Loading...
  • 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

Current Issue

    • A periodical characteristic-based resource
      prediction method for datacenter online services
       
      LIANG Yi,ZENG Shao-kang,LIANG Yan-de,DING Yi
      2020, 42(03): 381-390. doi:
      Abstract ( 101 )   PDF (1898KB) ( 199 )      Review attachment
      Online services, such as web service and stream processing, are the major workloads in modern datacenters. The variations in request arrival rates lead to the dynamic changes in the online services’ resource demands during their executions. Therefore, fast and accurate resource prediction of online services is essential to reasonable resource allocation in datacenters and guaranteeing the efficiency of service execution. However, existing resource prediction methods of online services can neither conduct the long-term accurate prediction, nor conduct the prediction with limited sample dataset and low time overhead. To this end, this paper proposes a resource prediction method of online services based on the periodic characteristic of requests, called PRP. Oriented to the periodic characteristics of online ser- vice requests, PRP adopts the autocorrelation function to recognize the changing period of the online services' resource usage. Then, based on the changing period, it divides the resource usage sample sequence and classifies the resource usage subsequences. Finally, based on the classified resource usage subsequences, it uses linear weighting to predict the resources of online services. Experimental results demonstrate that PRP outperforms the existing resource prediction methods in both the prediction accuracy and the computational efficiency.
       
      A node performance evaluation method for heterogeneous
      clusters based on PageRank and benchmarks
      HU Ya-hong1,WANG Yi-zhou2,MAO Jia-fa1
      2020, 42(03): 391-396. doi:
      Abstract ( 127 )   PDF (419KB) ( 147 )      Review attachment
      For a cluster to achieve its maximum throughput, data placement and task scheduling should be handled according to the performance of the cluster nodes. In a heterogeneous cluster, each node has quite different performance, and how to evaluate nodes’ performance is a challenge issue. Normally, nodes are evaluated by benchmarks, and different benchmarks evaluate the nodes from different aspects. PageRank algorithm is used by Google to rank web sites and now it is also applied to evaluate the influence of books or users' behavior, etc. A novel PageRank based node performance evaluation algorithm is proposed to take advantage of the evaluation results from different benchmarks. Firstly, each node is evaluated by mainstream benchmarks, such as LINPACK, NPB and IOzone. Secondly, PageRank algorithm is applied to calculate the nodes’ performance according to the execution results from each benchmark. In order to use PageRank algorithm, a graph model is established, and performance vectors and probability transition matrix are also calculated. The proposed algorithm can produce comprehensive evaluation results with low computational complexity.

       
      Image classification algorithm based on
      spiking neural network and mobile GPU computing
      XU Pin-jie1,2,WANG Hui-zhe1,2,LI Ce3,TANG Dan1,ZHAO Di1
      2020, 42(03): 397-403. doi:
      Abstract ( 186 )   PDF (943KB) ( 244 )      Review attachment
      Computer vision is designed to simulate human visual systems through machines, which is a hot spot in artificial intelligence and neuroscience research. As a classical task of computer vision, image classification has attracted more and more researches, especially the image classification algorithms based on neural networks perform well on various classification tasks. However, the traditional shallow artificial neural networks have weak feature learning ability and insufficient bio-interpretability, while the deep neural networks have the disadvantages of over-fitting and high power consumption. Therefore, the bio-interpretable classification algorithm in low power environments is still a challenging task. In order to solve the above problems, a classification method based on spiking neural network in Jetson TK1 development environment is designed. The main innovations of the research are as follows:
      (1) Designing a spiking convolution neural network for image classification;
      (2) Implementing the improved spiking neural network based on CUDA and deploying it in Jetson TK1.
       
       
      A hybrid parallel computing method of WKBZ normal model
      FAN Pei-qin1,3,LIU Xiao-yan2,GUO Wu-hong1,3,CUI Bao-long1,3
      2020, 42(03): 404-410. doi:
      Abstract ( 110 )   PDF (858KB) ( 139 )      Review attachment
      In order to solve the problem that the calculation of the underwater acoustic propagation model is large and it is difficult to meet the real-time and delicacy demands of underwater acoustic environment information assurance, based on the MPI+OpenMP hybrid parallel programming method, a hybrid parallel computing method of WKBZ normal model is proposed to realize the two-stage hybrid parallel computing of underwater acoustics. This method can effectively overcome the shortcomings of high communication cost of MPI and poor scalability of OpenMP by message transmission between nodes and shared memories in nodes, and it solves the fast calculation problem of underwater acoustic propagation. The test results show that this method can make good use of the multi-level parallel mechanism between the nodes in the SMP clusters and within the nodes, give full play to the advantages of the message passing programming model and the shared memory programming model, greatly reduce the MPI process communication time ,and effectively improve the scalability and efficiency of the parallel program.
       
       
      A DGA domain name detection method based on Transformer
      ZHANG Xin,CHENG Hua,FANG Yi-quan
      2020, 42(03): 411-417. doi:
      Abstract ( 205 )   PDF (725KB) ( 197 )     
      Existing DGA detection methods have achieved high detection accuracy, but there is a problem of high false alarm rate in abbreviated domain names. The main reason is that the abbreviated domain names have high randomness among characters and it is difficult for the existing detection methods to distinguish abbreviated domain names from DGA domain names. After analyzing the character characteristics of the abbreviated domain names, the detection of domain name character dependence is realized based on self-attention mechanism. Then, LSTM is used to improve the encoding way of Transformer model to better capture the location information of characters in domain names. A DGA domain name detection method (MHA) is constructed based on Transformer model. Experimental results show that the algorithm can effectively distinguish DGA domain names from abbreviated domain names, and get higher accuracy and lower false alarm rate.

       

       

       
       
      A certificateless public key authenticated searchable
      encryption scheme with efficient authorization
      LANG Xiao-li,CAO Su-zhen,LIU Xiang-zhen,ZHANG Yu-lei,WANG Fei
      2020, 42(03): 418-426. doi:
      Abstract ( 126 )   PDF (596KB) ( 159 )     
      This paper designs a security model of the public key authenticated searchable encryption scheme in efficiently authoried certificateless environments, and proposes a specific certificateless public key authenticated searchable encryption scheme with efficient authorization. In this scheme, the cloud server uses the signature of the data owner on the ciphertext keyword index to verify the data owner identity. Secondly, the data user authorizes the authorization server to verify the identity of the data user. If the data user is legal, the server help the data user to verify the validity of the ciphertext returned by the cloud server. At the same time, the data owner and the data user use the cloud server public key to generate the ciphertext keyword index and the trapdoor search credentials that satisfies  the transmission security on the public channel. Finally, the efficiency of the proposed scheme is verified by experimental simulation.

       
       
      An intrusion detection technology based on NBSR model
      ZHU Shi-song,BA Meng-long,WANG Hui,SHEN Zi-hao
      2020, 42(03): 427-433. doi:
      Abstract ( 172 )   PDF (602KB) ( 170 )     
      In order to better solve the problem of increasing the false positive rate of unknown intrusion behaviors by misuse detection in intrusion detection technology, an intrusion detection technology based on NBSR model is proposed. Firstly, in order to compensate for the lack of correlation analysis between features by the ReliefF feature selection algorithm, the Pearson correlation coefficient is introduced and the Relieff-P algorithm is proposed. Secondly, the Relieff-P algorithm is used to process the UNSW-NB15 dataset to remove irrelevant features and obtain a new feature subset. Finally, the naive Bayes classifier and the Softmax regression classifier are cascaded to form the NBSR classifier, and NBSR model was established. The experimental results on the UNSW-NB15 test set show that the NBSR model has lower false positive rate than other detection models.

       
      LoRa parameter matching optimization
      based on multi-objective genetic algorithm
       
      WANG Shuo-he1,LIU Xu1,LI Su-chen1,ZHANG Guo-ju2
      2020, 42(03): 434-440. doi:
      Abstract ( 120 )   PDF (688KB) ( 146 )     
      From the perspective of engineering application, LoRa wireless transmission system is required to have the characteristics of low system power consumption, short transmission distance and good system stability. Therefore, optimizing the matching parameters in its design is the key way to improve LoRa transmission performance. In this paper, aiming to achieve the lowest energy consumption, the longest transmission distance and the strongest system robustness of the LoRa wireless communication, the effective values of parameters such as SF, BW and CR are considered as the constraint condition, and the linear weighting method is adopted to convert the multi-objective optimization problem to a single-objective problem so as to work out the optimal solution. The simulation and actual test results show that the genetic algorithm applied to LoRa parameter matching is feasible and effective.

       

       
      A provably secure privacy-preserving multi-recipient
      heterogeneous aggregate signcryption scheme
      LIU Xiang-zhen1,ZHANG Yu-lei1,LANG Xiao-li1,LUO Guang-ping1,WANG Cai-fen2
      2020, 42(03): 441-448. doi:
      Abstract ( 96 )   PDF (529KB) ( 145 )     
      The heterogeneous aggregation signcryption technology not only solves the problem of communication under different cryptosystems, but also performs aggregate signature verification on multiple messages. This paper analyzes the heterogeneous signcryption scheme proposed by Niu et al. that ensures data privacy, and points out that both single signcryption and aggregation signcryption can be forged in this scheme, and there are passive attacks in the key generation center. Firstly, we detail a specific attack process, which illustrates that the Niu scheme generates passive attacks. Secondly, we improve the Niu scheme and prove that the improved scheme has no security holes through security analysis. Finally, performance analysis and simulation show that the improved scheme has the equivalent efficiency to the original scheme.

       
       
      A RFID tag number estimation algorithm
       based on bit estimation
      YANG Fan1,2,REN Shou-gang3,XU Huan-liang3,SUN Yuan-hao2,YANG Xing2
      2020, 42(03): 449-455. doi:
      Abstract ( 103 )   PDF (814KB) ( 106 )     
      In order to further improve the accuracy of the tag number estimation algorithms in dense tags environment, based on the analysis and comparison of the traditional bit estimation algorithms, a novel optimized bit estimation algorithm is proposed. Firstly, based on the binomial distribution theory, the observation of non-selected bits is used to calculate the idle bit rate. Then, a tag number estimation model in dense tags environment is established by determining the threshold of idle bit rate. Finally, the mathematical expression of the relationship between the estimated tag number and the slot consumption is derived. The simulation results show that the proposed algorithm has better estimation accuracy than the traditional bit estimation algorithms and is stable under different numbers of tags.

       

       

       
      An automatic transformation method
       from SysML model to AADL model
      MA Yan-yan1,YANG Zhi-bin1,2,JIANG Guo-hua1
      2020, 42(03): 456-466. doi:
      Abstract ( 285 )   PDF (1685KB) ( 234 )     
      The implementation of safety-critical systems requires multiple stages, including requirements, design, integration, verification, and testing. In recent years, model-driven development (MDD) has gradually become an important means for the design and development of safety-critical software. As no modeling language that can support the entire safety-critical system development lifecycle, we choose to integrate two widely used standard languages: system modeling language (SysML) and architecture analysis & design language of embedded real-time system (AADL). SysML and AADL provide two different views of the same system. While SysML provides a high-level system view for systems engineers, AADL establishes a low-level design view for architects, which combines hardware, opera- ting systems, and codes that realize all functions. This paper proposes an automatic transformation method from SysML to AADL. Firstly, SysML subset, called SubSysML, is defined, which mainly includes block definition diagram (BDD), internal block diagram (IBD), activity diagram (ACT) subset, and AADL Profile extended from IBD and BDD. Secondly, the transformation rules from SubSysML to AADL are defined and the transformation algorithm is designed. Then, the initial model of AADL is refined. Finally, EMF framework technology is used to implement the automatic transformation tool, and radar cases are adopted to verify the effectiveness of the proposed method.

       

       

       
      A lane line detection method based
      on dense segmentation network
      DING Hai-tao,SUN Rui,CHENG Xu-sheng,GAO Jun
      2020, 42(03): 467-473. doi:
      Abstract ( 143 )   PDF (887KB) ( 180 )     
      Most traditional lane detection algorithms rely on the combination of handcrafted features and heuristic algorithms, which are easily affected by factors such as vehicle occlusion and ground fouling. Aiming at the complicated problems that affect lane line detection, this paper considers lane line detection as a continuous segmentation problem, and proposes a lane line detection method based on dense segmentation network. To this end, dense blocks are used to construct a dense segmentation network (DSNet), so DSNet can reuse features to improve the performance of extracting lane line instance features and restoring the feature map resolution. At the same time, the proximity AND operation and Meanshift clustering algorithm are also introduced to process the output of DSNet, which reduces the influence of non-lane line pixels and makes the boundary of detection results more specific. Experiments show that the proposed algorithm can well solve the problem of vehicle occlusion and ground fouling, and can also determine the number of lane lines, which has better robustness and real-time performance.
       
      A rotation mean pulsation feature extraction method
      and its application in fuzzy face recognition
      ZHONG Guo-yun,WANG Meng-meng,WANG Yu-ling,CHANG Yan-rong,WU Zhong-liang
      2020, 42(03): 474-482. doi:
      Abstract ( 71 )   PDF (1114KB) ( 127 )     
      Aiming at the problem of poor recognition rate of blurred face caused by long-distance shooting of surveillance cameras at present, a method for extracting the pulsation feature of rotational mean with orderly global structural characteristics is proposed. In this method, several sampling points are selected equally on each vertical line of the image according to the order from top to bottom, and the mean pulsation method is used for coding. Firstly, the average values of all non-zero pixels on each vertical line are calculated, and the pixel values and average values of selected sampling points are compared and coded in turn in order to generate an 8-bit binary number. The range of the decimal values is the same as the range of the pixel values. The decimal number is the characteristic value of the whole vertical line, so the texture feature information describing each vertical line is extracted. Texture image fusion feature information is extracted by combining image preprocessing with histogram normalization. The experimental results show that the proposed method obviously outperforms deep learning in the field of fuzzy face recognition.
       
      An automatic design structure matrix generation
       system based on 3D CAD model
       
      LI Zhong-kai,YIN Wen-wei
      2020, 42(03): 483-492. doi:
      Abstract ( 128 )   PDF (1314KB) ( 159 )     
      Design structural matrix (DSM) has become a modeling and analytical tool for complex systems in the researches and practices of many areas. However, building DSM faces great difficulties, because the decomposition of each product and the different interpretation of the structural relationship are difficult to build in a standard model. This makes it difficult to realize the rapid generation of DSM and shorten its product design cycle. This paper introduces a new automatic DSM generation system based on CAD model. Firstly, the mates feature information of the top-level component in the structural feature tree of the CAD model in the SolidWorks platform is extracted by SolidWorks API technology. The extracted information is then arranged in a certain order and stored in the database. Secondly, the geometric relationship between the assembled components is analyzed in order to evaluate the degree of the influence of various types of mates on the connection relationship between the components. Thus, the corresponding rules of analysis and comparison are set up. Finally, the design of the automatic DSM generation system was successfully accomplished in Visual Basic. Meanwhile, some CAD models are used as examples to verify the feasibility and reliability of the developed system.

       
      Video stabilization algorithm based on
      optical flow method and Kalman filtering
      XIONG Wei1,2,WANG Chuan-sheng1,LI Li-rong1,LIU Min1,ZENG Chun-yan1
      2020, 42(03): 493-499. doi:
      Abstract ( 457 )   PDF (834KB) ( 223 )     
      Aiming at the video jitter problem caused by mobile phone shooting, a video stabilization algorithm based on optical flow method and Kalman filter is proposed. Firstly, the optical video method is used to pre-stabilize the dithered video, and Shi-Tomasi corner detection is performed on the pre-stabilized video frame. The LK algorithm is used to track the corner points, and then the RANSAC algorithm is used to estimate the affine transformation matrix between adjacent frames, so as to obtain the original camera path. Secondly, the smoothed camera path is optimized by the Kalman filter to obtain a smooth camera path. Finally, the relationship between the original camera path and the smooth path is used to calculate the compensation matrix between adjacent frames, and then the compensation matrix is used to geometrically transform the frames one by one, resulting in a stable video output. Experiments show that the proposed algorithm has good effects in processing six types of jitter videos. The PSNR value after image stabilization is increased by 6.631 dB compared with the original video, and the structural similarity (SSIM) between video frames is increased by 40%. The average curvature value is increased by about 8.3%.

       
      Facial expression recognition using
      feature fusion based on VGG-NET
      LI Xiao-lin1,2,3,NIU Hai-tao1,2
      2020, 42(03): 500-509. doi:
      Abstract ( 193 )   PDF (830KB) ( 234 )     
      Convolutional Neural Networks (CNN) and Local Binary Patterns (LBP) can only extract single features of facial expression images during facial expression feature extraction, so it is difficult to extract the precise features related to facial changes. In order to solve this problem, this paper proposes a facial expression recognition method using feature fusion based on deep learning. The method combines the LBP feature and the features extracted by the CNN convolutional layer into the improved VGG-16 network connection layer by weighting. Finally, the fusion features are sent to the Softmax classifier to obtain the probability of various features, and complete the basic six expression classifications. The experimental results show that the average recognition accuracy of the proposed method on the CK+ and JAFFE datasets is 97.5% and 97.62%, respectively. The recognition results obtained by the fusion features are significantly superior to that of single feature recognition. Compared with other methods, this method can effectively improve the accuracy of expression recognition and is more robust to illumination changes.

       

       

       
      Non-reference stereo image quality
      evaluation based on binocular fusion
      WANG Yang1,2,XIANG Xiu-mei1,2,LU Jia1,2,YU Zhen-xin1,2
      2020, 42(03): 510-516. doi:
      Abstract ( 103 )   PDF (825KB) ( 147 )     
      Aiming at the evaluation problem of symmetric distortion and asymmetric distortion image, a non-reference stereo image quality evaluation method based on binocular fusion is proposed. Firstly, the left and right viewpoint images of the stereo image are decomposed into Laplacian pyramid sequences respectively, and the fusion coefficients of each layer are determined by using the image ave- rage gradient and the region energy. On the basis of the binocular weighted model, the two sequences are merged layer by layer and the cyclopean image is reconstructed. Then, the multi-scale, multi-directional frequency domain transform features and the contrast, entropy, energy and inverse difference moment features of the left and right viewpoint images and the cyclopean images are extracted. Finally, feature parameters are trained as input to the support vector regression model and the image quality is predicted. The correlation analysis is performed under LIVE 3D phase I and LIVE 3D phase II image databases. The Pearson linear correlation coefficient and Spearman rank correlation coefficient reach 0.96 and 0.95 respectively. The results show that the prediction results of stereo image quality have higher consistency with subjective evaluation values.

       

       



       
       
      Health assessment of wind turbine generator
      based on improved stacked auto-encoder
       
      LIN Tao,ZHAO Cheng-lin,LIU Hang-peng,ZHAO Shen-shen
      2020, 42(03): 517-522. doi:
      Abstract ( 89 )   PDF (691KB) ( 258 )     
      Wind turbine generator has the characteristics of complicated structure and difficult maintenance. In order to evaluate its health, this paper combines the characteristics of denoising auto-encoder and sparse auto-encoder to improve the traditional stacked auto-encoder model, and uses the reconstruction error of the model to monitor the running state of the wind turbine generator. The reconstruction error obtained by off-line testing is compared with the reconstruction error obtained by online monitoring, and the health of the wind turbine generator is obtained by combining three different indicators. The health assessment model is trained and tested by using the actual data of a wind farm in Hebei Pro- vince. The example analysis shows that the method can effectively track the state change of the wind turbine generator and has the function of early identification of faults.

       

       

       

       

       
      Logistics transportation route research based
      on improved ant colony algorithm
      MA Gui-ping, PAN Feng
      2020, 42(03): 523-528. doi:
      Abstract ( 193 )   PDF (570KB) ( 277 )     
      In order to quickly solve the optimal path of logistics transportation in a complex traffic environment, on the basis of traditional ant colony algorithm, a logistics transportation path optimization model based on improved ant colony algorithm is proposed. Firstly, the model adds constraints based on transportation time, cost and average road smoothness factor to the traditional ant colony algorithm, and improves the updating method of traditional pheromone to limit the maximum and minimum pheromone concentration on the road, so as to change the transfer probability of path selection. Finally, simulation experiments are carried out on the improved ant colony algorithm, CSAACO algorithm and ACO algorithm. Under the same experimental environment, the distance and time reduction of the three algorithms are tested. The experimental data show that the improved ant colony algorithm has significantly shorter transportation distance and less transportation time than CSAACO algorithm and ACO algorithm. The improved ant colony algorithm has stronger global optimization ability, faster convergence speed, less time, and shorter optimal path, and improves the transportation efficiency of the entire logistics industry.
       
      A news recommendation algorithm based on user behavior
      LI Zeng, LIU Yu, LI Cheng-cheng
      2020, 42(03): 529-534. doi:
      Abstract ( 147 )   PDF (727KB) ( 205 )     
      In order to improve the efficiency and accuracy of news recommendation and reduce the repeated recommendation of similar content, through the research of user behavior and the analysis of user's news browsing behavior log, a news recommendation algorithm with Markov algorithm as the main algorithm is adopted. The algorithm is supplemented by the collaborative filtering algorithm and the content-based recommendation algorithm to establish a Markov model, thus realizing the application in intelligent news recommendation. Test results show that, compared with the traditional recommendation algorithm, the algorithm significantly improves the accuracy and execution efficiency, and its function is more intelligent.