Table of Content

    10 June 2020, Volume 40 Issue 6
    Artificial intelligence
    Review of anomaly detection algorithms for multidimensional time series
    HU Min, BAI Xue, XU Wei, WU Bingjian
    2020, 40(6):  1553-1564.  DOI: 10.11772/j.issn.1001-9081.2019101805
    Asbtract ( )   PDF (930KB) ( )  
    References | Related Articles | Metrics

    With the continuous development of information technology, the scale of time series data has grown exponentially, which provides opportunities and challenges for the development of time series anomaly detection algorithm, making the algorithm in this field gradually become a new research hotspot in the field of data analysis. However, the research in this area is still in the initial stage and the research work is not systematic. Therefore, by sorting out and analyzing the domestic and foreign literature, this paper divides the research content of multidimensional time series anomaly detection into three aspects: dimension reduction, time series pattern representation and anomaly pattern detection in logical order, and summarizes the mainstream algorithms to comprehensively show the current research status and characteristics of anomaly detection. On this basis, the research difficulties and trends of multi-dimensional time series anomaly detection algorithms were summarized in order to provide useful reference for related theory and application research.

    Survey of sub-topic detection technology based on internet social media
    LI Shanshan, YANG Wenzhong, WANG Ting, WANG Lihua
    2020, 40(6):  1565-1573.  DOI: 10.11772/j.issn.1001-9081.2019101871
    Asbtract ( )   PDF (666KB) ( )  
    References | Related Articles | Metrics

    The data in internet social media has the characteristics of fast transmission, high user participation and complete coverage compared with traditional media under the background of the rise of various platforms on the internet.There are various topics that people pay attention to and publish comments in, and there may exist deeper and more fine-grained sub-topics in the related information of one topic. A survey of sub-topic detection based on internet social media, as a newly emerging and developing research field, was proposed. The method of obtaining topic and sub-topic information through social media and participating in the discussion is changing people’s lives in an all-round way. However, the technologies in this field are not mature at present, and the researches are still in the initial stage in China. Firstly, the development background and basic concept of the sub-topic detection in internet social media were described. Secondly, the sub-topic detection technologies were divided into seven categories, each of which was introduced, compared and summarized. Thirdly, the methods of sub-topic detection were divided into online and offline methods, and the two methods were compared, then the general technologies and the frequently used technologies of the two methods were listed. Finally, the current shortages and future development trends of this field were summarized.

    Domain ontology driven approach for bidding webpage parsing
    MA Dongxue, SONG She, XIE Zhenping, LIU Yuan
    2020, 40(6):  1574-1579.  DOI: 10.11772/j.issn.1001-9081.2019101792
    Asbtract ( )   PDF (3054KB) ( )  
    References | Related Articles | Metrics

    In order to solve the low efficiency problem of parsing bidding webpages when using regular expression, a new automatic method was proposed based on bidding ontology model. Firstly, the structural features of bidding webpage texts were analyzed. Furthermore, a lightweight domain knowledge model on bidding ontology was constructed. Finally, a new algorithm for semantic matching and extraction of bidding webpage elements was introduced to realize the automatic parsing of bidding webpages. The experimental results show that, the accuracy and recall of the new method can reach 95.33% and 88.29% respectively by adaptive parsing. Compared with the regular expression method, the performance can be improved by 3.98 percentage points and 3.81 percentage points respectively. The proposed method can adaptively realize the structured parsing and extraction of semantic information in bidding webpages, and can satisfy the requirements of practical applications.

    Question classification of common crop disease question answering system based on BERT
    YANG Guofeng, YANG Yong
    2020, 40(6):  1580-1586.  DOI: 10.11772/j.issn.1001-9081.2019111951
    Asbtract ( )   PDF (719KB) ( )  
    References | Related Articles | Metrics
    As a key module of the question answering system, question classification is also a key factor that restricts the retrieval efficiency of the question answering system. Aiming at the problems of complicated semantic information and large differences of user questions in agricultural question answering system, in order to meet the needs of users to quickly and accurately obtain classification results of common crop disease questions, the question classification model of common crop disease question answering system based on Bidirectional Encoder Representations from Transformers (BERT) was constructed. Firstly, the question dataset was preprocessed. Then, Bidirectional-Long Short Term Memory (Bi-LSTM) self-attention network classification model, Transformer classification model and BERT-based fine-tuning classification model were constructed respectively, and the three models were used to extract information of questions and train question classification model. Finally, the BERT-based fine-tuning classification model was tested and the impact of dataset size on classification results was explored. The experimental results show that, the BERT-based fine-tuning common crop disease question classification model has the classification accuracy, precision, recall, weighted harmonic mean of accuracy and recall higher than those of the Bi-LSTM self-attention network classification model and the Transformer classification model by 2-5 percentage points respectively. On Common Crop Disease Question Dataset (CCDQD), it can obtain the highest accuracy of 92.46%, precision of 92.59%, recall of 91.26%, and weighted harmonic mean of accuracy and recall of 91.92%. The BERT-based fine-tuning classification model has advantages of simple structure, few parameters and fast speed, and can efficiently classify common crop disease questions accurately. So, it can be used as the question classification model for the common crop disease question answering system.
    Application of convolutional neural network with threshold optimization in image annotation
    CAO Jianfang, ZHAO Aidi, ZHANG Zibang
    2020, 40(6):  1587-1592.  DOI: 10.11772/j.issn.1001-9081.2019111993
    Asbtract ( )   PDF (695KB) ( )  
    References | Related Articles | Metrics

    Ranking function based annotation may cause more or fewer labels according to the probability predicted by the model in multi-label image annotation. Therefore, a Convolutional Neural Network with THreshold OPtimization (CNN-THOP) model was proposed. The model consists of Convolutional Neural Network (CNN) and threshold optimization. Firstly, CNN was used to train a model, which was used to predict the image, so as to obtain the prediction probability, and Batch Normalization (BN) layer was added to the CNN to effectively accelerate the convergence. Secondly, threshold optimization was performed by the prediction probabilities of the test set images obtained by the proposed model. After the threshold optimization process, an optimal threshold was obtained for each kind of label, so as to obtain a set of optimal thresholds. Only when the prediction probability of this kind of label was greater than or equal to the best threshold of this kind of label, the image would be labeled with this label. In the labeling process, the CNN model and a set of optimal thresholds were added to achieve more flexible multi-label labeling of the image to be labeled. Through the verification on 8 000 images in the natural scene image dataset, experimental results show that CNN-THOP has about 20 percentage points improvement on average precision compared to Ranking Support Vector Machine (Rank-SVM), and is about 6 percentage points and 4 percentage points higher respectively than Convolutional Neural Network using Mean Square Error function (CNN-MSE) in average recall and F1 value respectively, and has the Complete Matching Degree (CMD) reached 64.75%, which proves that the proposed method is effective in automatic image annotation.

    Automatic annotation of visual deep neural network
    LI Ming, GUO Chenhao, CHEN Xing
    2020, 40(6):  1593-1600.  DOI: 10.11772/j.issn.1001-9081.2019101774
    Asbtract ( )   PDF (3594KB) ( )  
    References | Related Articles | Metrics
    Focused on the issue that developers cannot quickly figure out the models they need from various models, an automatic annotation method of visual deep neural network based on natural language processing technology was proposed. Firstly, the field categories of visual neural networks were divided, the keywords and corresponding weights were calculated according to the word frequency and other information. Secondly, a keyword extractor was established to extract keywords from paper abstracts. Finally, the similarities between extracted keywords and the known weights were calculated in order to obtain the application fields of a specific model. With experimental data derived from the papers published in three top international conferences of computer vision: IEEE International Conference on Computer Vision(ICCV), IEEE Conference on Computer Vision and Pattern Recognition(CVPR) and European Conference on Computer Vision(ECCV), the experiments were carried out. The experimental results indicate that the proposed method provides highly accurate classification results with a macro average value of 0.89. The validity of this proposed method is verified.
    Relation extraction method based on dynamic label
    XUE Lu, SONG Wei
    2020, 40(6):  1601-1606.  DOI: 10.11772/j.issn.1001-9081.2019111959
    Asbtract ( )   PDF (708KB) ( )  
    References | Related Articles | Metrics
    Concerning the problem that the research methods of relation extraction for distant supervision datasets have a lot of label noise, a dynamic label method applied to the hierarchical attention mechanism relation extraction model was proposed. Firstly, a concept of generating dynamic label based on the similarity of relation categories was proposed. Since the same relation labels contain similar feature information, calculating the similarity of relation categories of feature information is helpful to generate the dynamic label corresponding to the feature information. Secondly, the scoring function of the dynamic label was used to evaluate whether the distant supervision label was noise and to determine whether a new label was needed to generate to replace the distant supervision label, and the influence of label noise on the model was suppressed by adjusting the distant supervision label. Finally, according to the dynamic label, the hierarchical attention mechanism was updated to focus on the effective instances, the importance of each effective instance was relearned and key relation feature information was further extracted. The experimental results indicate that, compared with the original hierarchical attention mechanism relation extraction model, the proposed method has the Micro and Macro scores increased by 1.3 percentage points and 1.9 percentage points respectively, realizes the dynamic correction of the noise label, and improves the relation extraction ability of the model.
    3D point cloud classification and segmentation network based on Spider convolution
    WANG Benjie, NONG Liping, ZHANG Wenhui, LIN Jiming, WANG Junyi
    2020, 40(6):  1607-1612.  DOI: 10.11772/j.issn.1001-9081.2019101879
    Asbtract ( )   PDF (689KB) ( )  
    References | Related Articles | Metrics

    The traditional Convolutional Neural Network (CNN) cannot directly process point cloud data, and the point cloud data must be converted into a multi-view or voxelized grid, which leads to a complicated process and low point cloud recognition accuracy. Aiming at the problem, a new point cloud classification and segmentation network called Linked-Spider CNN was proposed. Firstly, the deep features of point cloud were extracted by adding more Spider convolution layers based on Spider CNN. Secondly, by introducing the idea of residual network, short links were added to every Spider convolution layer to form residual blocks. Thirdly, the output features of each layer of residual blocks were spliced and fused to form the point cloud features. Finally, the point cloud features were classified by three-layer fully connected layers or segmented by multiple convolution layers. The proposed network was compared with other networks such as PointNet, PointNet++ and Spider CNN on ModelNet40 and ShapeNet Parts datasets. The experimental results show that the proposed network can improve the classification accuracy and segmentation effect of point clouds, and it has faster convergence speed and stronger robustness.

    Multi-agent collaborative pursuit algorithm based on game theory and Q-learning
    ZHENG Yanbin, FAN Wenxin, HAN Mengyun, TAO Xueli
    2020, 40(6):  1613-1620.  DOI: 10.11772/j.issn.1001-9081.2019101783
    Asbtract ( )   PDF (899KB) ( )  
    References | Related Articles | Metrics
    The multi-agent collaborative pursuit problem is a typical problem in the multi-agent coordination and collaboration research. Aiming at the pursuit problem of single escaper with learning ability, a multi-agent collaborative pursuit algorithm based on game theory and Q-learning was proposed. Firstly, a cooperative pursuit team was established and a game model of cooperative pursuit was built. Secondly, through the learning of the escaper’s strategy choices, the trajectory of the escaper’s limited Step-T cumulative reward was established, and the trajectory was adjusted to the pursuer’s strategy set. Finally, the Nash equilibrium solution was obtained by solving the cooperative pursuit game, and the equilibrium strategy was executed by each agent to complete the pursuit task. At the same time, in order to solve the problem that there may be multiple equilibrium solutions, the virtual action behavior selection algorithm was added to select the optimal equilibrium strategy. C# simulation experiments show that, the proposed algorithm can effectively solve the pursuit problem of single escaper with learning ability in the obstacle environment, and the comparative analysis of experimental data shows that the pursuit efficiency of the algorithm under the same conditions is better than that of pure game or pure learning.
    Video translation model from virtual to real driving scenes based on generative adversarial dual networks
    LIU Shihao, HU Xuemin, JIANG Bohou, ZHANG Ruohan, KONG Li
    2020, 40(6):  1621-1626.  DOI: 10.11772/j.issn.1001-9081.2019101802
    Asbtract ( )   PDF (1339KB) ( )  
    References | Related Articles | Metrics
    To handle the issues of lacking paired training samples and inconsistency between frames in translation from virtual to real driving scenes, a video translation model based on Generative Adversarial Networks was proposed in this paper. In order to solve the problem of lacking data samples, the model adopted a “dual networks” architecture, where the semantic segmentation scene was used as an intermediate transition to build front-part and back-part networks, respectively. In the front-part network, a convolution network and a deconvolution network were adopted, and the optical flow network was also used to extract the dynamic information between frames to implement continuous video translation from virtual to semantic segmentation scenes. In the back-part network, a conditional generative adversarial network was used in which a generator, an image discriminator and a video discriminator were designed and combined with the optical flow network to implement continuous video translation from semantic segmentation to real scenes. Data collected from an autonomous driving simulator and a public data set were used for training and testing. Virtual to real scene translation can be achieved in a variety of driving scenarios, and the translation effect is significantly better than the comparative algorithms. Experimental results show that the proposed model can handle the problems of the discontinuity between frames and the ambiguity for moving obstacles to obtain more continuous videos when applying in various driving scenarios.
    Extreme learning machine algorithm based on cloud quantum flower pollination
    NIU Chunyan, XIA Kewen, ZHANG Jiangnan, HE Ziping
    2020, 40(6):  1627-1632.  DOI: 10.11772/j.issn.1001-9081.2019101846
    Asbtract ( )   PDF (919KB) ( )  
    References | Related Articles | Metrics
    In order to avoid the flower pollination algorithm falling into local optimum in the identification process of the extreme learning machine, an extreme learning machine algorithm based on cloud quantum flower pollination was proposed. Firstly, cloud model and quantum system were introduced into the flower pollination algorithm to enhance the global search ability of the flower pollination algorithm, so that the particles were able to perform optimization in different states. Then, the cloud quantum flower pollination algorithm was used to optimize the parameters of the extreme learning machine in order to improve the identification accuracy and efficiency of the extreme learning machine. In the experiments, six benchmark functions were used to simulate and compare several algorithms. It is verified by the comparison results that the performance of proposed cloud quantum flower pollination algorithm is superior to those of other three swarm intelligence optimization algorithms. Finally, the improved extreme learning machine algorithm was applied to the identification of oil and gas layers. The experimental results show that, the identification accuracy of the proposed algorithm reaches 98.62%, and compared with the classic extreme learning machine, its training time is shortened by 1.680 2 s. The proposed algorithm has high identification accuracy and efficiency, and can be widely applied to the actual classification field.
    Xgboost algorithm optimization based on gradient distribution harmonized strategy
    LI Hao, ZHU Yan
    2020, 40(6):  1633-1637.  DOI: 10.11772/j.issn.1001-9081.2019101878
    Asbtract ( )   PDF (515KB) ( )  
    References | Related Articles | Metrics
    In order to solve the problem of low detection rate of minority class by ensemble learning model eXtreme gradient boosting (Xgboost) in the binary classification problem, an improved Xgboost algorithm based on gradient distribution harmonized strategy called Loss Contribution Gradient Harmonized Algorithm (LCGHA)-Xgboost was proposed. Firstly, Loss Contribution (LC) was defined to simulate the losses of the samples in Xgboost algorithm. Secondly, by defining Loss Contribution Density (LCD), the difficulty of samples being correctly classified in Xgboost algorithm was measured. Finally, a gradient distribution harmonized algorithm called LCGHA was proposed to dynamically adjust the one order gradient distribution of samples according to the LCD. In the algorithm, the losses of hard samples (mainly in minority class) were indirectly increased, and the losses of easy samples (mainly in majority class) were indirectly reduced, making Xgboost algorithm tend to learn the hard samples. The experimental results show that compared with three ensemble learning algorithms Xgboost, GBDT (Gradient Boosting Decision Tree) and Random_Forest, LCGHA-Xgboost has the recall increased by 5.4%-16.7%, and Area Under the Curve (AUC) improved by 0.94%-7.41% on multiple UCI datasets, and the Recall increased by 44.4%-383.3%, and AUC improved by 5.8%-35.6% on WebSpam-UK2007 and DC2010 datasets. LCGHA-Xgboost can effectively improve the classification and detection ability for minority class, and reduce the classification error rate of minority class.
    Data science and technology
    PiFlow: model driven big data pipeline framework
    ZHU Xiaojie, ZHAO Zihao, DU Yi
    2020, 40(6):  1638-1647.  DOI: 10.11772/j.issn.1001-9081.2019101793
    Asbtract ( )   PDF (1594KB) ( )  
    References | Related Articles | Metrics
    Big data processing with complex process mostly relies on pipeline systems. However, the pipeline systems of big data processing have some shortcomings in usability, function reusability, expansibility and processing performance. In order to solve the problems and improve the construction and development efficiency of big data processing environment and optimize the processing flow, a model driven big data pipeline framework called PiFlow was proposed. Firstly, the big data processing process was abstracted as a directed acyclic graph. Then, a series of components were developed to construct the data processing pipeline, and the pipeline task execution mechanism was designed. At the same time, in order to standardize and simplify the pipeline framework description, a model driven big data pipeline description language called PiFlowDL was designed, which described the big data processing tasks in a modular and hierarchical way. PiFlow configures the pipeline in a What You See Is What You Get (WYSIWYG) way, and integrates the functions such as status monitoring, template configuration, and component integration. Compared with Apache NiFi, it has the performance improvement of 2-7 times.
    Joint low-rank and sparse multiple kernel subspace clustering algorithm
    LI Xingfeng, HUANG Yuqing, REN Zhenwen
    2020, 40(6):  1648-1653.  DOI: 10.11772/j.issn.1001-9081.2019111991
    Asbtract ( )   PDF (1768KB) ( )  
    References | Related Articles | Metrics
    Since the methods of multiple kernel subspace spectral clustering do not consider the problem of noise and relation graph structure, a novel Joint Low-rank and Sparse Multiple Kernel Subspace Clustering algorithm (JLSMKC) was proposed. Firstly, with combination of low-rank and sparse representation for subspace learning, the relation graph obtained the attribute of low-rank and sparse structure. Secondly, a robust multiple kernel low-rank and sparsity constraint model was constructed to reduce the influence of noise on the relation graph and handle the nonlinear structure of data. Finally, the quality of relation graph was enhanced by making full use of the consensus kernel matrix by multiple kernel approach. The experimental results on seven datasets show that the proposed JLSMKC is better than five popular multiple kernel clustering algorithms in ACCuracy (ACC), Normalized Mutual Information (NMI) and Purity. Meanwhile, the clustering time is reduced and the block diagonal quality of relation graph is improved. JLSMKC has great advantages in clustering performance.
    Adaptive density peaks clustering algorithm
    WU Bin, LU Hongli, JIANG Huijun
    2020, 40(6):  1654-1661.  DOI: 10.11772/j.issn.1001-9081.2019111881
    Asbtract ( )   PDF (864KB) ( )  
    References | Related Articles | Metrics
    Density Peaks Clustering (DPC) algorithm is a new clustering algorithm with the advantages such as few adjustment parameters, no iterative solution and the capacity of finding non-spherical clusters. However, there are some disadvantages of the algorithm: the cutoff distance cannot be adjusted automatically, and the cluster centers need to be selected manually. For the above problems, an Adaptive DPC (ADPC) algorithm was proposed, the adjustment of adaptive cutoff distance based on Gini coefficient was realized, and an automatic acquisition strategy of clustering centers was established. Firstly, the calculation formula of cluster center weight was redefined by taking local density and relative distance into account at the same time. Then, the adjustment method of adaptive cutoff distance was established based on Gini coefficient. Finally, according to the decision graph and cluster center weight sort graph, the strategy of automatically selecting cluster centers was proposed. The simulation results show that, the ADPC algorithm can automatically adjust the cutoff distance and automatically acquire the clustering centers according to the characteristics of problem, and obtain better results than several commonly clustering algorithms and improved DPC algorithms on the test datasets.
    Over-sampling algorithm for imbalanced datasets
    CUI Xin, XU Hua, SU Chen
    2020, 40(6):  1662-1667.  DOI: 10.11772/j.issn.1001-9081.2019101817
    Asbtract ( )   PDF (749KB) ( )  
    References | Related Articles | Metrics

    In Synthetic Minority Over-sampling TEchnique (SMOTE), noise samples may participate in the synthesis of new samples, so it is difficult to guarantee the rationality of the new samples. Aiming at this problem, combining clustering algorithm, an improved algorithm called Clustered Synthetic Minority Over-sampling TEchnique (CSMOTE) was proposed. In the algorithm, the idea of the linear interpolation between the nearest neighbors was abandoned, and the linear interpolation between the cluster centers of minority classes and the samples of corresponding clusters was used to synthesize new samples. And the samples involved in the synthesis were screened to reduce the possibility of noise samples participating in the synthesis. On six actual datasets, CSMOTE algorithm was compared with four SMOTE’s improved algorithms and two under-sampling algorithms for many times, and CSMOTE algorithm obtained the highest AUC values on all datasets. Experimental results show that CSMOTE algorithm has higher classification performance and can effectively solve the problem of unbalanced sample distribution in the datasets.

    Cyber security
    Secure energy transaction scheme based on alliance blockchain
    LONG Yangyang, CHEN Yuling, XIN Yang, DOU Hui
    2020, 40(6):  1668-1673.  DOI: 10.11772/j.issn.1001-9081.2019101784
    Asbtract ( )   PDF (763KB) ( )  
    References | Related Articles | Metrics
    Blockchain technology is widely used in vehicular network, energy internet, smart grid, etc., but attackers can combine social engineering and data mining algorithms to obtain users’ privacy data recorded in the blockchain; especially in microgrid, data generated by games between neighboring energy nodes are more likely to leak user privacy. In order to solve such a problem, based on the alliance blockchain, a secure energy transaction model with a one-to-many energy node account matching mechanism was proposed. The proposed model mainly uses the generation of new accounts to prevent attackers from using data mining algorithms to obtain private data such as energy node accounts, geographical locations, and energy usage from transaction records. The simulation experiment combines the characteristics of the alliance chain, the number of new accounts generated by energy nodes, and the change of transaction verification time to give the analysis results of privacy protection performance, transaction efficiency, and security efficiency. The experimental results show that, the proposed model requires less time during the stage of transaction initiation and verification, has higher security, and the model can hide the transaction trend between adjacent users. The proposed scheme can be well applied to the energy internet transaction scenario.
    Access control model based on blockchain and user credit
    WANG Haiyong, PAN Qiqing, GUO Kaixuan
    2020, 40(6):  1674-1679.  DOI: 10.11772/j.issn.1001-9081.2019101780
    Asbtract ( )   PDF (703KB) ( )  
    References | Related Articles | Metrics

    Focusing on the problem that user privileges cannot dynamically change with time in the current access control and the security problems in the access control contract, an access control model based on Role-Based Access Control (RBAC) model, blockchain and user credit was proposed. Firstly, the roles were distributed to relevant users by the role publishing organization, and the access control strategy was stored in the blockchain through smart contract method. In the contract, the access credit threshold was set, and the contract information was verifiable, traceable and tamper-proof to any service provider organization in the system. Secondly, the final credit was evaluated by the model according to current credit, historical credit and recommended credit of the user, and the access privileges of the corresponding role was obtained based on the final credit. Finally, when the user credit reached the credit threshold set in the contract, the user can access the corresponding service organization. Experimental results show that the proposed model has certain fine granularity, dynamicity and security in the security access control.

    Wireless sensor network intrusion detection system based on sequence model
    CHENG Xiaohui, NIU Tong, WANG Yanjun
    2020, 40(6):  1680-1684.  DOI: 10.11772/j.issn.1001-9081.2019111948
    Asbtract ( )   PDF (656KB) ( )  
    References | Related Articles | Metrics
    With the rapid development of Internet of Things (IoT), more and more IoT node devices are deployed, but the accompanying security problem cannot be ignored. Node devices at the network layer of IoT mainly communicate through wireless sensor networks. Compared with the Internet, they are more open and more vulnerable to network attacks such as denial of service. Aiming at the network layer security problem faced by wireless sensor networks, a network intrusion detection system based on sequence model was proposed to detect and alarm the network layer intrusion, which achieved higher recognition rate and lower false positive rate. Besides, aiming at the security problem of the node host device faced by wireless sensor network node devices, with the consideration of the node overhead, a host intrusion detection system based on simple sequence model was proposed. The experimental results show that, the two intrusion detection systems for the network layer and the host layer of wireless sensor network both have the accuracy more than 99%, and the false detection rate about 1%, which meet the industrial requirements. These two proposed systems can comprehensively and effectively protect the wireless sensor network security.
    Website fingerprinting technique based on image texture
    ZHANG Daowei, DUAN Haixin
    2020, 40(6):  1685-1691.  DOI: 10.11772/j.issn.1001-9081.2019111981
    Asbtract ( )   PDF (1070KB) ( )  
    References | Related Articles | Metrics
    Website fingerprinting technique enables the local monitor to track which websites a user is visiting by capturing anonymous traffic between that user and the Tor (The onion router) entry nodes. Prior researches only extract part meta-data in the anonymous traffic to construct website fingerprints, and ignore much hidden fingerprint information inside the traffic. Therefore, a website fingerprinting technique named Image FingerPrinting (Image-FP) and based on deep convolutional neural network and image texture was proposed. Firstly, the anonymous communication traffic was mapped into Red-Green-Blue (RGB) images. Then, the Residual Network (ResNet) was used to construct the website fingerprinting model with automatic feature learning ability. In a closed-world scenario of 50 websites, Image-FP obtained classification accuracy of 97.2%, which is 0.4 percentage points higher than that of the state-of-the-art website fingerprinting attack technique. In the open-world scenario which is more realistic, Image-FP can identify the traffic of monitored websites with 100% accuracy, has the strongest accuracy and robustness among all fingerprinting techniques. The experimental results demonstrate that, the technique of converting anonymous traffic into images can preserve more features relevant to the website fingerprints, and further improve the classification accuracy while avoiding complex feature engineering
    Differential private average publishing of numerical stream data for wearable devices
    TU Zixuan, LIU Shubo, XIONG Xingxing, ZHAO Jing, CAI Zhaohui
    2020, 40(6):  1692-1697.  DOI: 10.11772/j.issn.1001-9081.2019111929
    Asbtract ( )   PDF (709KB) ( )  
    References | Related Articles | Metrics
    User health data such as heart rate and blood glucose generated by wearable devices in real time is of great significance for health monitoring and disease diagnosis. However, health data is private information of users. In order to publish the average value of numerical stream data for wearable devices and prevent the leakage of users’ privacy information, a new differential private average publishing method of wearable devices based on adaptive sampling was proposed. Firstly, the global sensitivity was introduced which was adaptive to the characteristic of small fluctuation of stream data average for wearable devices. Then, the privacy budget was allocated by the adaptive sampling based on Kalman filter error adjustment, so as to improve the availability of the published data. In the experiments of two kinds of health data publishing, while the privacy budget is 0.1, which means that the level of privacy protection is high, the Mean Relative Errors (MRE) of the proposed method on the heart rate dataset and blood glucose dataset are only 0.01 and 0.08, which are 36% and 33% lower than those of Filtering and Adaptive Sampling for differential private Time-series monitoring (FAST) algorithm. The proposed method can improve the usability of wearable devices’ stream data publishing.
    Advanced computing
    Discrete controller synthesis based resource management method of heterogeneous multi-core processor system
    AN Xin, XIA Jinwei, YANG Haijiao, OUYANG Yiming, REN Fuji
    2020, 40(6):  1698-1706.  DOI: 10.11772/j.issn.1001-9081.2019101865
    Asbtract ( )   PDF (905KB) ( )  
    References | Related Articles | Metrics
    Nowadays, with the development of semiconductor technology and the requirement of the diversification of applications, heterogeneous multi-core processors have been widely used in high-performance embedded systems. How to manage and distribute the available resources (such as processing cores) during running in order to meet the requirements in performance and power consumption of the system and the applications that the system runs is a main challenge that the system focuses. However, although some mainstream resource management techniques have achieved good results in terms of performance and/or power consumption optimization, they lack the strict reliability guarantee for the resource management component. Therefore, a method based on Discrete Controller Synthesis (DCS) was proposed to automatically and reliably design the online resource management scheme for heterogeneous multi-core systems, which applies DCS (which is formal and can construct management control components automatically) to the design of online resource management components for heterogeneous multi-core systems. In this method, the heterogeneous system’s running behaviors (such as how to distribute the processing cores to the applications) were described by using the formal models, and the online resource management problem was transformed to a DCS problem aiming at one system management objective (such as maximizing system performance). On this basis, the existing DCS tools were used to illustrate and validate the proposed method, and the scalability of the DCS method was evaluated.
    Greedy algorithm optimization based virtual machine selection strategy in cloud data center
    CAI Hao, YUAN Zhengdao
    2020, 40(6):  1707-1713.  DOI: 10.11772/j.issn.1001-9081.2019111988
    Asbtract ( )   PDF (575KB) ( )  
    References | Related Articles | Metrics
    In the virtual machine migration process, one of the most problems is how to select the candidate migrating virtual machine list from the abnormal physical hosts in cloud data center. Therefore, a Greedy Algorithm Optimization based Virtual Machine Selection algorithm (GAO-VMS) was proposed. In GAO-VMS, the virtual machines with the optimal objective functions would be selected to perform the migration and the candidate migration virtual machine list was formed subsequently. There are three kinds of greedy modes in GAO-VMS: Maximum Power Reduction Policy (MPR), minimum migration Time and Power Tradeoff policy (TPT) and Violated million instructions per second-Virtual Machines policy (VVM). GAO-VMS was evaluated on CloudSim simulator. Simulation results show that compared to the common virtual machine migration strategy, GAO-VMS reduces the energy consumption of cloud data center by 30%-35%, and reduces the number of virtual machine migrations by 40%-45% with 5% increment of the Service Level Agreement (SLA) violation rate and the joint index of SLA violation and energy. The proposed GAO-VMS strategy can be used for enterprises to construct green cloud computing center.
    Time-aware QoS prediction for SOA-based remote sensing image processing platform
    XU Jinrong, GUO Caiping, TONG Endong
    2020, 40(6):  1714-1721.  DOI: 10.11772/j.issn.1001-9081.2019101772
    Asbtract ( )   PDF (819KB) ( )  
    References | Related Articles | Metrics
    With the help of Service-Oriented Architecture (SOA), the remote sensing image processing algorithms can be abstracted into a set of component services. Then, flexible business requirements can be met through service selection and composition. In order to get the service components that meet the user’s Quality of Service (QoS) requirements for combination, QoS of all services should be firstly obtained. Indeed, the QoS of a service is unknown to its users who have never invoked the service before. Hence, many research work have been proposed to predict the missing QoS. However, these existing methods seldom take the temporal factors into consideration, which may decrease the prediction accuracy. In order to resolve the issue, a new QoS model based on time slice was firstly proposed by considering temporal factors. Furthermore, a time-aware QoS prediction method based on Collaborative Filtering (CF) was proposed. The experiments results on the WS-DREAM real dataset show that, the proposed time-aware QoS prediction method can obtain smaller Mean Square Error (MAE) and Root Mean Square Error (RMSE).In addition, some parameters may affect the time-aware QoS prediction performance. Thus, a set of experiments and analysis with various parameter combinations were carried out, which provides a certain reference for parameter selection.
    Adaptive most valuable player algorithm considering multiple training methods
    WANG Ning, LIU Yong
    2020, 40(6):  1722-1730.  DOI: 10.11772/j.issn.1001-9081.2019101815
    Asbtract ( )   PDF (687KB) ( )  
    References | Related Articles | Metrics
    The Most Valuable Player Algorithm (MVPA) is a new intelligent optimization algorithm that simulates sports competitions. It has the problems of low precision and slow convergence. An adaptive most valuable player algorithm considering multiple training methods (ACMTM-MVPA) was proposed to solve these problems. MVPA has a single initialization method, which is random and blind, reducing the convergence speed and accuracy of the algorithm. In order to enhance the level of the initial player and improve the overall strength of the initial team, the training phase was added before the competition phase of MVPA, and the neighborhood search algorithm and chaotic sequence and reverse learning algorithms were used to train and screen players; in order to enhance the player’s ability to self-explore and learn from the best player to make the player have the qualification to compete for the most valuable player trophy, an adaptive player evolution factor was added during the team competition phase. Experimental results on 15 benchmark functions show that the proposed algorithm outperforms MVPA, Particle Swarm Optimization (PSO) algorithm and Genetic Algorithm (GA) in optimization accuracy and convergence speed. Finally, an application example of ACMTM-MVPA in parameter optimization of storm intensity formula was given. The results show that this proposed algorithm is superior to accelerated genetic algorithm, traditional regression method and preferred regression method.
    Emotional bacterial foraging algorithm based on non-uniform elimination-diffusion probability distribution
    DONG Hai, QI Xinna
    2020, 40(6):  1731-1737.  DOI: 10.11772/j.issn.1001-9081.2019101725
    Asbtract ( )   PDF (632KB) ( )  
    References | Related Articles | Metrics
    In view of the uncertainty of chemotaxis step length and the lack of constancy of elimination diffusion probability in the optimization process of traditional bacterial foraging algorithm, in order to solve the problem of high-dimensional engineering optimization, an emotional bacteria foraging algorithm based on non-uniform elimination-diffusion probability was proposed. Firstly, in the chemotaxis step, the Gus distribution search mechanism was used to update the bacteria individual positions, so as to solve the problem of poor search ability and easy to fall into local optimum caused by bacteria swimming or flipping on each dimension in a random way. The emotion perception factor was introduced, and the sudden change of emotional intelligence was used to realize the adaptive chemotaxis step size, so as to avoid premature convergence of the algorithm. Secondly, in view of the probability constancy of bacterial individuals in the process of elimination-diffusion, the idea of using linear and non-linear probability distributions to replace the traditional constant distribution to realize non-uniform distribution was proposed. By introducing the random value of dynamic factor, the bacterial individuals in the undefined search space were limited, so as to save the calculation cost of the algorithm. Six benchmark functions were used in the test, and the test results show that: in the case of low calculation cost, except on Rosenbrock function, the proposed algorithm has low iteration times and good optimization quality on all functions, and the algorithm convergence comparison results show that the proposed algorithm has good convergence.
    Network and communications
    Vehicular ad-hoc network greedy routing algorithm based on path exploration
    TANG Xingfeng, XU Qingqin, MA Shiwei
    2020, 40(6):  1738-1744.  DOI: 10.11772/j.issn.1001-9081.2019101832
    Asbtract ( )   PDF (867KB) ( )  
    References | Related Articles | Metrics
    In order to improve the transmission efficiency of information between vehicles in the city and realize the information sharing between vehicles, aiming at the problem that the current multi-hop unicast routing algorithm based on geographical location forwarding in the Vehicular Ad-hoc NETwork (VANET) does not consider the specificity of the urban scene and cannot adapt to the high dynamicity of vehicles in the city, so that the data packets between vehicles may spread on the wrong path, resulting in high packet loss rate and long delay, a new greedy routing algorithm based on path exploration was proposed. Firstly, taken the data packet transmission delay as the standard, artificial bee colony algorithm was used to explore multiple routing paths planned by the digital map. Then, the multi-hop forwarding method of data packets between vehicles was optimized. Simulation results show that, compared with Greedy Perimeter Stateless Routing (GPSR) protocol and Maxduration-Minangle GPSR (MM-GPSR) improved algorithm, in the best case, the data packet arrival rate of the proposed algorithm increases by 13.81% and 9.64% respectively, and the average data packet end-to-end delay of the proposed algorithm decreases by 61.91% and 27.28% respectively.
    Multi-objective path planning algorithm for mobile charging devices jointing wireless charging and data collection
    HAN Yulao, FANG Dingyi
    2020, 40(6):  1745-1750.  DOI: 10.11772/j.issn.1001-9081.2019111933
    Asbtract ( )   PDF (594KB) ( )  
    References | Related Articles | Metrics
    The limited resources of wireless sensor network nodes cause the poor completeness and timeliness of data collection. To solve these problems, a multi-objective path planning model for Mobile Charging Devices (MCD) jointing mobile charging and data collection was established, and a Path Planning algorithm based on Greedy Strategy for MCD jointing wireless charging and data collection (PPGS) was proposed. Firstly, the monitoring area was divided into many seamless regular hexagon cells, so as to effectively reduce the number of cells visited by MCD. Then, the parameters such as the node energy and the quantity of data collection were predicted by using the Markov model, and the anchor minimum stopping time and anchor maximum waiting time for MCD were predicted based on the above. Compared with the existing Delay-Constrained Mobile Energy Charging algorithm (DCMEC) and Mobile Device Scheduling Algorithm and Grid-Based Algorithm (GBA+MDSA), the proposed algorithm has lower complexity and does not need to know the actual location information of nodes and anchors in advance. The simulation results show that, the proposed PPGS can guarantee the completeness and timeliness of data collection with a small number of MCD in wireless sensor network.
    Verification of control-data plane consistency in software defined network
    ZHU Mengdi, SHU Yong’an
    2020, 40(6):  1751-1754.  DOI: 10.11772/j.issn.1001-9081.2019101712
    Asbtract ( )   PDF (497KB) ( )  
    References | Related Articles | Metrics
    Aiming at the problem of inconsistency between the network policies of control layer and flow rules of data layer in Software Defined Network (SDN), a detection model for Verifying control-data plane Consistency (VeriC) was proposed. Firstly, the function of the packet processing subsystem was realized through the VeriC pipeline on the switch, and the function is sampling the data packet, and updating the tag field in the sampled data packet when the packet passing through the switch. Then, after the update was completed, the tag values were sent to the server and stored in the real tag value group. Finally, the real tag value group and the stored correct tag value group were sent to the verification subsystem to perform the consistency verification. As it failed, the two groups of tag values were sent to the localization subsystem to locate the switch with flow table entry error. A fat tree topology with 4 Pod was generated by ns-3 simulator, where the accuracies of consistency detection and faulty machine location of VeriC are higher than those of VeriDP, and the overall performance of VeriC is higher than that of 2MVeri model. Theoretical analysis and simulation results show that VeriC detection model can not only perform consistency detection and accurately locate the faulty switch, but also take shorter time to locate the faulty switch compared to other comparison detection models.
    Adaptive UWB/PDR fusion positioning algorithm based on error prediction
    ZHANG Jianming, SHI Yuanhao, XU Zhengyi, WEI Jianming
    2020, 40(6):  1755-1762.  DOI: 10.11772/j.issn.1001-9081.2019101830
    Asbtract ( )   PDF (1311KB) ( )  
    References | Related Articles | Metrics
    An Ultra WideBand (UWB)/ Pedestrian Dead Reckoning (PDR) fusion positioning algorithm with adaptive coefficient adjustment based on UWB error prediction was proposed in order to improve the UWB performance and reduce the PDR accumulative errors in the indoor Non-Line-Of-Sight (NLOS) positioning scenes and solve the UWB performance degradation caused by environmental factors. On the basis of the creative proposal of predicting the UWB positioning errors in complex environment by Support Vector Machine (SVM) regression model, UWB/PDR fusion positioning performance was improved by adding adaptive adjusted parameters to the conventional Extended Kalman Filter (EKF) algorithm. The experimental results show that the proposed algorithm can effectively predict the current UWB positioning errors in the complex UWB environment, and increase the accuracy by adaptively adjusting the fusion parameters, which makes the positioning error reduced by 18.2% in general areas and reduced by 48.7% in the areas with poor UWB accuracy compared with those of the conventional EKF algorithm, so as to decrease the environmental impact on the UWB performance. In complex scenes of both Line-Of-Sight (LOS) and NLOS including UWB, the positioning error per 100 meters is reduced from meter scale to decimeter scale, which reduces the PDR errors in NLOS scenes.
    Low SNR denoising algorithm based on adaptive voice activity detection and minimum mean-square error log-spectral amplitude estimation
    ZHANG Haoran, WANG Xueyuan, LI Xiaoxia
    2020, 40(6):  1763-1768.  DOI: 10.11772/j.issn.1001-9081.2019111880
    Asbtract ( )   PDF (2132KB) ( )  
    References | Related Articles | Metrics
    Aiming at the limitations of traditional noise reduction methods for acoustic signals in low Signal-to-Noise Ratio (SNR) environment, a real-time noise reduction algorithm was proposed by combining adaptive threshold Voice Activity Detection (VAD) algorithm and Minimum Mean-Square Error Log-Spectral Amplitude estimation (MMSE-LSA). Firstly, the background noise was estimated in VAD algorithm by probability statistics based on the maximum value of the energy probability, and the obtained background noise was updated in real time and saved. Then, the background noise updated in real time was used as the reference noise of MMSE-LSA, and the noise amplitude spectrum was updated adaptively. Finally, the noise reduction processing was performed. The experimental results on four kinds of acoustic signals in real scenes show that the proposed algorithm can guarantee the real-time processing of low SNR acoustic signals; and compared with the traditional MMSE-LSA algorithm, it has the SNR of the noise reduction signal increased by 10-15 dB without over-subtraction. It can be applied to practical engineering.
    Estimation of underdetermined mixing matrix based on improved weighted fuzzy C-means clustering
    SUN Jianjun, XU Yan
    2020, 40(6):  1769-1773.  DOI: 10.11772/j.issn.1001-9081.2019111882
    Asbtract ( )   PDF (1377KB) ( )  
    References | Related Articles | Metrics
    The Fuzzy C-Means clustering (FCM) algorithm has the defects of being sensitive to initial clustering center,being susceptible to noise point interference and poor robustness in solving the problem of speech underdetermined mixing matrix estimation. An improved WEighted FCM algorithm based on evolutionary programming (WE-FCM) was proposed to eliminate the defects. Firstly, the powerful search ability of Evolutionary Programming (EP) algorithm was used to optimize FCM for obtaining FCM algorithm based on EP (EP-FCM), in order to obtain a better initial clustering center. Then, the Local Outlier Factor (LOF) algorithm was used to perform weighting to reduce the effects of noise points. The simulation experiment results show that, the normalized mean square error value and the deviation angle value of the proposed algorithm were both much smaller than those of the classical K-means clustering, K-Hough, FCM algorithm based on Genetic Algorithm (GAFCM) and FCM algorithm based on Find Density Peaks (FDP-FCM) when the number of source signals were 3 and 4. The experimental results show that, the proposed algorithm significantly improves the robustness of FCM algorithm and the accuracy of mixing matrix estimation.
    Virtual reality and multimedia computing
    Object tracking algorithm based on correlation filtering and color probability model
    ZHANG Jie, CHANG Tianqing, DAI Wenjun, GUO Libin, ZHANG Lei
    2020, 40(6):  1774-1782.  DOI: 10.11772/j.issn.1001-9081.2019112001
    Asbtract ( )   PDF (3751KB) ( )  
    References | Related Articles | Metrics
    In order to solve the interference of similar background to object tracker in ground battlefield environment, an object tracking algorithm combining correlation filtering and improved color probability model was proposed. Firstly, based on the traditional color probability model, a color probability model emphasizing foreground was proposed by using the difference between foreground object histogram and background histogram. Then, a spatial penalty matrix was generated according to the correlation filter response confidence and maximum response position. This matrix was used to punish the likelihood probability of background pixel determined by the correlation filter, and the response map of the color probability model was obtained by using the method of integral image. Finally, the response maps obtained by the correlation filter and the color probability model were fused, and the maximum response position of the fusion response map was the central position of the object. The proposed algorithm was compared with 5 state-of-the-art algorithms such as Circulant Structure of tracking-by-detection filters with Kernels (CSK), Kernelized Correlation Filters (KCF), Discriminative Scale Space Tracking (DSST), Scale Adaptive Multiple Feature (SAMF) and Staple in tracking performance. The experimental results on OTB-100 standard dataset show that, the proposed algorithm has the overall accuracy improved by 3.06% to 55.98%, and the success rate improved by 2.24% to 54.97%; and under similar background interference, the proposed algorithm has the accuracy improved by 10.28% to 43.9%, and the success rate improved by 8.3% to 48.29%. The experimental results on 36 battlefield video sequences show that, the proposed algorithm has the overall accuracy improved by 2.2% to 45.98%, and the success rate improved by 3.01% to 58.27%. It can be seen that the proposed algorithm can better deal with the interference of similar background in the ground battlefield environment, and provide more accurate position information for the weapon platform.
    Adaptive color mapping and its application in void evolution visualization
    QIAO Jiewen, CHEN Wei
    2020, 40(6):  1783-1792.  DOI: 10.11772/j.issn.1001-9081.2019111889
    Asbtract ( )   PDF (3650KB) ( )  
    References | Related Articles | Metrics
    In order to improve the visualization effect of the void evolution of materials, an adaptive color mapping method based on data characteristics was proposed. Firstly, a number of control points were selected in the CIELAB color space to form an initial color path. Then, based on the proportion of the data characteristic values, the positions of control points were optimized and the color path was adjusted according to the constraints such as uniformity of color difference and consistency of brightness, so as to meet the requirement of control points following the data adaptively. Finally, the distribution of the perceptual difference sum was remapped by the equalization algorithm, and the perceptual uniformity of the color mapping was optimized to form the final color map. The experimental results show that, compared with traditional color mapping methods which only consider the color space and ignore the diversity of data, the proposed adaptive color mapping method has better identifiability of the characteristics of visualization results by fully considering the color proportion, the number of control points and self-adaptation, and guarantees the perceptual uniformity of the visualization results of the void evolution, improving the accuracy of visualization results and reduces the time required to observe effective information.
    High order TV image reconstruction algorithm based on Chambolle-Pock algorithm framework
    XI Yarui, QIAO Zhiwei, WEN Jing, ZHANG Yanjiao, YANG Wenjing, YAN Huiwen
    2020, 40(6):  1793-1798.  DOI: 10.11772/j.issn.1001-9081.2019111955
    Asbtract ( )   PDF (720KB) ( )  
    References | Related Articles | Metrics
    The traditional Total Variation (TV) minimization algorithm is a classical iterative reconstruction algorithm based on Compressed Sensing (CS), and can accurately reconstruct images from sparse and noisy data. However, the block artifacts may be brought by the algorithm during the reconstruction of image having not obvious piecewise constant feature. Researches show that the use of High Order Total Variation (HOTV) in the image denoising can effectively suppress the block artifacts brought by the TV model. Therefore, a HOTV image reconstruction model and its Chambolle-Pock (CP) solving algorithm were proposed. Specifically, the second order TV norm was constructed by using the second order gradient, then a data fidelity constrained second order TV minimization model was designed, and the corresponding CP algorithm was derived. The Shepp-Logan phantom in wave background, grayscale gradual changing phantom and real CT phantom were used to perform the image reconstruction experiments and qualitative and quantitative analysis under ideal data projection and noisy data projection conditions. The reconstruction results of ideal data projection show that compared to the traditional TV algorithm, the HOTV algorithm can effectively suppress the block artifacts and improve the reconstruction accuracy. The reconstruction results of noisy data projection show that both the traditional TV algorithm and the HOTV algorithm have good denoising effect but the HOTV algorithm is able to protect the image edge information better and has higher anti-noise performance. The HOTV algorithm is a better reconstruction algorithm than the TV algorithm in the reconstruction of image having not obvious piecewise constant feature and obvious grayscale fluctuation feature. The proposed HOTV algorithm can be extended to CT reconstruction under different scanning modes and other imaging modalities.
    Deformable medical image registration algorithm based on deep convolution feature optical flow
    ZHANG Jiagang, LI Daping, YANG Xiaodong, ZOU Maoyang, WU Xi, HU Jinrong
    2020, 40(6):  1799-1805.  DOI: 10.11772/j.issn.1001-9081.2019101839
    Asbtract ( )   PDF (1420KB) ( )  
    References | Related Articles | Metrics
    Optical flow method is an important and effective deformation registration algorithm based on optical flow field model. Aiming at the problem that the feature quality used by the existing optical flow method is not high enough to make the registration result accurate, combining the features of deep convolutional neural network and optical flow method, a deformable medical image registration algorithm based on Deep Convolution Feature Based Optical Flow (DCFOF) was proposed. Firstly, the deep convolution feature of the image block where each pixel in the image was located was densely extracted by using a deep convolutional neural network, and then the optical flow field was solved based on the deep convolution feature difference between the fixed image and the floating image. By extracting more accurate and robust deep learning features of the image, the optical flow field obtained was closer to the real deformation field, and the registration accuracy was improved. Experimental results show that the proposed algorithm can solve the problem of deformable medical image registration effectively, and has the registration accuracy better than those of Demons algorithm, Scale-Invariant Feature Transform(SIFT) Flow algorithm and professional registration software of medical images called Elastix.
    Lightweight human skeleton key point detection model based on improved convolutional pose machines and SqueezeNet
    QIANG Baohua, ZHAI Yijie, CHEN Jinlong, XIE Wu, ZHENG Hong, WANG Xuewen, ZHANG Shihao
    2020, 40(6):  1806-1811.  DOI: 10.11772/j.issn.1001-9081.2019101866
    Asbtract ( )   PDF (1242KB) ( )  
    References | Related Articles | Metrics
    In order to solve the problems of too many parameters, long training time and slow detection speed of the existing human skeleton key point detection models, a detection method combining the human skeleton key point detection model called Convolutional Pose Machines (CPMs) and the lightweight convolutional neural network model called SqueezeNet was proposed. Firstly, the CPMs with 4 stages (CPMs-Stage4) was used to detect the key points of the human images. Then, the Fire Module network structure of SqueezeNet was introduced into CPMs-Stage4 to reduce the model parameters greatly, and thus to obtain a new lightweight human skeleton key point detection model called SqueezeNet15-CPMs-Stage4. The verification results on the extended Leeds Sports Pose (LSP) dataset show that, compared with CPMs, SqueezeNet15-CPMs-Stage4 model has the training time reduced by 86.68%, the detection time of single image reduced by 44.27%, and the detection accuracy of 90.4%; and the proposed model performs the best in training time, detection speed and accuracy compared with three reference models improved VGG-16, DeepCut and DeeperCut. The experimental results show that the proposed model achieves high detection accuracy with short training time and fast detection speed, and can effectively reduce the training cost of the human skeleton key point detection model.
    lmage vignetting correction based on constrained log-intensity entropy under low-pass filtering
    ZHOU Siyu, BAO Guoqi, LIU Kai
    2020, 40(6):  1812-1817.  DOI: 10.11772/j.issn.1001-9081.2019101809
    Asbtract ( )   PDF (6022KB) ( )  
    References | Related Articles | Metrics
    Vignetting is the phenomenon that the intensity of the pixel in the image decreases along the radial direction. In order to solve the problem that it affects the accuracy of computer vision task and image processing, a method of single image vignetting correction based on constrained log-intensity entropy under low-pass filtering was proposed. Firstly, the vignetting model was established by using a sixth order polynomial function of even term. Secondly, the minimum log-intensity entropy of the target image was calculated by low-pass filtering. Under the constraint of the target value, the optimal parameter solution of the vignetting model was obtained, which can satisfy the change rule of the vignetting function and reduce the log-intensity entropy of the image. Finally, vignetting was eliminated by using inverse compensation of vignetting model. Vignetting correction results were evaluated by Structural SIMilarity index (SSIM) and Root Mean Square Error (RMSE). Experimental results show that the proposed method can not only effectively recover the brightness information of the vignetting area to get real and natural non-vignetting image, but also effectively correct the different degrees of vignetting with a good visual consistency.
    Painter artistic style extraction method based on color features
    WANG Qirong, HUANG Zhangcan
    2020, 40(6):  1818-1823.  DOI: 10.11772/j.issn.1001-9081.2019111886
    Asbtract ( )   PDF (998KB) ( )  
    References | Related Articles | Metrics
    Since the ineffectiveness of color features extracted by global and local feature extraction methods to describe the artistic style of painter, a new oil painting description method based on key region was proposed. Firstly, the information richness of oil painting region was defined by incorporating the proportion of primary color and color diversity to detect and select the key region of an oil painting. Secondly, the color features in 54 dimensions of the selected key region were used to describe the oil painting, those features were evaluated by Fisher Score, and the important features were selected to describe the key region of the oil painting, so as to highly reflect the painter artistic style. Finally, to verify the validity of the proposed method, the Naive Bayes classifier was used to realize oil painting classification. Two databases were established to perform simulation experiments. The database 1 includes 120 oil paintings by three painters, and the database 2 includes 513 oil paintings by nine painters from three different schools. The experimental results on database 1 show that, the accuracy of classification combined with Fisher Score is higher than the accuracy of ordinary classification, the accuracy of the proposed method for classifying oil paintings according to painter and school is 90% and 90.20% respectively. The experimental results on database 2 show that the accuracy of the proposed method for classifying oil paintings according to painter reaches 90%, which is 6.67 percentage points higher than that of Feature selected by Fisher Score(Features-FS), and the accuracy of the proposed method for classifying oil paintings according to school is 90.20%, which is comparable to that of Features-FS. The features extracted by the proposed oil painting description method based on key region can effectively describe the artistic style of painter.
    Real-time segmentation algorithm for bubble defects of plastic bottle based on improved Fast-SCNN
    FU Lei, REN Dejun, WU Huayun, GAO Ming, QIU Lyu, HU Yunqi
    2020, 40(6):  1824-1829.  DOI: 10.11772/j.issn.1001-9081.2019111926
    Asbtract ( )   PDF (756KB) ( )  
    References | Related Articles | Metrics
    When the bubbles of medical plastic bottles are detected, the arbitrariness of the bubble position in the bottle body, the uncertainty of the bubble size, and the similarity between the bubble characteristics and the bottle body characteristics increase the difficulty of detecting the bubble defects. In order to solve the above problems in the detection of bubble defects, a real-time segmentation algorithm based on improved Fast Segmentation Convolutional Neural Network (Fast-SCNN) was proposed. The basic framework of the segmentation algorithm is the Fast-SCNN. In order to make up for the lack of robustness of the original network segmentation scale, the ideas of the usage of the information between the channels of Squeeze-and-Excitation Networks (SENet) and the multi-level skip connection were adopted. Specifically, the deep features were extracted by further down-sampling of the network, the up-sampling operation was merged with SELayer module in the decoding stage, and the skip connections with the shallow layer of the network were increased two times at the same time. Four sets of experiments were designed for comparison on the bubble dataset with the Mean Intersection over Union (MIoU) and the segmentation time for single image of the algorithm used as evaluation indicators. The experimental results show that the comprehensive performance of the improved Fast-SCNN is the best, this network has the MIoU of 97.08%, the average segmentation time for a medical plastic bottle of 24.4 ms, and the boundary segmentation accuracy 2.3% higher than Fast-SCNN, which improves the segmentation ability of tiny bubbles, and this network has the MIoU improved by 0.27% and the time reduced by 7.5 ms compared to U-Net, and the comprehensive detection performance far better than Fully Convolutional Networks (FCN-8s). The proposed algorithm can effectively segment smaller bubbles with unclear edges and meet the engineering requirements for real-time segmentation and detection of bubble defects.
    Peptide spectrum match scoring algorithm based on multi-head attention mechanism and residual neural network
    MIN Xin, WANG Haipeng, MOU Changning
    2020, 40(6):  1830-1836.  DOI: 10.11772/j.issn.1001-9081.2019101880
    Asbtract ( )   PDF (1141KB) ( )  
    References | Related Articles | Metrics
    Peptide spectrum match scoring algorithm plays a key role in the peptide sequence identification, and the traditional scoring algorithm cannot effectively make full use of the peptide fragmentation pattern to perform scoring. In order to solve the problem, a multi-classification probability sum scoring algorithm combined with the peptide sequence information representation called deepscore-α was proposed. In this algorithm, the second scoring was not performed with the consideration of global information, and there was no limitation on the similarity calculation method of theoretical mass spectrum and experimental mass spectrum. In the algorithm, a one-dimensional residual network was used to extract the underlying information of the sequence, and then the effects of different peptide bonds on the current peptide bond fracture were integrated through the multi-attention mechanism to generate the final fragmention relative intensity distribution probability matrix, after that, the final peptide spectrum match score was calculated by combining the actual relative intensity of the peptide sequence fragmention. This algorithm was compared with Comet and MSGF+, two common open source identification tools. The results show that when False Discovery Rate (FDR) was 0.01 on humanbody proteome dataset, the number of peptide sequences retained by deepScore-α is increased by about 14%, and the Top1 hit ratio (the proportion of the correct peptide sequences in the spectrum with the highest score) of this algorithm is increased by about 5 percentage points. The generalization performance test of the model trained by human ProteomeTools2 dataset show that the number of sequences peptide retained by deepScore-α at FDR of 0.01 is improved by about 7%, the Top1 hit ratio of this algorithm is increased by about 5 percentage points, and the identification results from Decoy library in the Top1 is decreased by about 60%. Experimental results prove that, the algorithm can retain more peptide sequences at lower FDR value, improve the Top1 hit ratio, and has good generalization performance.
    Frontier & interdisciplinary applications
    Auto-registration method of ground based building point clouds based on line features and iterative closest point algorithm
    XU Jingzhong, WANG Jiarong
    2020, 40(6):  1837-1841.  DOI: 10.11772/j.issn.1001-9081.2019111978
    Asbtract ( )   PDF (1133KB) ( )  
    References | Related Articles | Metrics
    To overcome the shortcoming that the Iterative Closest Point (ICP) algorithm is easy to fall into local optimum, an auto-registration method of ground based building point clouds based on line features and ICP algorithm was proposed. Firstly, the plane segmentation was performed on point clouds based on normal consistency. Secondly, the outlines of point clusters were extracted by alpha-shape algorithm, and the feature line segments were obtained by the splitting and fitting process. Then, the feature line pairs were taken as the registration primitives, and the angle and distance between line pairs were used as similarity measures for same-name feature matching in order to achieve the coarse registration of building cloud points. Finally, with the coarse registration result as the initial value, the ICP algorithm was used to realize the fine registration of building point clouds. Two sets of partially overlapping building point clouds were used to carry out the experiments. The experimental results show that the proposed coarse-to-fine registration method can effectively improve the dependency of ICP algorithm on initial value and realize the effective registration of partially overlapping building point clouds.
    Optimization method of incremental split selection based on video queue length management
    WU Yiyuan, LIAN Peikun, GUO Jiangang, LAI Yuanwen, KANG Yaling
    2020, 40(6):  1842-1849.  DOI: 10.11772/j.issn.1001-9081.2019111986
    Asbtract ( )   PDF (1558KB) ( )  
    References | Related Articles | Metrics
    Concerning the phenomenon that the queues in the entrance lanes of the intersections are imbalanced or overflowed during the peak hours, an incremental split selection method based on video queue length management was proposed. Firstly, the queueing state at the end of the red time and the queueing length level at the end of the green time were judged. Then, the increment or decrement of the green time of each phase was calculated. Finally, the dynamic balance between the green time of each phase and the queue length of each entrance lane was realized with the purpose of balancing the queue lengths of the entrance lanes. The experimental results show that, the proposed optimization method can effectively balance the queue lengths of the entrance lanes, reducing the traffic delay and traffic congestion at the intersection. When the split does not match the queue length, the optimization method can quickly adjust the split to adapt to the change of the queue length.
    Optimization method of airport gate assignment based on relaxation algorithm
    XING Zhiwei, QIAO Di, LIU Hong’en, GAO Zhiwei, LUO Xiao, LUO Qian
    2020, 40(6):  1850-1855.  DOI: 10.11772/j.issn.1001-9081.2019111888
    Asbtract ( )   PDF (586KB) ( )  
    References | Related Articles | Metrics
    Aiming at the shortage of the airport gate resources and the disturbance caused by the actual flight arrival and departure time deviation from the planned time, a gate assignment scheduling method was proposed by adding buffer time between the adjacent flights in the same gate. Firstly, a robust gate assignment model with a goal to achieve minimum gate idle time and apron occupancy time was established. Then, a Lagrangian relaxation optimization algorithm based on double targets was designed, and the dual problem in the Lagrangian algorithm was solved by using the subgradient algorithm. Based on the operation data of a hub airport in China, the simulation results show that, compared with those of the original gate assignment scheme, the gate usage amount and the gate idle time of the proposed method is respectively reduced by 15.89% and 7.56%, the gate occupancy rate of the optimization scheme of proposed method is increased by 18.72% and the conflict rate is reduced to 3.57%, proving that the proposed method achieves the purpose of effectively improving the utilization and robustness of airport gates.
    Hierarchical segmentation of pathological images based on self-supervised learning
    WU Chongshu, LIN Lin, XUE Yunjing, SHI Peng
    2020, 40(6):  1856-1862.  DOI: 10.11772/j.issn.1001-9081.2019101863
    Asbtract ( )   PDF (2378KB) ( )  
    References | Related Articles | Metrics
    The uneven distribution of cell staining and the diversity of tissue morphologies bring challenges to the automatic segmentation of Hematoxylin-Eosin (HE) stained pathological images. In order to solve the problem, a three-step hierarchical segmentation method of pathological images based on self-supervised learning was proposed to automatically segment the tissues in the pathological images layer-by-layer from coarse to fine. Firstly, feature selection was performed in the RGB color space based on the calculation result of mutual information. Secondly, the image was initially segmented into stable and fuzzy color regions of each tissue structure based on K-means clustering. Thirdly, the stable color regions were taken as training datasets for further segmentation of fuzzy color regions by naive Bayesian classification, and the three complete tissue structures including nucleus, cytoplasm and extracellular space were obtained. Finally, precise boundaries between nuclei were obtained by performing the mixed watershed classification considering both shape and color intensities to the nucleus part, so as to quantitatively calculate the indicators such as the number of nuclei, nucleus proportion, and nucleus-cytoplasm ratio. Experimental results of HE stained meningioma pathological image segmentation show that, the proposed method is highly robust to the difference of staining and cell morphologies, the error of issue area segmentation is within 5%, and the average accuracy of the proposed method in nucleus segmentation accuracy experiment is above 96%, which means that the proposed method can meet the requirements of automatic analysis of clinical images and its quantitative results can provide references for quantitative pathological analysis.
2024 Vol.44 No.2

Current Issue
Honorary Editor-in-Chief: ZHANG Jingzhong
Editor-in-Chief: XU Zongben
Associate Editor: SHEN Hengtao XIA Zhaohui
Domestic Post Distribution Code: 62-110
Foreign Distribution Code: M4616
No. 9, 4th Section of South Renmin Road, Chengdu 610041, China
Tel: 028-85224283-803
Website: www.joca.cn
E-mail: bjb@joca.cn
Join CCF