-like image. However, the noise-like image can easily attract the attention of attackers, which increases the risk of being deciphered. Thus, a novel meaningful image encryption algorithm based on parallel compressive sensing and cellular neural network is proposed in this paper. In our scheme, the image is processed by three modules, namely, compression, encryption and hiding. Among them, the compression module is utilized to compress the plain images in parallel. After that, the encryption module converts the compressed image into a meaningless noise-like image by diffusion and scrambling. Finally, the hiding module realizes the visualization of the encrypted image by embedding the encrypted image into a carrier image, which helps reduce the interest of attacker in deciphering cipher images. Experimental results and analysis indicate that the proposed algorithm has satisfactory performance and can effectively protect image information from being leaked under the premise of reducing computational overhead.

Renxiu Zhang, Donghua Jiang, Wei Ding, Ya Wang, Yanan Wu, Yerui Guang, Qun Ding
36. Comments on “A Robust User Authentication Protocol with Privacy-Preserving for Roaming Service in Mobility Environments”

Guo, Xinglan Yang, Lei Wu, Tsu-Yang Chen, Lili Chen, Chien-MingRoaming service under the global mobile network (GLOMONET) means that users who use mobile devices can still use mobile devices in other regions or countries after leaving their region or country. When mobile users use roaming services, the communication information transmitted by wireless channels is easy to be tampered with and eavesdropped on by attackers. These attacks may expose the identity and location of remote users. Thus, mutual authentication among mobile users, foreign agents, and home agents play an important role. To ensure a secure roaming service in a mobile network, it is necessary to design an efficient and secure solution. Recently, Shashidhara et al. proposed a user authentication protocol for roaming service in the GLOMONET. In this paper, we find that there are some security vulnerabilities in their protocol, including perfect forward secrecy (PFS), key compromise impersonation attacks (KCIA), and known-session-specific temporary information attacks (KTIA).

Xinglan Guo, Lei Yang, Tsu-Yang Wu, Lili Chen, Chien-Ming Chen

Video, Image, and Others

Frontmatter
Chapter 37. Research on Construction Method of Massive Geographic Image Database for Power Grid EIA

With the continuous development of science and technology, the capacity of data is also showing a geometric increase. Therefore, this paper mainly studies how to store query, change, and replace a large number of geographic images. The text uses SQL server platform and ArcGIS geographic software processing platform. It solves the problem of processing limit exceeding the database in the face of massive database, insufficient network bandwidth compared with massive images, slow image display speed, storage problem of massive image data, and so on.

Yu Wu, Zun Li, Yang Guo, Songyang Zhang, Zhiguo Zhang, Zhentao Liu, Xutao Han, Wei Sun
Chapter 38. Automatic Director of Live Sport Based on Motion State

This paper introduces automatic director method of live sport based on motion state. The method is to combine the video frames of multi-channel cameras and use larger and more complete panoramic video for video content analysis and semantic event detection. Then, through the state analysis of the sport, it is determined whether the match at that moment is in Play or Break state. If the match is in Play state, it is adopted to place the motion target in the center of the guide screen. If the game is in the Break state, the sliding window method is adopted to build a HCRF model to mine the potential feature relationship between multimodal semantic clues and semantic events, so as to realize the detection of specific semantic events in sports videos, and to smoothly track the guide screen for extraction. We have done experiments on football match videos, and the results show that our method can generate coherent edited videos in line with audience psychology and achieve good visual effects, which also lays a foundation for the application of artificial intelligence to the automatic director of sports events.

Juan Wang, Longfei Zhang
Chapter 39. Image Dehazing Network Based on Multi-scale Feature Extraction

To remove image haze and make haze image scene clear, we proposed an image dehazing network based on multi-scale feature extraction (MSFNet) in this paper. The MSFNet first directly performs feature extraction on hazy images with three different resolutions to obtain fine feature maps and concatenates them with the rough feature maps extracted in the downsampling process for fusing and obtaining richer image information. Then, the fused feature maps are put into a network module composed of ResNeXt building blocks for network learning. Next, the feature maps extracted by upsampling are sequentially concatenated with the feature maps learned by the ResNeXt module for obtaining the residual image. Finally, the learned residual image is added to the input hazy image to obtain the image dehazing result. The experimental results on the SOTS dataset show that the MSFNet improves effectiveness of image dehazing.

Ting Feng, Fuquan Zhang, Zhaochai Yu, Zuoyong Li
Chapter 40. Research on the Challenges and Strategies of Enterprises in Reverse Logistics Cost Control Under B2C Mode

Generally speaking, B2C e-commerce enterprises have a high-sales return rate, resulting from the nature of the industry, and sales return means the increase in reverse logistics cost. Through the research on the composition of reverse logistics information system and reverse logistics process, it is found that B2C e-commerce enterprises face some problems in reverse logistics cost control, such as imperfect reverse logistics information system, unreasonable value utilization of returned and exchanged goods, huge storage cost and incomplete third-party logistics utilization. In order to achieve competitive advantage in the fierce market, e-commerce enterprises should establish a perfect reverse logistics information system, optimize the reuse value of returned and exchanged goods, endeavor to reduce the storage cost and fully introduce the third-party logistics. The countermeasures can provide some reference for e-commerce enterprises to do well in reverse logistics cost management under B2C mode. In this way, the enterprises can reduce reverse logistics cost and obtain healthy and sustainable development.

Chengxiao Ju
Chapter 41. Research on the Training Path of Big Data Application-Oriented Talents in Chinese Colleges and Universities

The big data technology is undergoing fast development around the world, and the market requirement for related talents is growing, while the insufficient supply of application-oriented talents has become an urgent issue that needs to be addressed. By analyzing the current situation of talent need in China's big data industry and the training of big data application-oriented talents in Chinese colleges and universities, this paper innovatively proposes the “three-integration training path” of big data application-oriented talents for colleges and universities. The “three-integration training path” mainly consists of the following three aspects: First, building the soft environment for talent training through “Integration of interdisciplinary”; Second, meeting the diversified needs of talents through “Integration between teaching and employment”; Third, developing multi-field practice scenarios through “integration between education and industry”. The “three-integration training path” can urge colleges and universities to conduct internal governance and reform by combining the industrial need, and provide valuable references for colleges and universities to train of big data application-oriented talents.

Xiaohong Ju
Backmatter
Metadaten
Titel
Advances in Smart Vehicular Technology, Transportation, Communication and Applications
herausgegeben von
Prof. Tsu-Yang Wu
Dr. Shaoquan Ni
Dr. Shu-Chuan Chu
Prof. Chi-Hua Chen
Prof. Margarita Favorskaya
Copyright-Jahr
2022
Verlag
Springer Singapore
Electronic ISBN
978-981-16-4039-1
Print ISBN
978-981-16-4038-4
DOI
https://doi.org/10.1007/978-981-16-4039-1

    Premium Partner