Loading...

Table of Content

    14 October 2020, Volume 0 Issue 10
    Root Cause Correlation Analysis of Chemical Enterprise Accident
    CHEN Zhuo, LI Xin, DU Jun-wei, YUAN Xi-ming
    2020, 0(10):  1-6.  doi:10.3969/j.issn.1006-2475.2020.10.001
    Asbtract ( 272 )   PDF (998KB) ( 264 )  
    References | Related Articles | Metrics
    The root causes of chemical accidents are mainly caused by the unsafe behavior of people, the unsafe state of machinery or materials, etc. Their essence is the defect of enterprise management. Mining the relationships between root causes, root causes and accidents is the key to prevent accidents and improve the level of enterprise safety management. Since the existing root cause analysis of accident investigation and the safety management index system are sparsely correlated, it is difficult to mine the correlations between management defects and accident evolution. So, collaborative filtering algorithm is used to fill in the missing score data in the accident investigation. The association rule algorithm based on the weighted support degree is used to mine the strong association rules between the root causes of the accidents, the root cause and the accident attribute. The experimental results show that compared with the existing algorithms, the association analysis algorithm based on weighted support can recommend more high-risk enterprises potential safety hazards and the evolutionary correlations between safety hazards and accidents, so as to scientifically guide the safe production of enterprises, realize the risk warning and accident prevention for the production process.
    A Recommendation Algorithm Based on Text Convolutional Neural Network
    YANG Hui, WANG Yue-hai, DOU Zhen-ze
    2020, 0(10):  7-11. 
    Asbtract ( 428 )   PDF (808KB) ( 422 )  
    References | Related Articles | Metrics
    The traditional matrix factorization model can not effectively extract the features of users and items, while the deep learning model can extract the feature information well. At present, the mainstream recommendation algorithm based on deep learning only uses the output of neural network or the product of item features and user features to make recommendation prediction, which can not fully mine the relationship between users and items. Based on this, this paper proposes a recommendation algorithm based on the combination of text convolutional neural network and bias singular value decomposition (BiasSVD). Text convolutional neural network (TextCNN) is used to fully extract the feature information of users and items, and then singular value decomposition method is used to make recommendations, which can deeply understand the document context information and further improve the accuracy of recommendation. After extensive evaluation and analysis on two real datasets of MovieLens, the recommendation accuracy of this algorithm is obviously better than that of ConvMF algorithm and mainstream deep learning recommendation algorithm.
    Attention and LSTM Based State Prediction of Equipment on Electric Power Communication Networks
    WU Hai-yang, CHEN Peng, GUO Bo, JIANG Chun-xia, LI Ji-xuan, ZHU Peng-yu
    2020, 0(10):  12-16. 
    Asbtract ( 327 )   PDF (713KB) ( 532 )  
    References | Related Articles | Metrics
    With the rapid growth of electric power communication networks, the importance of predicting the working state of online equipment is increasing as well. Since the running data of typical communication devices always come from heterogeneous resources, the prediction models have to be learned from features with high dimension, high sparsity as well as repetitive patterns. This problem severely restricts the performance of conventional machine learning approaches. This paper proposes a novel state prediction model based on the integration of attention mechanism and LSTM (Long Short-Term Memory). By a two-stage learning strategy, the attention mechanism can achieve both dimensionality reduction and feature extraction of original input. Meanwhile, most related features are extracted for final prediction from the end-to-end learning. Extensive experimental results on practical running data of electric power communication networks demonstrate that, the proposed method has high performance in the working state prediction problem.
    Spam E-mail Recognition Based on Cluster Analysis Algorithm
    GAI Xuan
    2020, 0(10):  17-22. 
    Asbtract ( 384 )   PDF (1256KB) ( 309 )  
    References | Related Articles | Metrics
    For spam recognition methods used in the past, in the face of today’s fast updating and a wide variety of word segmentation, it is difficult to accurately identify the key word segmentation in a e-mail, the application ability of the recognition methods needs to be further improved. To this end, a spam recognition method based on cluster analysis algorithm is proposed. Firstly, we preprocess e-mail samples to get the key word segmentation of the e-mail text content, remove the stop words, and calculate the weight of the word segmentation according to the frequency of the word segmentation in the e-mail text. Then, combining with the e-mail feature attributes, we construct the e-mail feature space, and quantify the e-mail feature. Lastly, the features of the e-mail are extracted and processed for dimensionality reduction, which is used as the input of the clustering algorithm, and the output result is iteratively calculated to complete the identification of spam. The experimental results show that the designed spam e-mail recognition method based on cluster analysis algorithm is more accurate in keyword extraction and word segmentation, and can accurately identify spam e-mails, which shows the practical application ability of the designed spam e-mail recognition method based on cluster analysis algorithm has been improved.
    Validity Check of Instructions and Actions of POF Entries
    FENG Dong, CHEN Xiao
    2020, 0(10):  23-30. 
    Asbtract ( 183 )   PDF (1147KB) ( 107 )  
    References | Related Articles | Metrics
    The POF protocol is a southbound interface protocol of SDN. Compared with the classic OpenFlow protocol, POF is protocol-independent and flexible. In recent research on the POF protocol, scholars have added actions and instructions such as comparison and jump to the POF, making each table entry more powerful. However, the increase in the number of instructions and functions has also brought greater instability to the interpretation and execution of entry instructions and actions by software switches. The purpose of this article is to improve the robustness of the system by statically checking for situations that cause software switches to crash in advance. We first analyze the characteristics of the POF matching action table and design a specific detection scheme. Then a control flow graph-based detection algorithm is proposed to find instruction errors, unreachable instructions, and loop blocks in the entries. Further, for a cyclic block, a detection algorithm based on a strongly connected component is proposed to determine the validity of the cyclic block. Experiments on POF switches show that the scheme described in this article can accurately detect common entry errors and provide reliability for software switches. At the same time, the detection is light and fast, so that it can detect flow table entries in real time.
    Development Status and Prospect of Blockchain as a Service
    LU Ge-hao, YANG Dan-ni
    2020, 0(10):  31-35. 
    Asbtract ( 269 )   PDF (915KB) ( 157 )  
    References | Related Articles | Metrics
    This paper discusses the research status of blockchain as a service, introduces the working principle of blockchain as a service through the combination of cloud computing and blockchain technology, discusses the basic design principles according to the service object of blockchain as a service, and analyzes the basic architecture of mainstream blockchain as a service platform. As a kind of cloud service, blockchain as a service is the cloud rental network platform of blockchain service facilities. According to the characteristics of its tenants, the computing resources, platform resources, software resources and hardware resources are shared to the maximum extent. The blockchain as a service ensures that the business scale of tenants can be flexibly scaled through the large-capacity resource pool, and the rented facilities can be shared and enjoyed exclusively, thus ensuring safe and reliable operation. In view of the blockchain as a service risk regulation, technical difficulties, performance and other issues, the current research deficiencies and future research directions are analyzed and prospected.
    Automatic Processing Method for Cache Conflict of  Multi-source Database Based on Rough Set
    LI Xi-min, LI Shu-qi
    2020, 0(10):  36-39. 
    Asbtract ( 126 )   PDF (722KB) ( 149 )  
    References | Related Articles | Metrics
    In order to improve the ability of multi-source database cache conflict adjustment, a multi-source database cache conflict automatic processing method based on rough set is proposed. The multi-source database cache data channeling model is constructed, the load balance scheduling method is used to balance the multi-source database cache, the multi-source database cache data feature mining is carried out with the fuzzy rough set feature extraction method, and the rough set features of multi-source database cache data are extracted. The multi-source database cache conflict is adjusted adaptively by multiple information reorganization and big data information fusion method. Through the distributed attributes of rough set, the cache conflict of multi-source database can be handled automatically. The simulation results show that the multi-source database cache has better equalization configuration ability, reduces the risk of data congestion, the number of rounds of conflict and the number of dead nodes, and improves the balance and security of multi-source database cache.
    Big Data Mining Algorithm of Heterogeneous Multi-core Platform Based on Semantic Segmentation
    ZHOU Xian-lai
    2020, 0(10):  40-43. 
    Asbtract ( 148 )   PDF (793KB) ( 183 )  
    References | Related Articles | Metrics
    In order to improve the accurate mining ability of heterogeneous multi-core platform big data, a method based on semantic segmentation is proposed. The fuzzy information detection model of heterogeneous multi-core platform big data is constructed, and the fuzzy directional clustering analysis of heterogeneous multi-core platform big data is carried out by using association features extraction method. The output autocorrelation features matching model of heterogeneous multi-core platform big data is constructed, and the features extraction and statistical analysis of heterogeneous multi-core platform big data are carried out by semantic features extraction method. The semantic dynamic features analysis model of heterogeneous multi-core platform big data is established, and the statistical characteristics of heterogeneous multi-core platform big data are extracted. According to the features extraction results of heterogeneous multi-core platform big data, the fuzzy C-means clustering method is used for big data clustering, and the semantic segmentation is used for adaptive optimization in the process of heterogeneous multi-core platform big data mining to realize the optimized mining. The simulation results show that the proposed method has higher accuracy and better features resolution, which can improve the mining and detection ability of heterogeneous multi-core platform big data.
    Segmentation of White Matter Lesions Based on 3D Full Convolutional Deep Neural Network
    ZHAO Xin, SHI De-lai, WANG Hong-kai
    2020, 0(10):  44-50. 
    Asbtract ( 174 )   PDF (1996KB) ( 111 )  
    References | Related Articles | Metrics
    The automatic segmentation of brain white matter lesions has an important auxiliary role in the clinical diagnosis and research of brain diseases. At present, researchers mainly use deep learning method to solve the problem of automatic segmentation of white matter lesions. Although some achievements have been achieved, there are still problems of low segmentation accuracy and small lesions can’t be segmented precisely. In this paper, a fully convoluted 3D deep neural network model is proposed, which integrates residual, pyramid pooling and attention mechanism. In this model, the residual net is used to avoid the gradient disappearance; pyramid pooling is used to aggregate more context information; attention mechanism is used to locate the reign of interest. All modules are connected in order to build a convolutional module chain with strong learning ability, and the up and down sampling are attached at both ends of the chain to form a complete end-to-end deep neural network model. The experiment is carried out on the MICCAI 2017 data set. Experimental results show that compared with other methods, the DSC score of this paper is 0.762, the recall rate is 0.727, the accuracy rate is 0.801, the specificity is 0.991, and the segmentation results are better than those mentioned in other literatures.

    A High Precision Microporous Plate Turbidity Identification Network
    LI Xi-ming, MA Li-xiao, ZENG Xiao-yin, WANG Xuan, SUN Jian, GUO Yu-bin
    2020, 0(10):  58-63. 
    Asbtract ( 118 )   PDF (1491KB) ( 106 )  
    References | Related Articles | Metrics
    A high precision microporous plate turbidity classification algorithm based on convolutional neural network is proposed. This algorithm mainly combines the traditional image processing technology with the convolutional neural network technology. Through the traditional image processing algorithm, round holes are cut from the microporous plate images taken naturally, and the cut round hole images are made into round hole data sets for the training, evaluation and testing of network models. At the same time, through the deep learning technology, multiple convolutional neural network models based on the depth-separable convolution kernel are designed and trained. Then, the turbidity classification model with the highest evaluation accuracy is selected and applied to the circular hole identification system, thus improving the work efficiency of researchers.
    Smoke Removal Algorithm of Gray-scale Image in Fire Field Based on Deep Learning
    MA Yue
    2020, 0(10):  64-68. 
    Asbtract ( 350 )   PDF (1827KB) ( 210 )  
    References | Related Articles | Metrics
    Due to the existence of a large number of smoke in the fire scene, the image clarity of the video monitoring system becomes blurred, and the contrast and clarity of the image decline, which can not provide effective visual support for evacuation and search and rescue. In view of this situation, this paper proposes a gray-scale image smoke removal algorithm based on deep learning. The network proposed in this paper is mainly composed of two parts: detection sub network and removal sub network. The former determines the specific location of smoke through residual learning network, and the latter uses dense U-shaped network to remove smoke while retaining the original background, and uses dense block to reuse low-level features to high-level features to further improve the accuracy of smoke removal. A large number of experimental results show that the network has better performances in removal effect and real-time, and the subjective evaluation and objective evaluation are better than other comparison algorithms.
    Product Image Similarity Algorithm Based on SIFT and Nearest Neighbor Matching
    WU Ying
    2020, 0(10):  69-75. 
    Asbtract ( 275 )   PDF (1779KB) ( 186 )  
    References | Related Articles | Metrics
    With the surge in the number of e-commerce users, various problems continue to emerge. Among them, the incidents of plagiarism and copying information from other stores in the same industry often occur, but plagiarized image information is more difficult to detect similarity than text information, because plagiarists often may crop, rotate or add filter to image information, in addition, they can process images with PS and other technologies, which makes it difficult to detect the similarity between the processed image and the original image. However, the manual comparison is inefficient and costly, so a system based on an algorithm that can quickly calculate the similarity of commodity images is required to solve this problem. Scale-invariant feature transform (SIFT) descriptors can solve the limitations of traditional algorithms for low similarity of rotated images. The accuracy of the algorithm SIFT is high, and it can describe rich feature information. Based on the introduction of the traditional image Hash algorithm, a similar nearest neighbor matching algorithm based on SIFT descriptor is proposed for the similarity comparison of electric drill product images. The original image is cropped, added filter, added contrast, rotated, added watermarks etc, respectively, and these processed images are all compared with the original images about similarity. The experimental results show that the similar nearest neighbor matching algorithm has better accuracy than Hash algorithm and SIFT algorithm, and it can identify plagiarized image information more accurately.
    Reconstruction of 3D Cartoon Character Modeling Based on Topology Analysis
    LIU Jian-gao, BIN Min
    2020, 0(10):  76-83. 
    Asbtract ( 312 )   PDF (1797KB) ( 160 )  
    References | Related Articles | Metrics
    Traditional methods do not determine the key feature points in the process of 3D animation character modeling reconstruction, which results in low matching degree and slow convergence speed of attitude image. Therefore, a 3D animation character modeling reconstruction method based on topology analysis is designed. Topological analysis is used to determine the point, line and face structure of 3D animation character modeling. The edge detection method based on fusion technology is used to extract feature points, and calculate the gradient value and gradient direction of feature points, so as to determine the key feature points and generate the main direction and feature descriptor. On this basis, the descriptor distance of the key points in the corresponding source database is calculated, and the feature points are matched according to the ratio of the minimum distance to the next small distance, and the 3D animation character modeling is reconstructed by interactive geometric constraint deformation. The test results show that the overall matching degree of the designed 3D animation character modeling reconstruction method is above 0.6, which is far higher than the matching degree of traditional reconstruction methods, and the convergence speed of the method is faster, which shows that it is suitable for the 3D animation character modeling reconstruction design.
    A Key Frame Automatic Selection Method for Moving Object
    CHEN Xiang, ZOU Qing-nian, XIE Shao-yu, CHEN Cui-qiong
    2020, 0(10):  81-89. 
    Asbtract ( 113 )   PDF (3110KB) ( 107 )  
    References | Related Articles | Metrics
    In order to avoid the accidents, such as mis-operation, strayed into the scene and so on, which are caused by the lack of on-site monitoring of substation, the scene reconstruction technology is usually adopted to monitor the object state in real time. The real-time model of substation equipment in the working state is analyzed, and a key frame automatic selection method of moving object is proposed. Firstly, the KLT algorithm with the pyramid structure is used to track feature points, then the change of the object attitude is calculated by the two-step decomposition method, and the real-time state of the object is predicted by the change of the object attitude in a short time. In order to evaluate the effectiveness of the proposed method, the relative error (RE), root mean square error (RMSE) and absolute trajectory error (ATE) of the proposed method are compared with the reference trajectory qualitatively and quantitatively. The experimental results show that the proposed algorithm reduces the data redundancy by about 40%-60%, and the real-time and robustness of the location and scene construction are improved effectively.
    Visualization of Radar Detection Range Based on 3D Earth Scene Under Interference Condition
    WU Zhong-de, LUO Xiao-fang, HOU Zeng-xuan, DUAN Peng-xuan, LI Nan-nan
    2020, 0(10):  90-96. 
    Asbtract ( 191 )   PDF (3441KB) ( 210 )  
    References | Related Articles | Metrics
    The 3D visualization of radar detection range is the key component of the display of modern battlefield electromagnetic situation. This paper puts forward a visualization method of radar detection range drawing based on the 3D earth scene. The method is based on the model of radar detection range in electronic countermeasures under ideal conditions and interference conditions, and combines with the coordinates transformation method between the geodetic coordinate system and geocentric coordinate system. By studying and designing a database based on the visualization data requirements of radar detection range under ideal conditions and interference conditions, based on the 3D earth scene platform ArcGlobe and OpenGL technology, we apply the visualization method to develop the marine battlefield electromagnetic situation display system. The experimental results show that the visualization method can accurately and vividly show the radar detection range in 3D earth scene under ideal conditions and interference conditions, and provide support for commanders to insight into the battlefield situation.
    A Pose Fusion Method of Augmented Reality and Visual Inertial Navigation Module
    NIE Lei-hang, NIE Yun, WANG Guo-wei
    2020, 0(10):  97-102. 
    Asbtract ( 155 )   PDF (2284KB) ( 163 )  
    References | Related Articles | Metrics
    Augmented reality technology based on visual markers and visual inertia odometer (VIO) technology have good complementarity. In this paper, a general pose fusion method is proposed to solve the problems of the current augmented reality system based on markers and VIO’s lack of geographic location information and cumulative errors. This method can transform the output of the same trajectory pose in arbitrary two different coordinate systems to the same coordinate system. Aiming at the problems in this paper, we realize the integration of the position and pose of the augmented reality based on visual markers and the visual inertial navigation module, and provide the real geographic information coordinates for the whole system by using the characteristics of the geographic information of the visual markers, so that the positioning system can be combined with the geographic information system. The image and IMU information output from Huawei P10 phone are collected in the form of real-time communication as the data source, and experiment is conducted on Ubuntu16.04 and Unity game engine. The results show that the proposed method can achieve accurate pose fusion effectively.
    Energy-Efficient Resource Allocation for Hybrid Spectrum Sharing Cognitive Radio Networks
    WANG Gui-zhu, LU Ling-yun, LI Xiang
    2020, 0(10):  103-109. 
    Asbtract ( 127 )   PDF (1332KB) ( 108 )  
    References | Related Articles | Metrics
    In cognitive radio networks, cooperative spectrum sensing can improve spectrum sensing detection performance by using multiple nodes sensing simultaneously. However, with the increase of the number of secondary users (SU), energy consumption increases and energy efficiency (EE) decreases. In order to solve this problem, an energy efficiency model based on hybrid spectrum sharing mode is constructed by combining two sharing modes of opportunistic spectrum access and underlay spectrum sharing. Meanwhile, three different fusion rules, reoccupation probability of primary user (PU) and reporting channel error are considered. Aiming at maximizing EE of the secondary system, Lagrange multiplier method and sub gradient descent algorithm are used for sensing time, the number of participants and the transmission power of users to be solved by iterative optimization. The simulation results show that the algorithm can achieve higher throughput and energy efficiency under the constraints of the minimum quality of service (QoS) and transmission power.
    Application of Timed Petri Net in Penetration Testing
    LI Hai-hao, WU Ya-feng, WANG Xiao-qiang
    2020, 0(10):  110-115. 
    Asbtract ( 155 )   PDF (772KB) ( 82 )  
    References | Related Articles | Metrics
    As an important part of penetration testing, the penetration test attack model has attracted the common attention of academia and industry. Existing penetration test attack models do not take into account the dynamic parameters in the penetration test attack process, and cannot describe when the vulnerability occurred. Taking the vulnerability as the basic unit, and the time interval of the place in the timed Petri net representing the occurrence interval of the vulnerability, this paper builds a penetration test attack model based on the timed Petri net. First, the vulnerability list is used as input to build a single vulnerability model. Then, the single vulnerability model collection is used to form a complete penetration test attack model through a model integration algorithm. Finally, the algorithm of penetration attack path selection is given, and the effectiveness of the algorithm of penetration attack path selection proposed in this paper is verified by simulation experiments.
    Multi-channel Cooperative Downloading Method for Vehicle Networking Based on Cooperative Vehicle Group
    JIANG Xu, YUAN Jing, ZHANG Zhen-xing
    2020, 0(10):  116-121. 
    Asbtract ( 120 )   PDF (1059KB) ( 94 )  
    References | Related Articles | Metrics
    Aiming at the problems of low throughput and superimposed communication domain of the single-channel cooperative downloading method in the current internet of vehicles, a multi-channel cooperative downloading strategy of vehicle networking based on opposite cooperative vehicle group (McDvg) is proposed. Utilizing the characteristics of large meeting probability, multiple retransmission of communication and extension of data transmission time of opposite cooperative vehicle group to form a vehicle group in a local area, we design a car selection strategy to select a suitable oncoming vehicle to carry the data required by the user to provide cooperative download for target vehicles. The method utilizes the opposite cooperative vehicle group to effectively extend the cooperative download time of the target vehicle, and adopts a multi-channel strategy to solve the communication conflict problem in the case of a single channel. At last, the scheme is verified by corresponding simulation experiments. The simulation results show that this method can significantly improve the download throughput and reduce the communication delay in the blind zone, so as to provide effective service support for the current internet of vehicles.
    Optimal Relay Selection Cooperative Communication Based on Polarization Diversity Technology with Equal Gain Combining
    LIU Bo, WANG Ming-wei
    2020, 0(10):  122-126. 
    Asbtract ( 150 )   PDF (663KB) ( 99 )  
    References | Related Articles | Metrics
    In multi-relay cooperative communication, Optimal Relay Selection (ORS) greatly simplifies network physical layer design by eliminating the need for accurate synchronization of all nodes. Equal Gain Combining (EGC) does not need to evaluate the decline amplitude, each branch has the same unit gain, weight coefficient does not change with the strength of fading channel signal at any time, which simplifies the complexity of the system design and implementation. Therefore, we propose in this paper, the polarization antenna is configured for the communication terminal, and the polarization diversity signal is processed by EGC without increasing the terminal volume, so as to overcome channel fading and further improve the performance of Decode-and-Forward ORS cooperative communication system. The theoretical analysis and simulation results show that the outage probability index of Decode-and-Forward ORS cooperative communication system is obviously better than that of the cooperative communication system without EGC polarization diversity technology in the same SNR and channel fading. Finally, EGC is compared with other polarized diversity merging techniques.