Computer and Modernization

Prediction of COVID-19 Based on Mixed SEIR-ARIMA Model

DONG Zhang-gong, SONG Bo, MENG You-xin

2022, 0(02): 1-6.

Asbtract ( 583 )

PDF (1176KB) ( 501 )

References | Related Articles | Metrics

Novel coronavirus pneumonia, referred to as COVID-19, is an acute infectious pneumonia caused by novel coronavirus, which is of highly infectious and generally susceptible to the population. Therefore, the prediction of the number of novel coronavirus pneumonia infections is not only beneficial for the country to make scientific decisions in the face of the epidemic, but also facilitates the timely integration of epidemic prevention resources. In this paper, a hybrid model SEIR-ARIMA constructed by the model SEIR based on the traditional infectious disease dynamics and the differential integrated moving average autoregressive model ARIMA is proposed to make prediction and analysis of the novel coronavirus pneumonia epidemic in different time periods and locations. From the experimental results, the prediction based on the SEIR-ARIMA hybrid model has better prediction effect than the common logistic regression Logistic, long short-term memory artificial neural network LSTM, SEIR model, and ARIMA model used for COVID-19 prediction. In order to truly reflect whether the improvement of the experimental effect originates from the advantage of combining SEIR and ARIMA models, this paper also implements the SEIR-Logistic hybrid model and SEIR-LSTM hybrid model, and compares the analysis with SEIR-ARIMA to conclude that both SEIR-ARIMA predictions achieve better prediction results. Therefore, the analysis of the development trend of COVID-19 based on the SEIR-ARIMA hybrid model is relatively reliable, which is conducive to the scientific decision-making of the country in the face of the epidemic and has good application value for the prevention of other types of infectious diseases in China in the future.

Improved Immune Particle Swarm Optimization Algorithm for Automatic Parallel Parking Based on Cubic Spline Interpolation#br#

WANG Zhe, WANG Long-da, LIU GANG, WANG Xing-cheng, WANG Zhong-jun, BAO Lu-jie

2022, 0(02): 7-12.

Asbtract ( 144 )

PDF (1211KB) ( 140 )

References | Related Articles | Metrics

It is difficult to obtain the smooth, accurate and optimal parking trajectory by using traditional automatic parallel parking optimization algorithm. For obtaining ideal optimal parking target trajectory, combined with the intelligent automatic parking theory, an automatic parallel parking method based on cubic spline interpolation is proposed. In order to improve the optimization performance for automatic parallel parking optimization algorithm effectively, an immune improved particle swarm optimization algorithm (IIPSO) based on cubic spline interpolation is proposed for choosing an appropriate parking position reference points by using shortest parking trajectory as optimization target. Firstly, for enhancing the global search performance and convergence velocity of particle swarm optimization (PSO), an adaptive mutation strategy is introduced. Secondly, an immune strategy is introduced to improve the global optimization ability of particle swarm optimization. The simulation results of test functions and the practical example of automatic parking indicate that the IIPSO algorithm proposed in this paper has better optimization precision and faster convergence speed.

Optimization of MD5 Decryption Algorithm Based on Many-core Sunway Processor

ZHANG Heng, ZHAO Rong-cai, DONG Ben-song,

2022, 0(02): 13-18.

Asbtract ( 166 )

PDF (1231KB) ( 105 )

References | Related Articles | Metrics

The current MD5 decryption algorithm cannot adapt to the Sunway architecture and cannot give full play to the performance advantage of SW26010 multi-core processor. Aiming at the above problems, optimization methods such as hashing initialization, loop unrolling, link variable optimization, 61-step optimization and application memory optimization are adopted to optimize the single core to improve the speed of the decryption algorithm. In addition, the optimized decryption algorithm is rewritten into the master-slave mode, and the computing tasks are assigned to 64 slave cores for parallel execution. The storage access mode of master-slave core is optimized to reduce the time overhead of memory access to the program. By the 5 groups of tests with different task loads, the experimental results show that the average acceleration ratio is 12.28 after optimization on the single core and 44.84 after optimization on the slave core. The experimental results show that the MD5 decryption algorithm optimization method based on SW26010 multi-core processor is feasible and effective.

Identification of Data-driven ADS-B Interference Source Signal Type

HU Yan, ZHUO Shu-long, SI Cheng-ke

2022, 0(02): 19-25.

Asbtract ( 206 )

PDF (1670KB) ( 89 )

References | Related Articles | Metrics

When extracting the subtle features of interference signals, the traditional identification methods of interference source signal types have some shortcomings, such as low accuracy and poor recognition effect. In this paper, a deep neural network based ADS-B interference signal modulation type recognition algorithm is proposed. Firstly, ADS-B signal and interference waveform are superimposed and mixed. Simulation signals are transmitted by controlling vector signal generator (VSG) and collected at the receiving end. Then, random noise is artificially added to the received baseband I and Q data, and based on this, tensor training sample datasets are constructed under various SNR scenarios. Finally, the training sample data are used to train the neural network designed in this paper, and the recognition performance of the traditional classification algorithm and that of the neural network algorithm proposed in this paper are compared and analyzed in the sample data set. Experimental results show that the neural network algorithm proposed in this paper has better recognition performance compared with the existing traditional recognition algorithms.

Similarity Measurement Method of Inf-ProA Information Activity Process Model

ZOU Meng-yuan, FAN Zhi-qiang, XU Luo, LIU Jie, LIANG Wan-lu

2022, 0(02): 26-32.

Asbtract ( 173 )

PDF (1056KB) ( 94 )

References | Related Articles | Metrics

With the development and advancement of the top-level design work of the entire army, a certain number of architecture design model assets have been gradually accumulated and formed in the use of the Inf-ProA framework and its supporting tools to carry out architectural design work in the military field. When architects are designing models, there is a wide range of requirements for referencing existing similar models. However, the existing architecture methods and tools are still unable to measure and recommend similar models with reference value. Based on the basic architecture design elements of the Information Activity Model (IAV-1a), through the research of similarity measurement on extensible language documents and unified modeling language model, this article puts forward a method that can be used to measure the similarity of Inf-ProA information activity process model. This method measures the similarity of the content and structure of the information activity process model, and considers the design similarity close to the final design results of the model, and uses the practical experience of architecture design to make the similarity measurement results more reasonable. The result of the similarity measurement can be used to provide a maximum similarity match for the model being designed, which is of great significance for recommending to the architect to complete the auxiliary design. The experimental results show that the method can effectively measure the similarity of the information activity process model.

Marine Group Target Mining Based on Imporved FP-growth

YUE Jian-cheng, WANG Yu-mei, WU Ya-fei, ZANG Yi-hua

2022, 0(02): 33-37.

Asbtract ( 113 )

PDF (753KB) ( 71 )

References | Related Articles | Metrics

The status of marine targets presents a complex and changeable situation. It needs to quickly excavate the group information of marine ships and provide group data support for mastering the situation of marine targets. This paper uses improved FP-growth algorithm to mine marine ships’ data, and uses the method of spatio-temporal segmentation to divide the targets area and mine frequent items. First, the original data is cleaned to get the effective data; secondly, the linear interpolation method is used to process the ship trajectory for subsequent calculation; then, FP-growth algorithm is used to build FP-tree; finally, the frequent term set is obtained to mine the information of marine ship groups. Aiming to the problem of low efficiency of association analysis based on itemset partition, this paper uses Hash table to split database and the method of node exchange to mine frequent itemsets, and compares the efficiency of the algorithm in memory consumption and time consumption. The test is done on AIS data set to verify the efficiency of the improved algorithm, with the given confidence and support of the target group information.

An Initialization Algorithm of HRG Model and Its Application in Link Prediction

ZHU Ding-kai, TIE Zhi-xin, HONG Shun-he

2022, 0(02): 38-44.

Asbtract ( 161 )

PDF (1150KB) ( 86 )

References | Related Articles | Metrics

In the process of using the hierarchical random graph (HRG) model to predict the link of the real network, it is necessary to construct an initial hierarchical random graph to initialize the Markov chain to run the Markov chain Monte Carlo sampling algorithm. In view of the inefficiency of the existing hierarchical random graph initialization scheme, this paper reconstructs the initial hierarchical random graph model and proposes a new hierarchical random graph model initialization algorithm. The algorithm is divided into two stages, in the first stage, similarity index (LHN-I index) is introduced to sort the edges in the network； In the second stage, the hierarchical random graph model is constructed by using the sorted edges. In this process, a method is designed to insert the network vertices into the hierarchical random graph model. The performance of the proposed algorithm is compared with the existing algorithm through three example networks. The experimental results show that the initial hierarchical random graph constructed by the proposed initialization algorithm not only has higher likelihood value, but also makes the Markov chain Monte Carlo algorithm converge faster, thus reducing the time consumption of link prediction. In addition, in the link prediction experiment, the improved link prediction algorithm based on hierarchical random graph model has better prediction accuracy than some link prediction algorithms based on similarity index.

Trustworthy Encryption Traffic Classification Method Based on RBF Neural Network

ZHANG Xiao-hang, LI Zheng, ZHU Xiao-ming, ZHANG Hai-feng, ZHAO Bo-yu

2022, 0(02): 45-51.

Asbtract ( 309 )

PDF (2888KB) ( 168 )

References | Related Articles | Metrics

Network traffic classification is widely used in research fields such as network resource allocation, traffic scheduling and intrusion detection systems. With the popularization of encryption protocols and the rapid development of network traffic, the traffic classifier based on deep learning has gradually attracted the attention of researchers due to its feature of automatically extracting features and high classification accuracy. But there has been no research on the credibility of network traffic classification. This article proposes a trustworthy deep learning model to classify encrypted network traffic. The proposed algorithm is based on the idea of RBF network and uses a new loss function and centroid update scheme for training. By using gradient penalty to force detection of input changes, it can effectively detect out-of-distribution data. On the two public ISCX VPN-nonVPN and USTC-TFC2016 traffic data sets, the proposed algorithm achieves 98.55% of the AUROC index, and has achieved the best out-of-distribution detection effect compared with similar algorithms. Extensive experimental results show that the proposed algorithm has high classification performance and can effectively detect out-of-distribution traffic data, which improves the credibility of the traffic classification model.

Underwater Localization Algorithm of Range Correction Based on Long Short-Term Memory

JI Ping, GUO Ying

2022, 0(02): 52-57.

Asbtract ( 128 )

PDF (928KB) ( 91 )

References | Related Articles | Metrics

Underwater sensor network enables humans to detect and develop marine resources， but the data collected by underwater sensor nodes loses their value when accurate localization information is lost. For many underwater localization algorithms that have been widely used are still difficult to achieve accurate ranging, therefore their localization accuracy is too low and unsatisfactory. This paper proposes an underwater localization algorithm based on Long Short-Term Memory modified ranging value to improve localization accuracy. The algorithm uses a variant model of Recurrent Neural Network, namely Long Short-Term Memory (LSTM), to improve the time difference of arrival ranging algorithm. LSTM is trained by the historical data of the marine environment and the ranging value. It can efficiently and accurately predict the current ranging correction value, so as to achieve the effect of optimizing the ranging error. And the effective combination of the above two is utilized to further improve the multilateral localization algorithm to achieve precise positioning of unknown underwater nodes. Finally, the simulation experiment and algorithm comparison prove that the algorithm proposed in this paper does have high localization accuracy and feasibility.

Substation Equipment Operation and Maintenance System and Its Construction Based on Digital Twin#br#

YANG Ke-jun, ZHANG Ke, HUANG Wen-li, CHEN Bo-wen

2022, 0(02): 58-64.

Asbtract ( 313 )

PDF (3097KB) ( 209 )

References | Related Articles | Metrics

In view of the problems of difficult periodic state control and low operation and inspection efficiency of substation equipment, based on the digital twin theory, a digital model and system based on the operation and and maintenance of real substation equipment is constructed. Firstly, the digital twins, which can reflect the real state of three kinds of equipment, namely converter transformer, condenser and GIS, are set up on the information layer. Secondly, the substation equipment of digital twins is analyzed with the historical big data of converter transformer, condenser and GIS, and then the next state of the substation state is predicted according to the collected state data, operation and inspection data of substation for realizing the data fusion of the physical layer and information layer in the actual substation of substation equipment. Lastly, the substation equipment operation and maintenance are considered as the experimental objects, the embedded information and physical fusion system is used to integrate and synchronize the operation and inspection data of substation equipment, forming the final digital twin framework system of substation equipment operation and inspection. Research shows that digital twin technology can improve the intelligent degree of substation equipment operation and inspection, and provide theoretical support for future substation construction.

Phase Partition of Batch Process Based on Instantaneous Frequency Response Function

LI Yu-bin, YU Tao

2022, 0(02): 65-69.

Asbtract ( 110 )

PDF (1093KB) ( 64 )

References | Related Articles | Metrics

Multi-phases characteristic is one of the essential characteristics of batch process, and the effective phase partition of batch process is the basis of fault monitoring. Most of the traditional phase partition methods focus on the input and output data of the process, which is sensitive to the input and output data mutation. In this paper, a phase partition method of batch process based on instantaneous frequency response function is proposed. This method uses instantaneous frequency response function to replace the input and output data for phase partition based on the instantaneous dynamic characteristics of the system. Wavelet transform is used to estimate the instantaneous frequency response function of the system and the kernel principal component analysis is used to reduce the dimension. The frequency response function is clustered by fuzzy C-means clustering to partition the phase. Experimental results show that the proposed method can realize the phase partition of batch process and has high robustness.

Review of Crowdsensing of Pollution Data Collection Technology

CAO Sheng-guo, WANG Sheng-ao, GENG Shu-meng

2022, 0(02): 70-78.

Asbtract ( 210 )

PDF (1383KB) ( 154 )

References | Related Articles | Metrics

Traditional pollution monitoring methods are mainly carried out by fixed base stations, but these are inflexibility and expensive, which are no longer sufficient to deal with the problems of growing pollution. A new data acquisition mode—Mobile Crowdsensing—provides a new idea for large-scale data perception. In order to grasp the research status of pollution data collection of Mobile Crowdsensing in time, the existing researches at home and abroad are systematically and comprehensively reviewed, and feasible schemes are provided for the application of crowdsensing in smart phones based on the existing research. Firstly, the development stages of pollution collecting technology is elaborated. Then, the advantages and disadvantages of different crowd-sourced pollution collection systems are compared and analyzed, and the advantages and disadvantages of the key technologies used and the applicable scenarios are explained. Finally, the problems existing in the collection of pollution data under crowdsensing are summarized, and the future research focus is prospected.

ElasticSearch Index Optimization Strategy for Engineering Data Retrieval

XU Xian-hui, WANG Shu-ying, ZENG Wen-qu

2022, 0(02): 79-84.

Asbtract ( 319 )

PDF (1114KB) ( 201 )

References | Related Articles | Metrics

With the development of manufacturing industry, various industries generate a large amount of engineering data during the manufacturing process, the data retrieval requirements of the modern engineering field requires that the corresponding results can be retrieved quickly and accurately through keywords. The retrieval of engineering data can be achieved by using ElasticSearch, but there is still space for optimization in terms of its performance. In order to solve this problem, based on the in-depth study of the underlying theory of ElasticSearch, the index creation, index fragmentation and index segment merging of ElasticSearch are optimized. Firstly, the ElasticSearch tokenizer is modified and a custom dictionary is configured. Secondly, an index sharding strategy based on the performance of the cluster node and the size of the index data is proposed. Finally, the timing of index segment merging based on node performance is optimized. Through the experiments based on the retrieval of subway engineering data, the experimental results show that the improvement method can indeed improve the data writing and query performance of ElasticSearch.

Fast Dimensionality Reduction Sorting Search Method Based on Feature Matching

XU Hui, TIE Zhi-xin, SHU Ying

2022, 0(02): 85-91.

Asbtract ( 108 )

PDF (1297KB) ( 73 )

References | Related Articles | Metrics

In the era of big data, more and more users or enterprises upload a large amount of data to cloud storage so as to reduce the pressure of local storage and obtain efficient management of data sharing services. As a result, searchable encryption technology arises at the right moment. Retrievable efficiency and guarantee of data security have always been the focus of research. Therefore, a fast dimensionality reduction sorting search method based on feature matching (DRFM) is proposed. Through the proposed feature score algorithm, the index feature vector of each document is created; through the proposed matching score algorithm, the query matching vector of query keywords is created. The K-L transform algorithm is used to reduce the dimensions of all document index feature vectors and query matching vectors to improve the efficiency of the algorithm. Theoretical analysis and experimental results show that the proposed scheme is efficient and feasible.

Short Text Classification Method Based on Support Vector Machine

ZHAO Yan-ping, WANG Fang, XIA Yang

2022, 0(02): 92-96.

Asbtract ( 167 )

PDF (681KB) ( 107 )

References | Related Articles | Metrics

This paper proposes an effective short text classification method based on support vector machine for short texts with sparse features, non-standard features and unclear topics. Due to the low accuracy and time efficiency of Chinese dependency grammar analysis, in view of the characteristics of client text consultation, this paper did not analyze the dependency grammar of sentences, but mainly uses syntactic features for analysis. Two syntactic features of substrings and subsequences of sentences are find out. Then three feature measure methods such as information gain, mutual information and chi-square statistics are used to realize feature selection effectively. Finally support vector machine method is used to classify text. The model proposed in this paper is applied into a set of real data, and the experimental results show that the average accuracy could reach 84.19%, thus verifying the robustness and effectiveness of the classification method.

Multi-view Point Cloud Registration Technology Based on K-means++

2022, 0(02): 97-101.

Asbtract ( 177 )

PDF (1725KB) ( 107 )

References | Related Articles | Metrics

A multi-view point cloud registration method based on K-means+〖KG-*3〗+ is proposed for the possibility of noise, outliers and occlusion in large scale point sets. Firstly, the random seeding technique of K-means+〖KG-*3〗+ algorithm is used to select the initialized center of mass from the subsampled multi-view point sets, and the clustering is completed according to the basic principle of the algorithm. Secondly, the point cloud data are stored in the K-D tree structure, and the nearest neighbor search algorithm is used to establish the corresponding relationship between the point sets, so as to improve the search efficiency of the corresponding point sets. Finally, the rigid transformation parameters between the point cloud data obtained by the clustering of each view and all views are calculated according to the scanning sequence by the iterative closest point algorithm, and the errors caused by pairwise registration are evenly spread to each view until the final registration result is obtained. Experiments on Stanford 3D point cloud datasets show that the proposed method has higher registration accuracy and robustness than partial multi-view registration algorithms in recent years.

Remote Sensing Image Classification Based on Advanced Capsule Neural Network

2022, 0(02): 102-107.

Asbtract ( 125 )

PDF (2127KB) ( 98 )

References | Related Articles | Metrics

Aiming at the problems of feature information loss and poor generalization ability caused by convolutional neural network (CNN) classification of remote sensing image, an improved capsule neural network classification model based on channel attention and mixed attention is proposed. Firstly, in order for the capsule neural network to adapt to large-size input images, the two maximum pooling layers are used in the feature extraction module. Secondly, in order to improve the classification accuracy, the SENet attention and CBAM attention are added to the last layer of the feature extraction module for improving the feature extraction module. Finally, the sample set is randomly divided into training set, verification set and test set, and the training set and veritication set are further used to train the model, the test set to test the model, and the AID data set is used to verity the generalization ability of model classification. The experimental results show that the accuracy and Kappa coefficient of the improved capsule neural network based on the SENet network are higher than other models, and the generalization ability is also. The overall classification accuracy and generalization ability of the proposed model are significant improved, thus verifying the feasibility and usability of the method.

Stencil Character Recognition of Paper Medicine Packaging Based on Mask-RCNN

WU Biao, ZHOU Qing-hua, ZENG Xiao-wei

2022, 0(02): 108-113.

Asbtract ( 197 )

PDF (2568KB) ( 280 )

References | Related Articles | Metrics

In order to realize the real-time detection of stencil characters on paper medical packaging, a stencil character recognition system based on image processing and deep learning is designed. The system first uses a variety of image processing methods to preprocess the image under the original lighting, thereby automatically extracting the region of interest in the image, and inputting it into the trained Mask-RCNN network for instance segmentation, then the pixel positions of different characters and their character values in each picture are obtained. The experimental results show that, compared with the traditional character recognition method, this method can solve the problem of insignificant gray-scale jumps in the stencil character pictures of paper medical packaging well, and accurately segment and mark the stencil characters in the picture of the paper packaging box. It has high practical value and its character recognition accuracy rate reaches 99%, which provides a new solution for the recognition and recording of stamped characters on the production line.

Detection of R Wave Based on Hilbert Transform and Adaptive Threshold

GUO Tian-yu, YAN Rong-guo, FANG Xu-chen, XU Yu-ling, TAO Zheng-yi

2022, 0(02): 114-119.

Asbtract ( 207 )

PDF (1348KB) ( 148 )

References | Related Articles | Metrics

This paper presents a R wave detection algorithm based on Hilbert transform and adaptive double threshold. Firstly, amplitude normalization and Hilbert envelopment analysis are performed on the pre-processed signal. Then, the R wave is detected by the adaptive double threshold methods. Finally, the location of the detected R wave is located according to the enhanced signal. 4 kinds of databases with different frequencies and signal noise ratio, like MIT-BIH Arrhythmia, QT, NST, European ST-T, and clinical collection of ECG data are used to evaluate the performance of the proposed algorithm. The Results show that the location of R wave in various irregular ECG signals with serious noise interference can still be accurately detected by the proposed algorithm. It has the sensitivity, positive and accuracy of the overall data detection reached 99.36%, 99.77% and 99.13% in the MIT-BIH arrhythmia database, and compared with the traditional Pan and Tompkins algorithms, the average consumption time of each record is greatly reduced, which proves that the proposed algorithm has good robustness and real-time performance.

Automatic Sleep Staging Based on 3CNN-BiGRU

TANG Jie, WEN Yuan-mei

2022, 0(02): 120-126.

Asbtract ( 156 )

PDF (1290KB) ( 222 )

References | Related Articles | Metrics

Aiming at the efficiency and accuracy of single-channel EEG signal sleep automatic staging, this paper proposes to use three-scale parallel Convolutional Neural Networks to extract sleep signal features and Bidirectional Gated Recurrent Unit 3CNN-BiGRU automatic sleep staging model to learn the internal time relationship between sleep stages. First, the model performs band-pass filtering on the original single-channel EEG signal, and uses the synthetic minority oversampling technique for class balance, and then sends it to the built model for training and verification experiments. Pre-training and fine-tuning training are used for optimizing the model, and 10-folds and 20-folds cross-validation is uses to improve training reliability. The experimental results of different models under different data sets show that the 3CNN-BiGRU model has achieved better training efficiency and better staging accuracy.

Table of Content