Computer and Modernization

A Multi-objective Flower Pollination Algorithm Based on Decomposition

CHEN Min-rong, HUANG Guang-jing

2019, 0(07): 1. doi:10.3969/j.issn.1006-2475.2019.07.001

Asbtract ( 202 )

PDF (1475KB) ( 137 )

References | Related Articles | Metrics

During the past decades, a variety of multi-objective evolutionary algorithms (MOEAs) have been widely used to solve all kinds of multi-objective optimization problems (MOPs). One of the representative MOEAs is the multi-objective evolution algorithm based on decomposition, called MOEA/D. The flower pollination algorithm (FPA) is a meta-heuristic optimization algorithm. However, to our best knowledge, so far there are few papers studying on FPA based on decomposition in the multi-objective optimization field. In this paper, under the framework of MOEA/D, we extend the initial FPA to decomposed-based multi-objective optimization field and further present a multi-objective FPA based on decomposition, called MOFPA/D. In addition, in order to keep the diversity of nondominated solutions in the external archive, we address a novel strategy, named grid-based segmentation of objective space to select some Pareto optimal solutions for output. The simulation results indicate that MOFPA/D is highly competitive with or superior to the initial MOEA/D in terms of solution convergence and diversity.

Machine Translation System Based on Self-Attention Model

SHI Yan, WANG Yu, WU Shui-qing

2019, 0(07): 9. doi:10.3969/j.issn.1006-2475.2019.07.002

Asbtract ( 257 )

PDF (948KB) ( 125 )

References | Related Articles | Metrics

In recent years, neural machine translation (NMT) has developed rapidly. The proposed Seq2Seq framework brings great advantages to machine translation. It can generate arbitrary output sequences after observing the entire input sentence. However, this model still has great limitations on the ability to capture long-distance information. The proposed recurrent neural network (RNN) and LSTM network were all proposed to improve this problem, but the effect is not obvious. The presentation of the attention mechanism effectively compensates for this deficiency. The Self-Attention model is proposed on the basis of attention mechanism, and an encoder-decoder framework is built based on Self-Attention. This paper explores the previous neural network translation model. The mechanism and principle of the Self-Attention model are analyzed. The translation system is realized based on Self-Attention model by TensorFlow deep learning framework. In the English-to-Chinese translation experiment, compared with the previous neural network translation model, it shows that the model has a good translation effect.

New Data Processing System Based on Edge Computing

WANG Shao-qiang, WANG Yu

2019, 0(07): 15. doi:10.3969/j.issn.1006-2475.2019.07.003

Asbtract ( 256 )

PDF (1085KB) ( 146 )

References | Related Articles | Metrics

The problems in the existing cloud computing model are analyzed, namely real-time, privacy protection, energy consumption and other aspects. Then, the concept of the edge computing model and its advantages over the cloud computing model are discussed, and the system architecture of the new data processing system is proposed. The operation and scheduling strategy of the data processing system is designed and implemented. The implemented functions are verified by specific examples. Finally, the application prospects and improvement directions of new data processing systems based on edge computing are discussed.

Heuristic Variable Selection Strategy Based on Variable Decision Level

LIU Yao, SONG Zhen-ming

2019, 0(07): 20. doi:10.3969/j.issn.1006-2475.2019.07.004

Asbtract ( 159 )

PDF (690KB) ( 96 )

References | Related Articles | Metrics

The heuristic branch strategy is an integral part of the SAT solver and directly affects the efficiency of the solver. Early heuristic branch decisions require traversal of the entire clause database, which is inefficient. With the emergence of variable state independent decaying sum (VSIDS) branch strategy, the efficiency of the SAT solver is improved, but the increments of the variables in the VSIDS strategy and its extended strategy are only related to the number of conflicts with the variable, and do not consider the influence of the decision layer of the variable in the branching strategy. Therefore, when a conflict occurs, if the scores of the variables related to the conflict are the identical but the decision levels are different, the choice of variables is random. Based on this, after expounding the importance of the decision level of variables, based on the VSIDS strategy, a heuristic variable selection strategy named HSVDL strategy based on variable decision level is proposed. Then the example shows that the HSVDL strategy is more likely to choose the variable with lower decision-making level than the variable with higher decision-making level in the variable decision-making stage, and the score is smaller, which reduces the memory occupation. Finally, experiments show that the HSVDL strategy can solve more examples, and the efficiency of the solver is also improved, indicating that the strategy has certain advantages.

A Horizontal Elastic Expansion and Contraction System Based on Kubernetes

TU Xue-zhen1, YANG Hai-chao2

2019, 0(07): 25. doi:10.3969/j.issn.1006-2475.2019.07.005

Asbtract ( 299 )

PDF (1502KB) ( 177 )

References | Related Articles | Metrics

The birth of Kubernetes has reduced application developers’ reliance on infrastructure and operations teams, it provides powerful tools for orchestrating and scheduling containers and virtual machines, which has become the de facto standard for the development and management of distributed cluster systems. Kubernetes monitors CPU and memory usage in the cluster through the core component HPA (horizontal pod autoscaler), and scales and shrinks the microservice cluster based on these metrics. However, these simple metrics can not meet the expansion and contraction requirements of the actual application. Although the improvements have been made in the latest version of the community, the actual results are still unsatisfactory. Based on the original platform, this paper designs a superimposed elastic expansion and contraction system E-HPA to make up for the defects and shortcomings of Kubernetes in horizontal elastic expansion and contraction, which can provide rich user-customizable metrics through flexible and simple configuration. The design ideas and specific implementation details of the system are expounded, and the experimental results are verified by taking the telecom business application as an example.

#br# A Moving Target Algorithm Based on Gaussian Mixture Model #br# and Eight-neighbor FDM with Morphological Processing

YANG Shu-guo1, HE Wen-jing1, LIU Yin-ling1, MA Zuo-lin1, HU Shuai2

2019, 0(07): 32. doi:10.3969/j.issn.1006-2475.2019.07.006

Asbtract ( 171 )

PDF (1412KB) ( 110 )

References | Related Articles | Metrics

For the extraction of the moving subjects in a video clip, this paper proposes a moving target detection algorithm based on improved Gaussian mixture model and eight-neighbor frame difference method. Firstly, some frames of a video are converted into grayscal images and a statistics model based on Gaussian mixture distribution is set up. Following up, the sketch of the moving part is acquired through eight-neighbor frame difference method. For a more precise subtraction, the Gaussian mixture model is added to the algorithm. Combining with morphological processing, a complete and precise foreground is extracted. Experimental results show that the proposed algorithm has better effect on videos with slow-moving objects and large shadow areas, compared to earlier algorithms, and it is robust in multiple occasions.

OLED Defective Pixel Recognition Based on K-means Clustering and LVQ Neural Network

JI Yan-ling1, LIN Zhi-xian1, TANG Qian1, GUO Tai-liang1, TANG Biao2

2019, 0(07): 37. doi:10.3969/j.issn.1006-2475.2019.07.007

Asbtract ( 251 )

PDF (1515KB) ( 134 )

References | Related Articles | Metrics

An improved K-means clustering segmentation and LVQ neural network classification method is proposed to identify defective pixels in inkjet printing process of organic light-emitting diode display panels. Firstly, the improved K-means clustering algorithm is used to segment the preprocessed printed pixels, then the connected-domain horizontal rectangle is used to determine the coordinates and geometric features of each printed pixel, and the texture features are extracted by the gray-scale co-occurrence matrix, finally, the features are classified by LVQ neural network to complete the marking and classification statistics of the defective pixels. The results show that the proposed algorithm recognition rate of the article is obviously better than other common classification recognition algorithms, the average defect detection rate of the algorithm is 100%, the classification accuracy rate is 98.9%, and the single pixel detection time is 8.3 ms.

A Cattle Face Detection Method Based on Improved NMS

GOU Xian-tai, HUANG Wei, LIU Qi-fen

2019, 0(07): 43. doi:10.3969/j.issn.1006-2475.2019.07.008

Asbtract ( 181 )

PDF (2579KB) ( 143 )

References | Related Articles | Metrics

A cattle face detection method based on improved Faster R-CNN is proposed for multi-face detection of livestock cattle identity authentication. Inception v2 is used to replace ZF network as the basic network of Faster R-CNN, which improves the model accuracy significantly. The NMS (Non-Maximum Suppression) is optimized for the multi-cattle detection scenario, so that the recall rate of the model is significantly improved. Comparing with other detection models, the improved model is superior to others in accuracy and recall.

A Survey of Dynamic Human Facial Expression Synthesis 〖JZ〗Approaches Driven by Model Features

CHEN Song, YUAN Xun-ming

2019, 0(07): 47. doi:10.3969/j.issn.1006-2475.2019.07.009

Asbtract ( 176 )

PDF (2156KB) ( 144 )

References | Related Articles | Metrics

This paper surveys the overall methodologies for generating dynamic facial expression on 2D and 3D human face models and categorizes them into several classes based on their intrinsic properties. Even though there exists a considerable body of previous works, this topic is still gaining very active research attention. According to the dimensionality of their final output, the methods are categorized into class working on 2D image plane and class working on 3D facial manifold surfaces. The general approach of synthesizing 2D human facial expression includes: Feature-points driven facial expression synthesis method, generation methods based on facial proportion map, Laplacian transfer based expression generation method, methods based on general expression mapping functions, synthesis methods based on active expression model and the latest approaches using deep learning techniques. On the aspect of synthesizing 3D facial expression, the classifications contain: 3D shape regression-based methods, synthesis methods based on deformation, approaches based on pseudo muscle models, algorithms based on a motion-vector analysis, and the latest approaches using deep learning techniques. For each category, we describe its basic methodology along with its advantages and limitations. This survey paper is expected to help researchers to better position their potential future direction in the context of existing solutions.

A SLAM Technology Combining Area Detection and Semantic Segmentation

WANG Wen, XU Yi-bai, LU Shan, FENG Yu

2019, 0(07): 55. doi:10.3969/j.issn.1006-2475.2019.07.010

Asbtract ( 361 )

PDF (1519KB) ( 115 )

References | Related Articles | Metrics

This paper proposes a real-time location and mapping (SLAM) technology that combines region detection and semantic segmentation. The precision of inter-frame pixel matching in the visual odometer (VO) process is realized by introducing high-precision image descriptor SIFT. In order to reduce the computational complexity caused by the introduction operation, we design a real-time region detection algorithm to detect the region of interest (ROI) between adjacent frames, so that the SIFT descriptors are extracted and matched only in the ROI region. At the same time, in order to improve the accuracy of the bundle adjustment (BA) and reduce the cumulative error, the paper combines the semantic information. The semantic map is implemented by a real-time semantic segmentation algorithm. Compared with the original SLAM scheme, the proposed method can improve the accuracy of SLAM and meet the requirements of real-time localization and mapping. Finally, we verify the effect of the scheme in the power operation scenario.

Face Recognition Based on Laboratory Security Service Robot

HUANG Wu-fei1, ZENG Ji-chang1, HUANG Jin-quan1, LI Si-han1, CHEN Yu-jun1, CHEN Zhao1, ZHENG Zhi-shuo1,2

2019, 0(07): 61. doi:10.3969/j.issn.1006-2475.2019.07.011

Asbtract ( 143 )

PDF (1012KB) ( 95 )

References | Related Articles | Metrics

A face recognition system has been designed for robots. At the first stage, the system captures face image through camera, and then completes the face recognition on PC. Finally, the PC sends instructions to the robot. The recognition process is to calculate the Euclidean distance of the facial 128D vector between image captured by camera and those stored in database, then to determine identity through the result of the Euclidean distance. The results show that the system has good recognition, uncomplicated structure, and can be widely applied.

Online Monitoring and Fault Diagnosis System for #br# Smart Substation Secondary Equipment in Provincial Power Grid

XIE Min1, SHAO Qing-zhu1, WANG Tong-wen1, SHAO You-shen2, WANG Hai-gang1, YU Bin1

2019, 0(07): 65. doi:10.3969/j.issn.1006-2475.2019.07.012

Asbtract ( 173 )

PDF (1694KB) ( 113 )

References | Related Articles | Metrics

To meet the demands of secondary equipment information for power grid operation and intelligent dispatching, the online monitoring and fault diagnosis system for smart substation secondary equipment is constructed in Anhui provincial power grid to realize the remote inspection and operation, state monitoring and warning, analysis of fault diagnosis and action behavior of relay protection devices. Combined with the practice and application of pilot system, this paper introduces the construction demand and realization of the hardware structure, software structure and functional module. And a look-ahead is presented for future technology development in promoting the online monitoring and fault diagnosis system. Through the construction and application of the system, it provides decision-making basis for dispatchers to deal with power grid faults quickly and effective technical support for the fault elimination and condition based maintenance of secondary equipment in intelligent substations.

Fault Classification of Process Layer Network in Intelligent Substation Based on ANP-SVM

ZHANG Kai-yan1,2, PAN Yang3, LOU Ji-chao4

2019, 0(07): 72. doi:10.3969/j.issn.1006-2475.2019.07.013

Asbtract ( 161 )

PDF (965KB) ( 153 )

References | Related Articles | Metrics

Aiming at the inefficiency and data set noise of the existing process layer network fault classification in intelligent substation, this paper proposes an ANP-SVM based process layer network fault classification algorithm. Firstly, the improved separation interval method is used to optimize the selection of kernel parameters and error parameters of SVM, and then the anti-noise sample data is input into the optimized SVM, which makes the classification more accurate and efficient. The experimental results show that the algorithm has good performance in the process layer network fault classification.

Intelligent Search Engine of Inner Mongolia Electric Power Marketing Data

LIU Yan-hang, MI Jia, HAN Xue, KONG Fan-chun

2019, 0(07): 78. doi:10.3969/j.issn.1006-2475.2019.07.014

Asbtract ( 122 )

PDF (898KB) ( 137 )

References | Related Articles | Metrics

With the construction of power grid and information technology in Inner Mongolia for many years, power marketing requires higher and higher data integration ability and information search speed. Therefore, in order to improve the data search ability of power companies, this paper puts forward an overall framework of intelligent search engine on how to search target data quickly. The search engine adopts the mode of combination of distribution and centralization in its structure. Through the improvement of database index establishment method and the improvement of index algorithm with adaptive ability evolved from genetic algorithm, the intelligent marketing data search method provided by the intelligent marketing data search engine based on large data is greatly improved. It improves the retrieval speed and accuracy of the results, and becomes an indispensable user assistant in the process of marketing business processing.

Indoor Localization Algorithm Based on Semi-supervised #br# Learning of Global Manifold Geometry

LI Shi-bao, WANG Sheng-zhi, ZHANG Xin, CHEN Hai-hua, LIU Jian-hang, HE Yi-jing

2019, 0(07): 82. doi:10.3969/j.issn.1006-2475.2019.07.015

Asbtract ( 161 )

PDF (1077KB) ( 98 )

References | Related Articles | Metrics

The construction of radio map is time consuming and labor intensive in the conventional wireless local area network (WLAN) indoor localization systems. In order to solve this problem, the paper proposes a semi-supervised manifold alignment radio map construction approach based on the global geometry of manifold structure. The proposed method utilizes a small number of labeled RSS which requires a huge time consuming to collect and plenty of unlabeled data that is easy to obtain. Then, the locations of plenty of unlabeled data can be obtained by calibrating the solution of the manifold alignment of objective function. In addition, the geodesic distance is utilized to capture the global geometry of manifold feature which can fully exploit the correspondence characteristics of the labeled RSS and its coordinates. Thus, it can improve the accuracy of radio map with limited labeled RSS data. The extensive experiments demonstrate that the proposed method can construct an accurate radio map at a low manual cost, as well as achieve a high localization accuracy.

Discovery of Hierarchy Organization Structure of Business System#br# with Hierarchical Network Structure

XIONG Yong-fang, LIN You-fang, WU Zhi-hao

2019, 0(07): 88. doi:10.3969/j.issn.1006-2475.2019.07.016

Asbtract ( 178 )

PDF (3139KB) ( 115 )

References | Related Articles | Metrics

The business system architecture diagram is one of the important tools for operations to maintain data center. Automatic combing architecture can significantly improve the maintenance efficiency. The architecture diagram of the business system is a hierarchical network structure. Therefore, for this problem of hierarchical organization structure discovery, it is not only necessary to detect the cluster-group but also to locate the hierarchical position of the cluster-group in the system architecture diagram. Because the quality of the cluster-group detection is directly affects cluster-level positioning, so accurate detection of cluster-group is very crucial. Community detection helps to reveal the relationship between individuals in a complex network structure, but the server cluster in the business system does not conform to the density-based community definition of traditional community. Therefore, this paper proposes a method of functional cluster positioning (FCP) based on the function of the server to detect the cluster-group and locate the cluster hierarchy. We construct a similarity network of servers based on server similarity of connectivity and attribute. The cluster characteristics of the network conform to the definition of traditional community, so we can detect the cluster-group by algorithms of traditional community, and then locate the functional level of the cluster-group in the business system based on the business flow so as to realize the discovery of hierarchy organization structure of the business system. The experimental results on the real data set show that the FCP method proposed in this paper can automatically and accurately discover the hierarchy organization structure of the business system in the data center.

Detecting Extreme Driving Behaviors Based on Mobile Crowdsensing

ZHAI Shu-ying1, LI Ru1, GUO Yang2

2019, 0(07): 97. doi:10.3969/j.issn.1006-2475.2019.07.017

Asbtract ( 132 )

PDF (1844KB) ( 87 )

References | Related Articles | Metrics

In order to improve driving safety and reduce traffic accidents, this paper proposes an extreme driving behavior recognition method based on mobile crowdsensing. The collected data of sensors related to users’ smart phones are preprocessed, and then the context information of passengers is identified by means of the dynamic step number detection and the random forest methods. According to different extreme driving behaviors, smart phones of passengers in different positions are selected for data collection, and the impact of different phone places and multi-feature fusion is also considered in designing the extreme driving behavior detection method. Regarding the potential inconsistency of the results from passengers in different positions, a Bayesian voting model is proposed to solve the problem. Experimental results from real world datasets indicate that our method can effectively identify the extreme driving behavior of drivers.

Cruise System for Service Robots Meeting in Same Trajectory

CHEN Hua-zhen, XIA Guo-qing, CHEN Ze-feng

2019, 0(07): 104. doi:10.3969/j.issn.1006-2475.2019.07.018

Asbtract ( 125 )

PDF (2448KB) ( 81 )

References | Related Articles | Metrics

Service-oriented cruise robots have a great influence on traffic and transportation, and the research and development of service-oriented robots is increasing. To solve the problem of self-cruise strategy of service-oriented robots, a self-cruise system of service-oriented robots meeting in same trajectory and different directions based on MK60FX512VLQ15 microcontroller is proposed. The system achieves path planning by capturing path images by camera and processing by MK60FX512VLQ15 microcontroller. The safe distance control between robots is realized by using ultrasonic ranging module. The real-time communication between robots is realized by Bluetooth 2.0. At the same time, a host computer software is developed by using the QT Creator software to realize the image processing and analysis of the robot and remote online debugging parameters. The experimental results show that the orbital-seeking service of the system can effectively realize the intersection of robots in a given path under complex scenarios.

Medical Big Data and Privacy Leakage

SHANG Jing-wei1,2, JIANG Rong1,2, HU Xiao-han1,2, SHI Ming-yue1,2

2019, 0(07): 111. doi:10.3969/j.issn.1006-2475.2019.07.019

Asbtract ( 223 )

PDF (840KB) ( 181 )

References | Related Articles | Metrics

Big data, especially in the field of medical big data, is directly related to the life and health of human beings. With the development of big data and the acceleration of informatization, the rapid popularization of medical and health information platform and digital medical equipment and instruments has led to the explosive growth of data in the medical field, with various types and complicated relationships. Sensitive medical data security is also a matter of great concern. While medical data provide help for human health, the protection of related sensitive data has become a hot topic for scholars, practitioners and the general public. This article starts with big data’s basic concept, through the analysis to the current stage privacy leakage and the medical big data related research, unifies the big data domain correlation research to the current privacy leakage behavior, the protection technology and so on question carries on the classification elaboration. It is hoped that it will be helpful to the further research of scholars in this field.

Trust-based Dynamic Multi-level Access Control Model

ZHANG Peng, ZHOU Liang

2019, 0(07): 116. doi:10.3969/j.issn.1006-2475.2019.07.020

Asbtract ( 217 )

PDF (980KB) ( 148 )

References | Related Articles | Metrics

Aiming at the problems of limited scalability, inadequate security and permission allocation in traditional role-based access control (RBAC) models, in order to improve the compatibility of access control and refine the granularity of access control, this paper proposes a trust-based dynamic multi-level access control model (TBDMACM). TBDMACM guarantees data confidentiality and security of access process by obtaining corresponding authorization through the static roles and dynamic trust of users. The experimental results show that the proposed method can effectively prevent malicious access and better solve the problem of system privilege scalability.

A Security Mechanism for Front-end Data Interaction

WU Wan-jian1, WU Xiao-yong2

2019, 0(07): 122. doi:10.3969/j.issn.1006-2475.2019.07.021

Asbtract ( 297 )

PDF (872KB) ( 102 )

References | Related Articles | Metrics

The traditional front-end integration mode has been developed into a more common front-end separation mode. Although the front-end separation frame solves the problem of strong coupling between components, there is little mention of safety-related descriptions. According to the advantages and disadvantages of the current technology, based on the characteristics of RSA and AES algorithms, we studied and designed a security mechanism for data exchange between front and rear ends. It enhanced the stability and security of data transmission through RSA and AES hybrid encryption and trans-coding. The correctness of the data was verified by using the irreversible hash algorithm. The relevant business logic of the algorithm key was extracted to form a separate key management service.

Table of Content