Table of Content

05 July 2021, Volume 0 Issue 06

Previous Issue    Next Issue

Multi-UAV Hunting Based on Improved Whale Optimization Algorithm

LING Wen-tong, NI Jian-jun, CHEN Yan, TANG Guang-yi

2021, 0(06):  1-5. 

Asbtract ( 374 )   PDF (1931KB) ( 419 )  

References | Related Articles | Metrics

UAV hunting is a challenging and realistic task. In order to enable UAVs to hun moving targets successfully and effectively, a multi-UAV hunting algorithm based on dynamic prediction of hunting points and improved whale optimization algorithm is proposed. When the environment is unknown and the target motion trajectory is unknown, this paper first uses polynomial fitting to predict the target motion trajectory, obtains the prediction point by dynamically predicting the number of steps, sets up hunting points around it, and then uses the two-way negotiation method to make reasonable assign each target point. Aiming at the shortcomings of the whale optimization algorithm that it is easy to fall into the local optimum, a method based on adaptive weights and changing the position of the spiral is proposed to improve the development ability and search ability of the algorithm. Finally, several experimental simulations were carried out in different experimental environments, and the experimental results showed the effectiveness of the proposed algorithm.

Short Text Classification Based on Improved CHI and TF-IDF

DAI Ji-peng, SHAO Feng-jing, SUN Ren-cheng

2021, 0(06):  6-11. 

Asbtract ( 333 )   PDF (941KB) ( 134 )  

References | Related Articles | Metrics

In order to improve the effect of classifying short texts with a small amount of data, and effectively reduce the feature dimension of the feature space, aiming at the defects of the traditional CHI statistical method and the TF-IDF weight calculation method, this paper proposes a new factor of word class and frequency to improve the feature selection method, and consequently to enhance the classification accuracy. As the traditional CHI statistical method is sensitive to low-frequency words, and the TF-IDF weight calculation method ignores the distribution of feature items between and within classes, the paper introduces the factor of word class and frequency to improve the traditional CHI statistical method and the TF-IDF weight calculation method, and uses the two methods in combination to reduce the interference caused by low-frequency words, with consideration to the special situation of the distribution of feature words within and between classes. The paper uses the XGBoost classification algorithm to apply the proposed method in the classification experiment of topic text with small amount of data and short text. The experimental results show that, compared with the traditional CHI and TF-IDF methods, the feature selection method with factor of word class and frequency improves the classification accuracy on the balanced and unbalanced corpus, and greatly reduces the memory usage. 

A Capacity Satisfaction Calculating Method Based on Inf-ProA Architecture Framework

FAN Zhi-qiang, CAO Jiang, NIU Chan, WANG Li-ting

2021, 0(06):  12-17. 

Asbtract ( 219 )   PDF (743KB) ( 106 )  

References | Related Articles | Metrics

Capability satisfaction is an important indicator to measure system construction. The ability satisfaction analysis based on the architecture framework can realize the analysis and judgment of the ability satisfaction degree in the system design stage, avoiding the increase of the construction cost in the later period caused by insufficient early analysis. In order to analyze capability satisfaction in the early stage, a capability satisfaction calculation method based on Inf-ProA architecture framework is proposed. At first, we build a core architecture elements to trace relationships meta model based on ability. Then, the weight calculation method and the satisfaction calculation method of the core architecture element tracking meta-model is constructed. Moreover, a capability satisfaction method is proposed based on the weight, satisfaction value and the tracking relation chain of the core elements. Finally, an application case is used to verify this proposed method, and the experimental results show that this method can effectively analyze the capability satisfaction.

DF-SSD: A One-stage Small Target Detection Algorithm Based on Deconvolution and Feature Fusion

WANG Liang-wei, CHEN Mei, LI Hui, LI Huan-juan, SHI Ruo, DAI Zhen-yu

2021, 0(06):  18-23. 

Asbtract ( 279 )   PDF (1801KB) ( 104 )  

References | Related Articles | Metrics

Aiming at the problem of the SSD model’s poor detection performance on small targets, the DF-SSD algorithm was proposed, its technical contributions include a one-stage detector method based on deconvolution and feature fusion and an improved default bounding boxes’ size calculation algorithm. Deconvolution and feature fusion can increase the semantic information of shallow feature layers. In DF-SSD algorithm, the improved default bounding boxes’ size calculation introduces the characteristics of the data set, which can effectively use each default bounding box for training and prediction. Compared with the improved R-SSD and RSSD models based on SSD, the DF-SSD method has higher detection accuracy. At the same time, DF-SSD’s detection overhead is only 1/2 of R-SSD and 1/5 of DSSD. The MAP of the DF-SSD on the VOC2007 and DIOR data sets is 1.4 and 3.6 percentage points higher than that of SSD respectively. Meanwhile, DF-SSD’s MAP of small targets of ship, vehicle, windmill, and cat increased 23.2, 12.6, 8 and 4.8 percentage points respectively. The results show DF-SSD effectively improves the detection accuracy of small targets and has a faster detection speed.

Risk Assessment Model of Accounting Resource Sharing Management Based on Genetic Algorithm

YANG Xiao-dong

2021, 0(06):  24-28. 

Asbtract ( 173 )   PDF (1042KB) ( 95 )  

References | Related Articles | Metrics

As the risk assessment value of accounting resource sharing management is not the global optimal value, the assessment result is not accurate, and the assessment time is prolonged. Therefore, a risk assessment model of accounting resource sharing management based on genetic algorithm is proposed. By selecting the risk assessment index of accounting resource sharing management, using Delphi method to determine the weight of the risk assessment index, this paper constructs the quantitative model of accounting resource sharing management risk assessment. On this basis, this paper uses genetic algorithm to calculate the optimal solution of parameters, makes risk assessment rules to determine the risk degree, so as to realize the assessment of accounting resource sharing management risk based on genetic algorithm. According to the experimental results, compared with the traditional risk assessment model of accounting resource sharing management, the risk assessment model of accounting resource sharing management greatly reduces the assessment time, and can accurately assess the risk of accounting resource sharing management, which fully shows that the model has better assessment performance.

A Collaborative Filtering Algorithm Based on Information Entropy and Improved Similarity

HUANG Hao, CHEN Li

2021, 0(06):  29-34. 

Asbtract ( 185 )   PDF (1831KB) ( 77 )  

References | Related Articles | Metrics

In order to reduce the noisy data and data sparsity problems in the collaborative filtering algorithm, and improve the accuracy of the algorithm, a collaborative filtering algorithm based on information entropy and improved similarity is proposed. The user information entropy model is used to judge the noise data to eliminate the interference of the noise data on the experimental results; the improved similarity calculation method for sparse data is used, and all the score data are used. Rather than relying on common scoring items to calculate, it is of great help to alleviate the impact of sparse data on the accuracy of the recommended results. Experimental results show that the algorithm can eliminate the influence of noisy data on the results to a certain extent, alleviate the interference of data sparseness on the accuracy of recommendation results, improve the accuracy of the recommendation algorithm, and alleviate some common problems in traditional recommendation system algorithms. Compared with the traditional collaborative filtering algorithms, the accuracy of the algorithm is higher.

Application of Deep Learning in Defect Detection of Mobile Phone Data Interface

LIU Ya-dong, LUO Yin-sheng, CAO Yang-yang, SONG Wei

2021, 0(06):  35-40. 

Asbtract ( 217 )   PDF (2439KB) ( 82 )  

References | Related Articles | Metrics

In order to better detect the defects of the mobile phone data interface, this paper proposes a detection algorithm based on Faster R-CNN. The specific research method is to replace RoIPooling in the Faster R-CNN detection architecture with RoIAlign to solve the deviation of the target return position caused by the two quantifications in the RoIPooling calculation process. The ResNet50 fusion FPN network is used as a feature extraction network to improve the model’s detection effect on small target defects. Finally, the test set is used for prediction. Experiments show that the mean average accuracy (mAP) of the proposed algorithm in this paper has reached 91.17%, which is 24.72 percent points higher than mAP when VGG is used as the feature extraction network, and is 2.58 percent points higher than mAP when ResNet50 is used alone as the feature extraction network. Therefore, the algorithm proposed in this paper has a significant effect on mobile phone data interface defect detection, and provides a more effective detection method for mobile phone data interface defect detection.

Dynamic Gesture Recognition Based on Space-time Feature Extraction of Neural Network

LIN Zhi-wei, ZHU Wen-zhang, CHEN Hao

2021, 0(06):  41-47. 

Asbtract ( 228 )   PDF (2794KB) ( 105 )  

References | Related Articles | Metrics

Aiming at the existing 3D convolution method of dynamic gesture recognition with large number of computational parameters, it is difficult to extract 2D convolution of long-time-series images in terms of time dimension. In this paper, a gesture recognition method based on the combination of 2D convolutional neural network and long and short term memory network is proposed. Firstly, spatial features are extracted based on 2D convolutional neural network, and then features in time dimension are extracted by interrelation of sequential images through long and short term memory network. In order to verify the validity of the algorithm in this paper, using the acquisition of 7 kinds of dynamic hand gestures and IsoGD public data sets to verify this proposed algorithm, the experimental results show that under using online enhancement algorithm the paper in the collection on a set of dynamic hand gestures recognition rate reached 87.14%, IsoGD public data sets on recognition rate of 57.89%, compared with the existing method, the recognition rate is improved.

Image Fusion Algorithm  Based on Improved NSCT Infrared and Visible Light

YANG Bin, HUANG Run-cai, WANG Cong-ao

2021, 0(06):  48-53. 

Asbtract ( 165 )   PDF (3224KB) ( 66 )  

References | Related Articles | Metrics

Aiming at the problem that the target details are easily lost in the process of infrared and visible light image fusion, an image fusion algorithm of combining non-subsampled contourlet transform (NSCT) with principal component analysis (PCA) is proposed. First, NSCT is used to decompose the source image to obtain low-frequency and high-frequency sub-band images. In the low-frequency sub-band coefficients, because PCA can highlight the main information of the image, the principal component analysis method fusion rule is selected. In the high-frequency sub-band, relatively high-level coefficients express the most detailed information in the source image, the absolute maximum method fusion rule is selected, while the low-level coefficients represent rougher information, the absolute maximum and regional standard deviation fusion rules can be selected. It can be concluded from the experimental results that the algorithm is superior to other algorithms in the fusion effect of target information and detailed information in infrared and visible images, and has better image visual effects.

Road Structure Based Steering Control Without Localization for Vehicle out of Garage

XIONG Ying, ZHOU Ya-qi, MAO Xue-song

2021, 0(06):  54-60. 

Asbtract ( 138 )   PDF (2106KB) ( 73 )  

References | Related Articles | Metrics

To solve the problem that unmanned vehicles cannot acquire their positions for planning a path in global reference frame when they are driving in an underground garage where the localization signal is weak, a steering control method was deduced theoretically by considering geometry of the road edge for aiding vehicles out of garage. Firstly, driving scene of the vehicle and low speed vehicle model were given. Then, vehicle pose relative to road edge and turning curvature at turning point were deduced theoretically, in addition to the steering angle control method at each road segment. Finally, driving behavior by the method was simulated under the condition that road edge data are ideal and non-ideal, respectively. Simulation results show that the method can realize self-driving without using location if the road edge measurement error is limited.

Surface Defect Detection Based on Improved Deep Metric Learning Algorithm

WANG Wei, , YU Hou-yun,

2021, 0(06):  61-68. 

Asbtract ( 204 )   PDF (2275KB) ( 103 )  

References | Related Articles | Metrics

An algorithm based on deep metric learning is proposed for the surface defect detection of small batches and multiple varieties of industrial products. The algorithm improves the VGG16 network model, which is more suitable for mapping the original image to the latent space; for the task of product surface defect detection, a conditional triplet loss function is proposed to strengthen the fitting ability of neural network. When the defect is judged in the latent space, the classification model based on the KNN algorithm in the original metric learning algorithm is discarded, and the classification method based on the Gaussian distribution probability is proposed. When new types of products are detected, on the basis of the trained network model, the network is finely tuned by using the image data of the new product as input. Through the above improvements to the deep metric learning algorithm for the defect detection task, after K-Fold cross-validation on the button defect data set, the accuracy on different query sets is over 90%, and the highest can reach 99.89%, by providing 50 non-defective samples and 50 defective samples for training the network. And compared with traditional deep metric learning algorithms, the accuracy is increased by about 10%. The experimental results show that the improved deep metric learning algorithm can well solve the surface defect detection problem of small batch and multi-variety industrial products.

Load Forecasting Method of Distribution Network Based on Neural Network

LI Yan-sheng

2021, 0(06):  69-73. 

Asbtract ( 142 )   PDF (2912KB) ( 120 )  

References | Related Articles | Metrics

The distribution network connected to high permeability distributed photovoltaic should reduce the load of distribution network to a certain extent. Due to the difference of the coupling characteristics of the load, photovoltaic output and meteorological factors in the distribution network, and the strong randomness, it is difficult and randomness to predict the net load of the distribution network. In order to realize the short-term prediction of the net load of the fluctuating distribution network, the short-term prediction model of the neural network is constructed based on the long-short term memory (LSTM). The short-term prediction model of photovoltaic output and the hourly load prediction model of distribution network are built by LSTM, and cross validation is used to optimize the structure super parameters of each LSTM predictor. The net load of distribution network is obtained by comparing the predicted results. From the analysis of the experimental results, it can be seen that the LSTM method can adaptively mine the correlation between photovoltaic output characteristics and historical load forecasting objects. Compared with the support vector regression (SVR) method, this method has high prediction accuracy and simple process.

Multi Feature Load Forecasting Model for LSTM Network in Mobile Cloud Computing

CHEN Si-yu, ZHUANG Yi, LI Jing

2021, 0(06):  74-85. 

Asbtract ( 160 )   PDF (2341KB) ( 77 )  

References | Related Articles | Metrics

Aiming at the problem that the load of mobile virtual machine changes greatly and it is difficult to be predicted accurately, this paper proposed an AR-LSTM-ED load forecasting model based on joint feature selection, which can carry out single-step and multi-step prediction of cloud host load. In this paper, the joint feature selection method was used to obtain other load series related to the target prediction load series, and the undecimated wavelet transform method suitable for online prediction was used to decompose the target prediction features into subsequences which could be easier to be predicted. Finally, these sequences and target prediction sequences were input into AR-LSTM-ED model. The model used the long-short term memory encoder-decoder network to predict the target load, which could capture the long-term dependence. The autoregressive model (AR) was further combined to predict the linear data in the load. We used the Google cloud computing data set to verify the algorithm. The experimental results show that the proposed method achieves better performance.

A Routing Algorithm for Wireless Sensor Networks in Concave Mountain Terrain

ZHANG Yan-hu, YAN Li-juan, KONG Shu-mei

2021, 0(06):  86-90. 

Asbtract ( 145 )   PDF (860KB) ( 65 )  

References | Related Articles | Metrics

Wireless sensor networks have been widely used in people’s life. Based on the self-organization and energy consumption characteristics of WSN (Wireless Sensor Networks) and LEACH routing protocol algorithm, this paper proposes a wireless sensor network algorithm suitable for concave mountain terrain that optimizes the information transmission direction of nodes. The algorithm improves the WSN networking method and adopts the method of preferring cluster heads closer to the base station for networking. Firstly, the network randomly generates the first batch of cluster-head nodes. Secondly, each cluster-head node collects the information of the nodes in the cluster, and meanwhile collects the remaining energy and position coordinate information of each node, which is then summarized and sent to the base station. Thirdly, the base station determines the cluster-head node of the next round according to the obtained information and broadcasts it to the whole network. Finally, each node selects the cluster head closer to the base station to network into the cluster to complete the information transmission. Through the simulation test on Matlab simulation software, the experimental results show that the algorithm described in this paper can effectively improve the wireless network life cycle, to a certain extent, balance the energy consumption of each node of the wireless network, and extend the service life of the network.

Load Balancing Technology Under Big Data Architecture Based on HBase

LEI Ming, JIANG Han-sheng, WU Guo-liang, ZHAO Yu-juan, LIANG Jian

2021, 0(06):  91-95. 

Asbtract ( 178 )   PDF (1129KB) ( 87 )  

References | Related Articles | Metrics

 With the continuous growth of the scale and type of meteorological data, meteorological data has gradually entered the stage of massive service, and meanwhile providing more agile data services based on the background of big data has become an urgent demand for business development. In this paper, a load balancing algorithm and strategy based on HBase system is proposed for semi/unstructured meteorological data. In the actual test and comparison, it is found that the system can meet more than 2 million grid points and 100 concurrent scenarios, and the query speed is within 2 s. Compared with the load balancing algorithm which is not added load balance, the system data response speed is improved by 42.69 times, which can effectively meet the actual business needs.

High Precision Cymometer Based on FPGA

DONG Bo, WANG Zhi, YU Hang, LIU Bo

2021, 0(06):  96-99. 

Asbtract ( 240 )   PDF (1334KB) ( 92 )  

References | Related Articles | Metrics

In order to meet the requirements of hardware engineers for high precision and high bandwidth frequency measurement instruments, a high precision cymometer based on FPGA is designed. The frequency meter consists of peripheral voltage following circuit and serial communication circuit, frequency divider module, frequency measurement module and serial communication module on FPGA. The Cyclone IV chip of Altera company is used as the control core. Firstly, the signal to be measured is stabilized and isolated by voltage follower, and then the stabilized signal is connected to the frequency divider module. The frequency divider module divides the frequency signal into low-frequency and high-frequency signals with the limit of 1 kHz, and uses periodic frequency measurement method and pulse counting method to measure the frequency of low-frequency signal and high-frequency signal respectively. The measured frequency data can be uploaded to the upper computer through the serial port in real time. After testing, the frequency meter can achieve 1 Hz accuracy, 200 MHz frequency measurement bandwidth and multi-channel detection.

A Dynamic Public Key Searchable Encryption Scheme Against Keyword Guessing Attack

LI Zhi-yi, WANG Xue-ming, WANG Ze-xian

2021, 0(06):  100-106. 

Asbtract ( 176 )   PDF (915KB) ( 69 )  

References | Related Articles | Metrics

Ciphertext search can be used to protect user’s privacy when searching in plaintext. However, the index struction of most existing PEKS schemes are based on file keyword pairs, and each search needs to traverse all files, which will make the search efficiency of the scheme low. Moreover, the existing PEKS schemes only support static search operations, and the files in the cloud can not be updated dynamically, which will inevitably lead to waste of resources and inconvenience in using. In view of the existing problems, this paper proposes a more efficient dynamic public key searchable encryption scheme, which has certain performance advantages compared with other schemes, and proves that the scheme has the security under the adaptive dynamic keyword selection attack under the random oracle model, and can resist the keywords guessing attack.

Robust Medical Image Encryption Algorithm Based on Compound Chaos

GAO Guo-jing, LYU Qing-wen

2021, 0(06):  107-112. 

Asbtract ( 177 )   PDF (3711KB) ( 82 )  

References | Related Articles | Metrics

Aiming at the low degree of image scrambling of the current single chaotic image encryption algorithm, which makes the encryption result vulnerable to known attacks and the problem of low security, a robust medical image encryption algorithm based on compound chaos is proposed. The feedback mechanism of image encryption is set as the theoretical support of algorithm implementation, and read the original image of robust medical. According to the establishment conditions of the chaotic model, a compound chaotic model is constructed. Choosing the Logistics chaotic map, generating a compound chaotic sequence and performing multiple iterations, scrambling and encrypting the pixels of the initial image of the robust medical image, replacing and diffusing, and generating the corresponding key to achieve the encryption of the robust medical image. Compared with the traditional encryption algorithm, the number of scrambling bits of the paper method is increased by 886 bits, that is, the degree of scrambling is increased by 8.7%, and the correlation of image sequences is low and the ability of anti attack is strong. The experiment results show that the security of this robust type of medical image encryption algorithm is high.

A Novel Hyper Chaotic Image Encryption Algorithm Using Four Directional Diffusion Based on Matrix

GE Bin, CHEN Gang, FANG Rui, LIAO Zhong-zhi

2021, 0(06):  113-119. 

Asbtract ( 217 )   PDF (3603KB) ( 104 )  

References | Related Articles | Metrics

A new hyper chaotic image encryption algorithm using four directional diffusion on matrix is proposed to overcome the low efficiency of most hyper chaotic image encryption algorithm. Firstly, the original sequence generated by the hyper chaotic system is quantized and reconstructed to generate the key matrix with the same size as the plaintext image. Secondly, the initial vector required by the diffusion process is generated by the low-dimensional chaotic system. Finally, the vectorized operation is used to complete the global diffusion of pixel information from the top, bottom, left and right. The simulation results show that the algorithm can resist the attacks such as plaintext selection and differential analysis, and the time complexity of the algorithm is only O(M+N), which has both security performance and operation efficiency, and has a good effect  in the field of image real-time secure communication.

Blind Watermarking Algorithm of 3D Mesh Models Based on Logistic Chaotic Encryption

YAO Xiao-lu, ZHANG Guo-you, WANG Jiang-fan, CUI Jian

2021, 0(06):  120-126. 

Asbtract ( 181 )   PDF (3823KB) ( 74 )  

References | Related Articles | Metrics

In order to protect the copyright information of 3D model data, a blind watermarking method based on Logistic chaotic encryption is proposed. Firstly, the binary image generated by dot matrix font is used as watermark information, and the watermark image is scrambled and encrypted. Then, the mesh simplification method is used to display the 3D mesh model in multiple layers, and the local set area of each vertex in the middle layer is calculated, and then the watermark information is embedded by modifying the small and dense vertices in the middle layer. Finally, the watermark strength is adaptively embedded according to the curvature of the neighborhood of the vertex. Watermark extraction is the  inversion process of watermark embedding. Experimental results show that the algorithm is robust to affine transformation, simplification, and smooth attacks, which not only achieves the purpose of protecting the copyright of the three-dimensional model, but also improves the security of the watermark system.