Table of Content

    24 September 2020, Volume 0 Issue 09
    A Clustering Data Collection Algorithm in WSN Based on Compressed Sensing
    ZHANG Lei, CUI Juan-juan, LI Jing-jing
    2020, 0(09):  1-5.  doi:10.3969/j.issn.1006-2475.2020.09.001
    Asbtract ( 173 )  
    References | Related Articles | Metrics
    In order to reduce the wireless sensor network transmissions and its energy consumption, a data collection method combining K-means balanced clustering and Compressed Sensing (CS) theory is proposed based on the characteristics of spatio-temporal correlation of WSN node data. Firstly, K-means clustering algorithm is used to divide the network into clusters. Then, each cluster head node transfers the collected data to the Sink node of the base station based on the spatial-temporal CS. Finally, Sink node uses OMP algorithm to accurately reconstruct the collection data. The simulation results show that this algorithm effectively reduces the data traffic of wireless sensor network and measurement required in the reconstruction of compressed sensing algorithm.
    Time Series Forecasting Model Based on LSTM-Prophet Nonlinear Combination
    ZHAO Ying, ZHAI Yuan-wei, CHEN Jun-jun, TENG Jian
    2020, 0(09):  6-11.  doi:10.3969/j.issn.1006-2475.2020.09.002
    Asbtract ( 725 )  
    References | Related Articles | Metrics
    At present, the single prediction model has low prediction accuracy for complex nonlinear time series and can not capture the composite characteristics of time series well. Therefore, this paper proposes a time series prediction model based on back propagation neural network combination of Long Short-Term Memory-Prophet (LSTM-Prophet). The prediction values obtained from the long short-term memory network and Prophet prediction model are combined by BP neural network to obtain the final prediction value. Then, a comparative experiment is designed and implemented between the proposed model and three individual models. The accuracy and validity of the proposed model are verified by data sets from three different fields. The experimental results show that the proposed prediction model has high prediction accuracy, good universality, and application prospect.
    Analysis of Hot Topics of College Online Public Opinion Based on Baidu Tieba
    LI Jin-hai, HU Xu
    2020, 0(09):  12-18.  doi:10.3969/j.issn.1006-2475.2020.09.003
    Asbtract ( 301 )  
    References | Related Articles | Metrics
    The college online public opinion is the collection of attitudes, cognitions, opinions, and emotions on public events directly related to college students or they are interested in in the context of campus. With the popularity of network applications, the college online public opinion may develop in any uncontrollable direction. Therefore, the study of college online public opinion is an important work in college management. This paper uses the Taizhou college Bar as the data source, and uses Web crawlers to collect the content of the topic posts and the number of replies of the Taizhou college Bar as experimental data. Through Python data analysis technology, the hot topics of college online public opinion is analyzed, and the data analysis results are demonstrated through the data visualization technology. The causes and effects of hot topics of college online public opinion is studied based on the results of data analysis. Finally, the paper proposes guidance strategies to promote the overall development of college students under the Internet environment and promote the campus harmonious construction.
    A Multi-feature Fusion Algorithm for Label Generation of Educational Resources
    LI Wen, WEN Yong-jun, TANG Li-jun,
    2020, 0(09):  19-24.  doi:10.3969/j.issn.1006-2475.2020.09.004
    Asbtract ( 136 )  
    References | Related Articles | Metrics
    In the form of tags, educational resources can be accurately described in a simple and effective way, and the messy and huge educational resources in the Internet can be classified efficiently, so that users can browse and obtain educational resource information conveniently and the utilization rate of educational resources  is improved. There are many methods to generate text tags in natural language processing, but the description of features is not comprehensive. Therefore, the method of label generation for multi-feature fusion is studied. Combining with the characteristics of Chinese text, adding TF-IDF weights and location information weights on the basic of TextRank algorithm, considering the information of words in the corpus and the position information in the article, the labels including corpus information and position information are generated to form a multi-feature fusion algorithm for label generation. The test results and analysis show that the maximum F-measure value of the improved TextRank algorithm is 0.571 and its average value is 0.34, which is better than the commonly TextRank algorithm and TF-IDF algorithm, and the improved TextRank algorithm can effectively improve the quality of educational resource labels, which is beneficial to better utilization and management of educational resources.
    A Multi-source and Multi-dimensional Government Data Sharing Analysis System for Macroeconomic Analysis
    LI Xin-hua, WANG Yong, YAN Jia-jing, HOU Wen-gang
    2020, 0(09):  25-31.  doi:10.3969/j.issn.1006-2475.2020.09.005
    Asbtract ( 139 )  
    References | Related Articles | Metrics
    In the context of the “digital economy” national strategy, to provide timely, comprehensive and accurate macroeconomic information for different types of users like government departments, industries and enterprises and form a big-data-analysis-based decision-making system, government shared data of many different business fields, data structures and data sources is gathered by Jiangxi provincial information center, and a macro-economic big data analysis system is built. Combined with the traditional time series characteristics and other macro-economic characteristics which affect key indicators of social economy, a method and model database about applying economic big data in monitoring and early warning macro-economy is built on the basis of time series analysis, regression model. Furthermore, double logarithm linear regression model and ARIMA model are adopted to carray out real-time monitoring and intelligent prediction research on social economy to provide a reliable reference for macro-control and precise implementation of relevant government departments.
    A Hybrid Prediction Method on Graph Convolutional Network with Single Time-series Feature
    LI Hao-tian, SHENG Yi-qiang,
    2020, 0(09):  32-36.  doi:10.3969/j.issn.1006-2475.2020.09.006
    Asbtract ( 178 )  
    References | Related Articles | Metrics
    In recent years, graph neural networks have been widely discussed in the field of deep learning. However, most of the researches are based on graph nodes and carry out classification and regression prediction under the premise of multi-dimensional attributes. Forecasting does not produce the desired results on single time-series of feature. This paper proposes a time-series graph convolutional network that can predict features in a complex graph network based on single time-series of feature of the node. By parameterizing the adjacency matrix in the traditional graph convolution network, the algorithm solves the problem of parameter degradation under a single feature condition, and combines the sequential learning method of the LSTM network to integrate the timing information into the training process, which improves the training accuracy. Experiments on the traffic flow data set PeMS and Los show that the prediction accuracy is better than that of mainstream algorithms such as GCN, T-GCN, GRU, LSTM.
    An Adaptive Watermark Embedding Algorithm Based on Entropy Theory
    ZHANG Shuai, YANG Xue-xia
    2020, 0(09):  37-42.  doi:10.3969/j.issn.1006-2475.2020.09.007
    Asbtract ( 177 )  
    References | Related Articles | Metrics
    In order to improve the imperceptibility and robustness of the digital image with watermark, a watermark embedding scheme based on image information entropy and edge entropy theory combined with fruit fly optimization algorithm is proposed. Firstly, the carrier image is divided into blocks, and the information entropy and edge entropy of each block are calculated. The two entropy values of each block are added and sorted. Then, the blocks with higher entropy value are selected according to the capacity of embedded watermark, and each bit of watermark is embedded into the blocks which are decomposed by wavelet transform and singular value. Finally, in order to further balance the contradiction between the transparency and robustness of the embedded watermark, the intensity of the embedded watermark is optimized by using the fruit fly algorithm. The experimental results show that the proposed method has better imperceptibility and is more robust than similar algorithms in the face of multiple types and intensity of simulated attacks.
    Registered Users Public Key Searchable Encryption Scheme with Secure-channel Free
    DENG Yin-juan, DU Hong-zhen, DOU Xiao-xia
    2020, 0(09):  43-48.  doi:10.3969/j.issn.1006-2475.2020.09.008
    Asbtract ( 153 )  
    References | Related Articles | Metrics
    In the existing SCF-PEKS scheme architecture, encrypting the server keyword public key depend on the user’s public key, which limit the service only to a certain user. The use of searchable encryption is greatly limited because users without the public key corresponding to the private key cannot search the data. The paper proposes a more efficient SCF-PEKS scheme which allows users to register for use based on composite-order bilinear group. This scheme allows multiple users to search the data without secure channels, users who need data search services finish keyword searches through registration. and the keyword public key encryption of the server no longer depends on the users public key. The scheme is proved to be secure against chosen keyword attack(IND-SCF-CKA) based on the decisional subgroup assumption in the standard model, which has a higher computing efficiency compared with current SCF-PEKS schemes.
    A Carousel Greedy Algorithm for Positive Influence Dominating Set Problem in Social Network
    WAN Ke
    2020, 0(09):  49-53.  doi:10.3969/j.issn.1006-2475.2020.09.009
    Asbtract ( 212 )  
    References | Related Articles | Metrics
    The problem of the smallest positive influence dominating set in a social network is a NP-hard combinatorial optimization problem. Aiming at this problem, there are two typical greedy algorithms with fast solving speed at present, but the quality of greedy solution needs to be improved. The carousel greedy method is to improve the quality of greedy solution without increasing the time complexity of greedy algorithm, and for some classical NP-hard problems, the experimental results show that the carousel greedy method can effectively improve their greedy algorithms. In this paper, the carousel greedy method is combined with the two greedy algorithms that have positive influence on the dominant set to improve the solution quality of the greedy algorithm, so as to propose the corresponding rotation greedy algorithm. The experimental results on several real social network instances show that, compared with the original greedy algorithms, the solution quality of the proposed carousel greedy algorithm is improved.
    Classification Method of Motor EEG Signals Based on Fractional Fourier Transform
    HUANG Xiao-shuang
    2020, 0(09):  54-59.  doi:10.3969/j.issn.1006-2475.2020.09.010
    Asbtract ( 155 )  
    References | Related Articles | Metrics
    Motor imaging EEG signals are typical non-linear and non-stationary, thus the traditional classification method based on single feature extraction is difficult to achieve better classification performance. Aiming at this problem, the Fractional Fourier Transform (FrFT) is introduced into the feature extraction process of EEG signals. Firstly, FrFT is used to analyze signals, the useful information is extracted from different dimensions while expanding the feature domain, and the feature vectors is formed. Then Support Vector Machine (SVM) classifier is used to classify the proposed feature vectors. Finally, the experiment is carried out using Graz data. The experimental results show that the proposed method can achieve up to 92.57% correct classification results, which is significantly higher than the traditional classification method using single feature extraction.
    A Feature-enhanced Tri-CNN Pedestrian Re-identification Method
    ZHOU Fang-yu, CHEN Shu-rong
    2020, 0(09):  60-65.  doi:10.3969/j.issn.1006-2475.2020.09.011
    Asbtract ( 139 )  
    References | Related Articles | Metrics
    Aiming at the problem of low resolution of extracted high level features and low recognition rate caused by occlusion in person re-identification, a feature enhanced pedestrian re-identification method based on Tri-CNN is established. Firstly, PCA dimensionality reduction is performed on the image features extracted from the pooling layer, and more discriminating pedestrian features are extracted according to CCA fusion features. Secondly, the spatial recursive model (SRM) is introduced to detect the features of occluded pedestrians in multiple directions, so as to improve the recognition rate of occluded pedestrians. Finally, according to the Euclidean distance measurement criterion, the distance between positive and negative sample pairs is verified respectively, and the loss function of Softmax and Triplet is combined to optimize the network model, so as to determine whether it is the same pedestrian. Experiments on MARS and ETHZ data sets show that the proposed method can effectively solve the problem of general occlusion recognition and significantly improve the accuracy of pedestrian re-identification.
    A Multi-stage Remote Sensing Image Object Detection Method
    MENG Xi-ting, JI Lu-yan,
    2020, 0(09):  66-72.  doi:10.3969/j.issn.1006-2475.2020.09.012
    Asbtract ( 146 )  
    References | Related Articles | Metrics
    The research of object detection in natural scenes and remote sensing scenes is extremely challenging. Although many advanced algorithms have achieved excellent results in natural scenes, the complexity of remote sensing images, the diversity of object scales, and the dense distribution of object make the research on remote sensing image object detection slow. This paper proposes a novel multi-category object detection model which can automatically learn the weights of feature fusion and highlight object features at the same time. As a result, the model achieves effective detection of small objects and densely distributed objects in complex remote sensing images. The experimental results of the model on public datasets DOTA and NWPU VHR-10 show that the detection effect exceeds that of most classical algorithms.
    A Denoising Method Based on CEEMD Wavelet Packet
    SUN Xiao-juan, WANG Li
    2020, 0(09):  73-76.  doi:10.3969/j.issn.1006-2475.2020.09.013
    Asbtract ( 178 )  
    References | Related Articles | Metrics
    Due to the characteristics of EEG is easy to be disturbed by noise, a method of the denoising EEG using CEEMD wavelet packet is proposed. Firstly, the EEG is decomposed by CEEMD to get a set of intrinsic mode function(IMF) components. Then the IMF component containing noise is denoised by wavelet packet threshold while the low-frequency IMF component of the signal is retained. Finally, the IMF components denoised by wavelet packet threshold and the reserved IMF components are accumulated and reconstructed to obtain the EEG after noise reduction. The experimental result shows that denoising of EEG using CEEMD wavelet packet can retain effectively the detailed characteristics of EEG when restraining noise, so as to achive good denoising performance.
    A Point Cloud Registration Algorithm Based on Multi-core Parallel and Dynamic Threshold
    LI Yun-chuan, WANG Xiao-hong, CHEN Si-ji, GE Yi-pan, LI Chuang
    2020, 0(09):  77-82.  doi:10.3969/j.issn.1006-2475.2020.09.014
    Asbtract ( 149 )  
    References | Related Articles | Metrics
    Aiming at the disadvantages of error correspondence points and low precision in point cloud registration, this paper proposes a point cloud registration algorithm based on multi-core parallel and dynamic threshold. This algorithm adopts the improved SAC-IA to complete rough registration for point cloud, and uses mainly OpenMP to realize the parallel extraction of the normal vector of point cloud query points, FPFH and parallel search of the correspondence points, so that the speed of the entire registration algorithm can be maintained or even improved. This paper uses the improved ICP algorithm to achieve registration in the point cloud fine registration. The improvement points focus on the culling of the error correspondence points and the dynamic determination of threshold. The center of gravity of registration points is used as the reference points. According to the dynamic threshold, the point pairs distance constraint is used to remove the error correspondence points. The experimental results show that the registration speed of this algorithm is improved when the registration accuracy is improved.
    Lightweight YOLOv3 Traffic Sign Detection Algorithm
    BAI Shi-lei, YIN Ke-xin, ZHU Jian-qi
    2020, 0(09):  83-88.  doi:10.3969/j.issn.1006-2475.2020.09.015
    Asbtract ( 282 )  
    References | Related Articles | Metrics
    Traffic sign detection has always been a hot topic in the field of autonomous driving. In deep learning algorithms, YOLOv3 and Faster R-CNN have obtained excellent object detection performance, but when detecting small targets, there are cases of missed detection. In order to accurately and quickly identify small targets in traffic sign detection, this paper proposes a lightweight YOLOv3 traffic sign detection algorithm. Multi-scale feature maps are obtained through convolutional neural networks while using both shallow and deep feature extraction. Deep features can effectively keep the detection accuracy from falling, and shallow features can effectively improve the accuracy of small target detection tasks. The model is compressed by the pruning algorithm, the trained model is sparsely trained, some unimportant convolution kernel channels are deleted, and the pruned model is fine-tuning to maintain the balance of parameters in the model file. The experimental results show that the accuracy of the model is improved by 2.3% by extracting the multi-scale feature map, the weight of the model is reduced by 70% by compressing the model with the pruning algorithm, and the detection time of the model is saved by 90%. Therefore, a lightweight traffic sign detection model with stronger robustness can be deployed on mobile embedded devices, without taking up large GPU computing resources but improving detection efficiency.
    Ship Detection in SAR Images of Ocean Based on Matrix Decomposition
    ZHENG Hui-min,
    2020, 0(09):  89-94.  doi:10.3969/j.issn.1006-2475.2020.09.016
    Asbtract ( 150 )  
    References | Related Articles | Metrics
    To solve the problem that the original image of the ocean is inconsistent with the data structure of the low-rank and sparse matrix decomposition model, this paper proposes a new ship detection method of the SAR images about ocean based on matrix decomposition. Firstly, the method needs to reconstruct the SAR images about ocean with similar structure. Then a new matrix decomposition model with higher resolution and faster decomposition speed, which is solved by the augmented Lagrange multiplier method, is designed according to the properties of recombination matrix, and the sparse components of the ship target are extracted without any clutter model and detection statistics. Finally, the morphological operation is used to realize the ship target detection in the SAR images about ocean. The experimental results based on the real data from Gaofen-3 SAR satellite show that compared with the existing ship detection methods based on robust principal component analysis, the proposed method can extract the ship target accurately from the sea clutter with better shape and faster speed, and has better robustness in dealing with complex sea conditions.
    Prediction of Debris Flow Based on Improved Extreme Learning Machine
    ZENG Ding, ZENG Yong
    2020, 0(09):  95-99.  doi:10.3969/j.issn.1006-2475.2020.09.017
    Asbtract ( 516 )  
    References | Related Articles | Metrics
    In order to improve the accuracy of debris flow prediction, an improved Extreme Learning Machine(ELM) algorithm based on DBSCAN clustering is proposed in this paper. Firstly, DBSCAN algorithm is used to cluster the training data about debris flow. Secondly, the ELM classifier is trained by classifying the different training sets obtained by clustering. Finally, the ELM classifier is used to predict the data of the prediction sets. The experimental results show that the accuracy of the improved ELM algorithm in predicting the occurrence of debris flow is 91.6% on average. Compared with the traditional ELM algorithm, the stability of the improved ELM algorithm is significantly improved. Compared with the traditional ELM algorithm, BP neural network and Fisher prediction method, the improved ELM algorithm has higher prediction accuracy.
    Base Station Location Planning Based on Improved Whale Optimization Algorithm
    TANG Li-qing, YING Zhong-yu, LUO Yun
    2020, 0(09):  100-105.  doi:10.3969/j.issn.1006-2475.2020.09.018
    Asbtract ( 195 )  
    References | Related Articles | Metrics
    Base station location planning is a significant optimization problem in network communication, and there is a great impact on the quality of network communication. Based on the constrained conditions of base station location planning, this paper constructs a base station location planning optimization model with the network coverage as the optimization index. The traditional optimization algorithms have some disadvantages such as slow convergence rate, easy to fall into local optimal, so this paper proposes an improved whale optimization algorithm. Firstly, aiming at improving the algorithm convergence rate, an adaptive changing strategy for convergence factor decreasing with the iteration number nonlinearly is introduced to improve global convergence ability. Then, the variation disturbance which obeys normal distribution is applied in some individuals to avoid premature convergence of the algorithm. The simulation results of the benchmark functions and the test example of base station location planning test problem show that the improved algorithm proposed in this paper can obtain a more ideal optimal solution and has faster convergence rate.
    Research and Simulation Analysis of Auxiliary Seat Requirement for Elderly Based on Jack
    YANG Qin, WANG Jia-bin, WANG Wei-xing
    2020, 0(09):  106-111.  doi:10.3969/j.issn.1006-2475.2020.09.019
    Asbtract ( 152 )  
    References | Related Articles | Metrics
    With the growth of age, the body condition is constantly changing. After entering the old age stage, the functions of the body are significantly weakened, unable to complete some regular actions normally. When using some conventional tools, it is necessary to use some devices with auxiliary functions to achieve better results. By the study of the physical parameters and muscle strength changes of the elderly, this paper determined the functional and objective requirement of the elderly for the auxiliary seat on the basis of in-depth research. The basic model of the auxiliary seat is established according to the functional requirement. Jack software is used to analyze the ergonomic simulation of the elderly auxiliary seat. Based on the objective requirement of the elderly, the comfort, visual field, forcesolver and reachability is analyzed. The results show that the ergonomic performance of the auxiliary seat is relatively good, but some aspects need to be improved.
    Encoder-Decoder Photovoltaic Power Generation Prediction#br# Model Based on Attention Mechanism#br#
    SONG Liang-cai, SUO Gui-long, HU Jun-tao, DOU Yan-mei, CUI Zhi-yong
    2020, 0(09):  112-117.  doi:10.3969/j.issn.1006-2475.2020.09.020
    Asbtract ( 152 )  
    References | Related Articles | Metrics
    The weather factors that affect the output of photovoltaic power generation systems have great volatility and discontinuities. Therefore, it is necessary to create a suitable prediction model to accurately predict the characteristics of photovoltaic output to ensure the efficient operation of the power generation network. This paper selects the appropriate historical photovoltaic power generation data through the maximal information coefficient, uses it as one of the features to reconstruct the input data, and attention mechanism is introduced on the Encoder-Decoder model constructed by LSTM neurons to obtain an attention-based Encoder-Decoder photovoltaic power generation prediction model. The analysis of actual photovoltaic power plant examples verifies the accuracy and applicability of the proposed model in predicting photovoltaic power generation.
    A Wheel Polishing Workstation System Based on Multi-industrial Robot
    LIU Hai-long, ZHANG Lei, WU Hai-bo
    2020, 0(09):  118-121.  doi:10.3969/j.issn.1006-2475.2020.09.021
    Asbtract ( 469 )  
    References | Related Articles | Metrics
    This paper proposes a design scheme of dual-industrial robot collaborative polishing work station, which takes automobile wheel hub polishing as the object of the research and aims at the difficult problem of trajectory planning and automatic production coordination of its inner hollow-out polishing. Firstly, according to the actual production process of the automobile wheel hub, SolidWorks and RobotStudio are used to design relevant model components and build the overall layout of the workstation. Then the electrical control model of Smart dynamic logic components, sensors and I/O communication network are created and designed. Finally, the system function verification and simulation are realized by offline programming of multi-industrial robots. The proposed scheme provides design basis and experimental platform for the realization of the auto wheel polishing automatic production line, shortens the development cycle of the production line, saves the operator time, and improves working efficiency.
    A Scheduling Method of Smart Grid Based on Genetic Algorithm
    WU Hai-wei, WANG Xiao-zhong, ZHU Fa-shun,
    2020, 0(09):  122-126.  doi:10.3969/j.issn.1006-2475.2020.09.022
    Asbtract ( 242 )  
    References | Related Articles | Metrics
    With the development of smart grid, the scale and types of data obtained from dispatch control systems increase exponentially, and dealing with these data is relatively complex. In order to better perform power dispatching to provide corresponding decision support for power system and better serve customers to meet the users power needs at different times, this paper proposes a G-DSM algorithm based on genetic algorithm, which can control server appliances. In the algorithm, the load scheduling problem is defined as cost minimization problem and solved by genetic algorithm. The algorithm combines with the large amount of power big data obtained from the user side to plan the users power demand, reduces the users cost and peak power load, thereby avoiding the waste of power resources and improving the work efficiency of the power grid. Experimental results show that the algorithm has good feasibility and is easy to implement in actual operation.