Computer and Modernization

Multi-core Parallelism Packet Classification Algorithm Based on Dimensions Decomposition

TANG Zhi-bin1,2， ZENG Xue-wen1,2， CHEN Xiao1,2

2020, 0(02): 1. doi:10.3969/j.issn.1006-2475.2020.02.001

Asbtract ( 210 )

PDF (1050KB) ( 132 )

References | Related Articles | Metrics

In order to achieve high-speed packet classification, a multi-core parallelism packet classification algorithm is proposed in this paper. Based on the idea of dimension decomposition and Bit Vector (BV), the algorithm parallelizes the classification process with multiple cores using the intra-packet scheme, which shortens the processing time of single packet, improves the system throughput, and ensures that the output sequence is consistent with the input sequence. The experimental results show that the parallelism algorithm can achieve 92700 rules per second in terms of preprocessing speed and 5.37 Mpps throughput on the Cavium OCTEON CN6645 multi-core network processor platform. When the packet size is larger than or equal to 256 Byte, the parallelism algorithm can achieve 10 Gbps line speed processing, and its performance is higher than that of HiCut and PCIU algorithm under the same conditions.

A Minimum Cost Maximum Flow Algorithm Based on #br# Improved Heap Optimization Dijkstra Algorithm

DENG Guo-qiang, HAN Ying-zheng

2020, 0(02): 8. doi:10.3969/j.issn.1006-2475.2020.02.002

Asbtract ( 331 )

PDF (533KB) ( 175 )

References | Related Articles | Metrics

The minimum cost path for searching for sinks in the surviving network by the shortest path algorithm is the main way to solve the minimum cost maximum flow in the flow network, and the Dijkstra algorithm is one of the most efficient shortest path algorithms. In this paper, by demonstrating that there is no negative power circle in the surviving network, the improved heap optimization Dijkstra algorithm is used to search the remaining network for the minimum cost path to improve the efficiency of the algorithm. The experimental results show that compared with the classical minimum cost maximum flow algorithm based on the shortest path fast algorithm and the minimum cost maximum flow algorithm based on Bellman-Ford algorithm, the improved algorithm proposed in this paper has higher time efficiency.

Predict Model for Foreign Exchange Based on Grey-Markov

WEI Qing-zheng， YANG Yun， LI Ling-yan， WEI Hai-zhou

2020, 0(02): 12. doi:10.3969/j.issn.1006-2475.2020.02.003

Asbtract ( 239 )

PDF (473KB) ( 108 )

References | Related Articles | Metrics

With the development of the economy and the improvement of the income level of residents, exchange of foreign exchange has gradually become a daily demand for people. However, the state has strict management for foreign exchange and limits for personal exchange quotas. When the personal foreign exchange expenditure exceeds the prescribed amount, it is considered to be suspected of splitting foreign exchange.In order to better grasp the suspected foreign exchange split in the future, we need more accurate forecasts.The historical data is used by the grey prediction model, and the predicted value of the combined prediction model is obtained by combining the Markov prediction model. The research and application of this paper show that the grey Markov combination forecasting model has higher precision than the single forecasting model and can predict the future data more accurately.

A Cyclic Optimizing RRT Algorithm for UAV Path Planning

XIAO Zhi-cai， YIN Gao-yang， YAN Shi

2020, 0(02): 16. doi:10.3969/j.issn.1006-2475.2020.02.004

Asbtract ( 276 )

PDF (980KB) ( 155 )

References | Related Articles | Metrics

To solve the optimizing problem of rapidly-exploring random tree algorithm (RRT), a cyclic optimizing RRT algorithm is proposed. By introducing path length constraint into RRT algorithm as heuristic condition, useless points in search space are exterminated, near optimal path can be obtained. By introducing path length constraint of planned feasible path into next algorithm execution as heuristic condition, useless points in search space are exterminated through cyclic iterative strategy， path length is shortened after each operation. The near optimal path will be found after multiple operations, then the problem that the basic RRT algorithm for UAV path planning is overcome, path length constraint as heuristic condition sufficiency is used， meanwhile, a series of feasible alternative path are obtained, they can be quickly chosen through cooperative arrival time in coordination mission. Simulation results demonstrated that this proposed method can fast get safe and near optimal path that takes into account the dynamic constraints of UAV.

A Commercial Site Recommendation Algorithm Based on Taxi GPS Trajectory and POI Data

JIA Chong, FENG Hui-fang, YANG Zhen-juan

2020, 0(02): 21. doi:10.3969/j.issn.1006-2475.2020.02.005

Asbtract ( 285 )

PDF (1007KB) ( 149 )

References | Related Articles | Metrics

Focused on the issue of commercial site selection, a commercial site recommendation algorithm based on taxi GPS trajectory and POI of urban is proposed. Firstly, the taxi GPS trajectory and POI data of the city are preprocessed and map matching is performed. Secondly, we split the city into traffic zones and analyze the traffic flow characteristics between traffic zones by using OD matrix. Combining the POI distribution characteristics and semantic attributes in the traffic zones, a commercial site recommendation model based on the OD matrix and the POI data is constructed. Finally, the effectiveness and practicability of the proposed algorithm are verified by taxi GPS trajectory and POI data of Lanzhou. The recommendation results are visualized at the traffic zones level. The experimental results show that this recommendation algorithm can not only recommend reasonable commercial site selection, but also provide an immediately visual quantitative analysis in decision making for commercial site selection. At the same time, it can provide decision-making basis for urban public service facilities spatial layout planning.

Loan Risk Prediction Method Based on SMOTE and XGBoost

LIU Bin, CHEN Kai

2020, 0(02): 26. doi:10.3969/j.issn.1006-2475.2020.02.006

Asbtract ( 414 )

PDF (769KB) ( 374 )

References | Related Articles | Metrics

In recent years, the rapid development of online credit loan results in both continuous growth of total amount of loan and the continuous rise of probability of default. Therefore, it is of great practical significance for online credit enterprises to prevent the risk of Internet finance by studying the risk of loan. Aiming at loan-related problems including the non-balanced distribution, a large number of noise, and high dimension, a loan risk prediction method based on SMOTE and XGBoost is proposed. Through the feature engineering, the dimensionality reduction and denoising of the data set are realized. For the non-equilibrium problem of the data, the SMOTE algorithm is used to oversample the number of positive and negative samples. Based on above-mentioned work, this paper builds an XGBoost classification model, compares it with some traditional classification algorithms, and conducts comparison of validity of the prediction results under different positive and negative sample proportions. The experiment shows that XGBoost algorithm has better effect in loan risk prediction model in comparison with traditional classification models, and the increase of the proportion of minority samples through the use of SMOTE algorithm can improve the effectiveness of prediction results.

Ant Colony Algorithm Based on Clustering Integration for Solving Large-scale TSP Problems

YE Jia-qi1, FU Qiang1,2, HE Yi-jia1, YE Hao1

2020, 0(02): 31. doi:10.3969/j.issn.1006-2475.2020.02.007

Asbtract ( 380 )

PDF (970KB) ( 213 )

References | Related Articles | Metrics

Ant Colony Algorithm (ACA) is a Travelling Salesman Problem (TSP) to effectively solve the combination optimization. However, with the increased scale of TSP, traditional ACA has failed to effectively solve a large-scale TSP. The paper proposes a solving method based on Improved AP Ant Colony Algorithm (IAPACA) for large-scale TSP. With the AP clustering, the TSP is divided into sub-problems, for which the optimal solution is sought. Then the consequence of the problem is acquired through combination of the sub-problems with improved scheme. Finally an experiment simulation for test calculating example from TSPLIB standard library is conducted. The experimental results show that IAPACA has better effect than that of the traditional ACA.

Data Consistency Mechanism for Authoritative DNS

WANG Qian， YAN Xia-li， YE Jue-yu， ZHANG Hai-kuo， LI Zhen-hui

2020, 0(02): 36. doi:10.3969/j.issn.1006-2475.2020.02.008

Asbtract ( 278 )

PDF (732KB) ( 190 )

References | Related Articles | Metrics

As a distributed system, authoritative DNS service adopts multi-copy data storage and multi-node service mode, which puts forward requirements on the consistency of resolution data. Byzantine fault tolerance becomes the key problem of authoritative DNS. According to the characteristics of authoritative system, this paper presents a DNS data consistency mechanism. Based on PBFT algorithm and self-designed data consistency checksum, in the process of node data synchronization, the data negotiation scheme is implemented to eliminate the influence of Byzantine node, and ensures the analytic node to obtain consistent and reliable data. The results of consistency analysis of analytic data prove that in the untrustworthy environment, the guarantee mechanism can effectively reduce the probability of inconsistency of resolution data, and enhance the reliability of authoritative DNS.

Research on Beidou Secure Transport Protocol Based on SM9 Identity Password

WU Ke-he， CHEN Hong-xiang， LI Wei

2020, 0(02): 41. doi:10.3969/j.issn.1006-2475.2020.02.009

Asbtract ( 173 )

PDF (733KB) ( 97 )

References | Related Articles | Metrics

With the wider coverage and higher positioning accuracy of Beidou satellite, Beidou communication is widely used in the field of positioning and navigation communication. In view of Beidou’s transmission security and considering Beidou’s transmission characteristics, a secure transmission protocol based on SM9 identifier password is designed, which reduces the frequency of Beidou’s transmission, improves the speed of negotiation completion, and ensures data integrity, security and reliable transmission.

Research and Implementation of Security Access System for #br# Video Terminal Based on Business Secret Algorithm

TANG Zi-zhuo1， WU Ke-he1， LI Wei1， ZHANG Xian-kang2， CUI A-jun2

2020, 0(02): 46. doi:10.3969/j.issn.1006-2475.2020.02.010

Asbtract ( 208 )

PDF (1399KB) ( 119 )

References | Related Articles | Metrics

With the continuous development of power network video surveillance system, the security problem of video terminal access to the surveillance system is also increasingly prominent. Information leakage, data tampering and malicious attacks need to be solved urgently.Aiming at the application status and security risks of power network video surveillance system, combined with the requirements of power security protection, this paper designs a video terminal security access system based on business secret algorithm, which can realize the security access of power video terminal. Finally, the system can satisfy the security and protection requirements of power video surveillance through experimental test and analysis.

Hamming Neural Network Model for Tetris Game

LIU Chang-ping1， LIU Hai1， XIA Meng1， YIN Guang-cai2

2020, 0(02): 51. doi:10.3969/j.issn.1006-2475.2020.02.011

Asbtract ( 332 )

PDF (738KB) ( 155 )

References | Related Articles | Metrics

Artificial intelligence has been widely applied to computer games. A hamming neural network games model was proposed for Tetris game. In this model, a pattern matrix was created according to blocks of different position on a given blocks map. 〖JP3〗The falling block acted as a waiting match block. Pattern match was performed in the feedforward layer of hamming neural network to obtain the hamming distance among patterns. The optimistic position and shape were achieved in the recursive layer of hamming neural network. This module and its game were realized using MATLAB to cumulate automatically blocks. Compared with other evaluation function and algorithms, this model has its own generic characters and can be adapted to modeling of other computer games.

A Fusion Social Network Recommendation Model Based on User Behavior Mining

ZHANG Chuang-ji

2020, 0(02): 55. doi:10.3969/j.issn.1006-2475.2020.02.012

Asbtract ( 231 )

PDF (684KB) ( 291 )

References | Related Articles | Metrics

Big data processing technology and parallel computing method are used to mine the user behavior characteristics of social network, and the intelligent recommendation of social network is realized. A fusion social network recommendation model based on user behavior mining is proposed. The association rule distribution model is used to detect the user behavior characteristics of the fused social network, and the ontology information and association rules items of the user behavior of the fused social network are extracted, and the fuzzy decision model of the joint recommendation of social network is constructed. The joint information entropy eigenvalue of user behavior is calculated, and the fuzzy C-means clustering method is used to classify and recognize the extracted features. User behavior mining and adaptive recommendation based on social network fusion are realized according to classification and recognition results. The simulation results show that the proposed method has a high precision and a high confidence level.

Knowledge Portrait Model for Research Project Documents

WU Di， AI Zhong-liang， LIU Zhong-lin， LI Chang-bao

2020, 0(02): 60. doi:10.3969/j.issn.1006-2475.2020.02.013

Asbtract ( 228 )

PDF (790KB) ( 132 )

References | Related Articles | Metrics

Aiming to improve the ability of accurate and intelligent identification of the knowledge points of the research project documents produced by research activities, we analyzed the document structure and proposed a method for establishing document knowledge portrait. A multi-level knowledge portrait fitting the structure of documents is designed. It identified the key points of the document knowledge automatically and extracted the knowledge points according to the multi-granularity of the semantic paragraph. The accuracy of knowledge extraction is used to test the accuracy of the model. The experimental results show that the model is more accurate than the traditional method for document knowledge description and can be used in practical application.

Prediction of Polyproline Type II Secondary Structure Based on Convolutional Neural Network

LIU Yang， MENG Ai

2020, 0(02): 65. doi:10.3969/j.issn.1006-2475.2020.02.014

Asbtract ( 292 )

PDF (740KB) ( 217 )

References | Related Articles | Metrics

Polyproline type II helix is a special and rare protein secondary structure. In order to save the time and cost of determine the structure by experimental method, a deep learning algorithm based on convolution neural network is designed to predict polyproline type II helix. First of all, the protein sequence information feature is encoded to generate feature matrix, which includes amino acid orthogonal code, physical and chemical properties of amino acids and position-specific scoring matrix. Secondly, the normalized feature matrix is inputted into convolution neural network to automatically extract the local deep features of protein sequence and output the prediction results of polyproline type II helix. The experimental results show that the performance of this algorithm is better than six traditional machine learning algorithms such as support vector machine.

Research Progress and Application of Behavior Tree Technology

LIU Rui-feng， WANG Jia-sheng， ZHANG Hao-long， TIAN Meng-fan

2020, 0(02): 76. doi:10.3969/j.issn.1006-2475.2020.02.016

Asbtract ( 452 )

PDF (1147KB) ( 289 )

References | Related Articles | Metrics

Artificial intelligence will not only change the production and life of human beings, but also fundamentally change the winning mechanism and fighting mode of modern wars, and give birth to new fighting means and ideas. Behavior tree will be a good medium for the application of artificial intelligence in the military field. This paper first introduces the basic principle of Behavior Tree (BT), and then elaborates and analyzes the development status of behavior tree from the aspects of implementation and design application. Secondly, the characteristics of behavior tree, finite state machine and hierarchical finite state machine are compared from the aspects of hierarchy, maintainability, code coupling, expansibility and reusability. Finally, the development trend of behavior tree combined with machine learning and expert system, and its application prospects in interactive software platform, unmanned autonomous systems and combat simulation systems are analyzed.

Remaining Useful Life Prediction of Lithium-ion Batteries #br# Based on VMD and GPR Algorithm

WU Yi, WANG You-ren

2020, 0(02): 83. doi:10.3969/j.issn.1006-2475.2020.02.017

Asbtract ( 273 )

PDF (940KB) ( 168 )

References | Related Articles | Metrics

The lithium-ion batteries often suffer from sudden and occasional capacity regeneration due to the complex and discontinued working conditions. Thus, the capacity degradation data shows the nonlinear and nonstationary trend, which makes it difficult to achieve high prediction accuracy. Therefore, a remaining useful life(RUL) prediction method of lithium-ion batteries based on variational mode decomposition (VMD) and Gaussian process regression(GPR) is proposed to treat this problem. Firstly, VMD is employed to decompose the capacity degradation data and to extract the global degradation, local regeneration, and random fluctuation components. Then, different GPR prediction models are built for each component by choosing suitable kernel functions. Lastly, the predicted components are superimposed to obtain the capacity prediction result and the predicted RUL. The proposed method is validated through a case study using NASA dataset. Results show that the proposed method outperforms the GPR models without VMD decomposing.

A Modified Object Detection Algorithm Based on SSD

SU Meng, LI Wei

2020, 0(02): 89. doi:10.3969/j.issn.1006-2475.2020.02.018

Asbtract ( 212 )

PDF (770KB) ( 181 )

References | Related Articles | Metrics

Single Shot MultiBox Detector(SSD), one of the most prevalent object detection algorithms based on deep learning, greatly shortens the time of object detecting, whose accuracy of detection is acceptable. However its accuracy cannot meet the need of practical application. Attention mechanism is helpful in improving the performance of convolutional neural network and the accuracy of detection. This paper modifies SSD with attention mechanism and proposes a modified object detection algorithm based on SSD. The modified SSD has been tested with Pascal VOC dataset. The modified SSD achieves 78.5% mAP in Pascal VOC2007 test set and 77.1% mAP in Pascal VOC2012 test set, which are higher than original SSD 4.2 percentage points and 4.7 percentage points separately.

A Lightweight Target Detection Network Based on Channel Rearrangement

XU Han-zhi， AI Zhong-liang， ZHANG Zhi-chao

2020, 0(02): 94. doi:10.3969/j.issn.1006-2475.2020.02.019

Asbtract ( 214 )

PDF (714KB) ( 157 )

References | Related Articles | Metrics

Tiny YOLO and YOLOv3-tiny are two lightweight target detection algorithms known for their outstanding speed performance. Based on these two network models, combining packet convolution and improved channel rearrangement algorithm and the original loss function， this paper constructs a new faster network model which improves the detection accuracy by improving the loss function of YOLOv3. The PASCAL VOC and COCO datasets were trained and tested respectively. The speed of the network model was faster than 265 pictures per second, and the accuracy was higher than Tiny YOLO and similar to YOLOv3-tiny.

Rubber Plunger Defect Detection Method #br# Based on Super Pixel Segmentation and Random Forest

SUN Shi-fan， YE Ming， LIU Kai

2020, 0(02): 99. doi:10.3969/j.issn.1006-2475.2020.02.020

Asbtract ( 306 )

PDF (2939KB) ( 139 )

References | Related Articles | Metrics

For the small-sized rubber plunger end face with a diameter of 3 mm, it was difficult to segment the defect contour by the interference of light spots, dust and texture, So a defect detection system combining SLIC(simple linear iterative clustering) and RF(random forest) algorithms was proposed. Firstly, Hough transform and anisotropic filtering were used as image preprocessing. Then SLIC algorithm based on super pixel segmentation was used to segment and extract defect regions. Finally, the five-dimensional shape feature of the obtained defect area was used as the RF classifier feature vector for defect classification prediction. The results show that the SLIC algorithm is 0.128 s faster than the traditional adaptive threshold segmentation algorithm, and the segmentation effect is much better than the traditional algorithm, the defects as small as 0.5 mm can be accurately segmented, the overall inspection process takes less than 1.5 s on average. At the same time, the accuracy rate of RF classification is 97.3%. Therefore, the defect detection system of this paper meets the requirements of accuracy and real-time of online detection, which can be used in practical work.

Face Verification and Application Research of ID Photos Based on DCNN

LI Shuo1, BIAN Qing-shan2, LIU Chuan-wen2, LIU Ming-tao1, ZHANG Lin-tao1

2020, 0(02): 104. doi:10.3969/j.issn.1006-2475.2020.02.021

Asbtract ( 170 )

PDF (1285KB) ( 114 )

References | Related Articles | Metrics

In different authentication scenarios, it is difficult to adapt the existing methods to face recognition under different authentication photos for the sake of the influence of age span, dress-up and lack of samples, which cannot conform to the practical application requirements. For the sake of solving the above problems, it puts forward a different identification method on the basis of the deep convolution neural network. This method makes the improvement of VGG network adapted to different document photo recognition, realizing end-to-end autonomous learning of face features, eliminating the influence of age span, dress-up and other factors. In addition, the method cuts down the trainable parameters to 1/6 of the original network structure, thus ensuring the identification accuracy while greatly reducing the training time of the model. According to the experimental results, after training on the self-built data set and CAS-PEAL-R1 public data set under the college graduation examination scene, the verification accuracy and recall rate of this method were 6.29 and 7 percentage points higher than the original method respectively, which can conform to the different document examination needs under various application scenarios.

Research on Bird Detection Algorithm for Transmission #br# Lines Applicable to Mobile Terminal

CUI Wen-chao, LI Yuan-bo, WANG Min-jian

2020, 0(02): 110. doi:10.3969/j.issn.1006-2475.2020.02.022

Asbtract ( 235 )

PDF (914KB) ( 189 )

References | Related Articles | Metrics

Transmission line safety is the premise of safe and stable operation of the power grid, but frequent bird activities have seriously affected the transmission line. In order to solve the drawbacks of the traditional bird-repelling method, the researchers use deep learning algorithms for bird detection. However, deep learning algorithm needs to run on a server with good performance, which will inevitably cause network delay and cannot be used to drive birds in real time. Therefore, bird detection should be carried out at the mobile terminal, but the existing target detection algorithm model is large and cannot be directly applied to the mobile terminal. Therefore, this paper proposes a bird detection algorithm for transmission line suitable for mobile terminal, which will be in the YOLO v3 model. This algorithm replaces the basic network darknet-53 in YOLO V3 model with the lightweight feature extraction network MobileNet, which achieves bird detection of transmission lines at mobile terminal. The experimental results show that the accuracy of the model can reach 83.57% and the detection speed reaches 61 fps in the bird detection task of the transmission line. It can be stably operated on the mobile terminal platform of 4 GB memory, which can meet the accuracy and real-time requirements of the bird detection task of transmission line and have a good application prospect.

Research on Intelligent Detection Technology for Illegal Wearing #br# in Power Operation by Introducing Self-Attention

MO Bei-bei, WU Ke-he

2020, 0(02): 115. doi:10.3969/j.issn.1006-2475.2020.02.023

Asbtract ( 263 )

PDF (4039KB) ( 176 )

References | Related Articles | Metrics

With the rapid development of power grid construction, the scale of technical support personnel in operation site is expanding continuously. Operation site belongs to high-risk work site, illegal wearing protective equipment will seriously endanger the workers. In order to improve the inefficiency of traditional manual supervision, this paper uses a real-time deep learning algorithm to detect illegal wearing behavior. The algorithm combines the real-time object detection network YOLOv3 and Self-Attention mechanism, uses the DANet structure for reference, and embeds the Self-Attention module at the high level layers of YOLOv3 network to better mine and learn location relations and channel relations of feature maps. The experimental results show that the mAP and Recall of this algorithm reached 94.58% and 96.67%. Compared with YOLOv3, its mAP and Recall increased by 12.66% and 2.69%. The accuracy of the model is significantly improved, which can meet the detection requirements of the task and improve the intelligence of power grid.

Research on Dangerous Goods Detection Algorithms Based on Improved Adaboost and LBP

NIU Dao-hong1， MA Xiao-dong1, WU Xue-bing1, WANG Fang1,2

2020, 0(02): 122. doi:10.3969/j.issn.1006-2475.2020.02.024

Asbtract ( 176 )

PDF (1601KB) ( 123 )

References | Related Articles | Metrics

An improved algorithm of dangerous goods detection based on Adaboost and LBP is proposed, which can improve the accuracy and speed of identification. It solves the problem that the detection accuracy of dangerous goods decreases due to the influence of brightness, illumination and other interference factors. The improved algorithm incorporates HSV color space classification of positive samples in the training stage, which improves the detection efficiency of cascade classifiers, and extracts eigenvalues with the improved LBP algorithm. Compared with traditional object detection methods, the accuracy is improved by 2 percent point to 93.29%. Finally, the algorithm is transplanted to the rescue manipulator platform. The experimental results show that the improved detection algorithm can identify dangerous goods accurately and quickly in the actual environment detection, and the training efficiency is obvious. At the same time, it has good robustness under different illumination conditions and meets the practical requirements.

Table of Content