Loading...

Table of Content

    03 December 2020, Volume 0 Issue 11
    Pulmonary Nodule Aided Diagnosis System Based on Target Detection Algorithm
    XI Xiao-qian, LIU Wei
    2020, 0(11):  1-7. 
    Asbtract ( 816 )   PDF (1589KB) ( 852 )  
    References | Related Articles | Metrics
    According to statistics, lung cancer is one of the diseases with the highest morbidity and mortality rate in the world. With the maturity of computer-aided diagnosis (CAD) and convolutional neural networks (CNN), the diagnosis and treatment in medical field are becoming more and more intelligent. In this paper, an automatic detection method of lung nodules based on target detection algorithm is presented, and a set of image processing flow of CT of lung parenchyma is presented, which combines threshold segmentation algorithm with digital morphological processing. After training and learning 1186 lung nodules in LUNA16 data set, we observe the evaluation results of YOLO V3 model in the data set to verify the model. The accuracy of the experimental results is 92.18%, the average detection time of each image is 0.035 seconds. In order to verify the validity of YOLO V3 model, this paper compares it with existing algorithms such as SSD, CNN, U-Net and so on. At the same time, this paper designs an auxiliary diagnosis system of pulmonary nodules based on CAD technology, realizes the human-computer interaction, and provides a simple and clear auxiliary diagnosis tool for doctors.
    Surface Defect Detection of Aluminum Profile Based on Image Fusion and YOLOv3
    ZHANG Lei, LANG Xian-li, WANG Le
    2020, 0(11):  8-15. 
    Asbtract ( 333 )   PDF (5406KB) ( 255 )  
    References | Related Articles | Metrics
    China is a large country in the manufacture of aluminum profiles, and the quality inspection of aluminum profiles is of great significance. Aiming at the problems of low detection efficiency and relatively weak stability of traditional manual visual inspection methods, the feature extraction of the single YOLOv3 method is not prominent, and the detection accuracy is limited, this paper proposes a method for detecting surface defects of aluminum profiles based on image fusion and YOLOv3. First, image enhancement and spatial filtering methods are used to preprocess the original image to obtain a processed image; then the feature extraction and matching ideas in SLAM are used for reference to extract and match the original image and the processed image; then image fusion is performed to obtain the final processed image; subsequently, the K-means algorithm is used for clustering and tuning parameters optimization. Finally, a single-stage object detection model YOLOv3 is used to detect the surface defects of aluminum profiles. An end-to-end full convolutional neural network is used to complete the input from the original image to the output of the object categories and confidence in the Bounding box and box. The experimental results show that the average success rate of surface defect classification detection by this method is 98.33%, which is 3.75 percent points higher than the single YOLOv3 method; the mAP value of the validation set is 88.81%, which has an increase of 4.18 percent points. It has stronger feature extraction and generalization capabilities, and can accurately detect surface defects, perform classification and localization.
    Target Tracking Algorithm Based on Multimodal Data
    ZHOU Jing-wei, HAN Li-xin, LI Xiao-shuang
    2020, 0(11):  16-22. 
    Asbtract ( 256 )   PDF (3661KB) ( 305 )  
    References | Related Articles | Metrics
    In order to solve the problems of target occlusion and complex background in target tracking, a target tracking algorithm based on multimodal data is proposed. First, pixel-level fusion of each modal data is performed to reduce the impact of insufficient information in single-modal data on the tracking results. Then, different features are extracted and filtered from the fused image. At the same time, in order to solve the problem of model tracking failure caused by a single model drift, the response graph obtained by filtering is merged at the decision level. Finally, the tracking result is obtained according to the peak value of the fused response graph. In addition, an occlusion detection module is added in the tracking process to enhance the model robustness. The evaluation of the algorithm on the Princeton tracking benchmark shows that, compared with other mainstream algorithms, the target tracking algorithm based on multimodal data improves the tracking accuracy on target occlusion videos by 8.4% and the coincidence success rate by 3.3%. It has a good anti-occlusion effect.
    Lightweight Image Super-Resolution Based on Convolutional Neural Network
    LIANG Chao, HUANG Hong-quan
    2020, 0(11):  23-27. 
    Asbtract ( 216 )   PDF (2031KB) ( 152 )  
    References | Related Articles | Metrics
    In recent years, deep convolutional neural networks have performed well in solving single image super-resolution problems. For improving the disadvantage that, the deeper the layer number of convolutional neural network is, the greater the amount of calculation is, the slower the real-time reconstruction speed is, combined with the existing convolutional network model, a lightweight network structure is proposed. This paper reduces the number of network layers in the neural network layer and uses channel split to build a multi-scale enhanced structure of local features. Then it combines the residual network for model construction. Experiment results show that, compared with LapSRN method, VDSR method, and traditional interpolation method, this method is faster in real-time reconstruction and is not weaker than others in peak signal to noise ratio and structural similarity.
    Strategy of “Fighting the Landlord” Based on Deep Convolutional Neural Network
    XU Fang-jing, WEI Kun-peng, WANG Yi-song, PENG Qi-wen, YU Xiao-min
    2020, 0(11):  28-32. 
    Asbtract ( 318 )   PDF (1163KB) ( 236 )  
    References | Related Articles | Metrics
    Deep neural network has made amazing achievements in various foreign games. In recent years, convolutional neural network has gained great attention because of its unique unit structure, and has been frequently used in game AI agents, such as AlphaGo and Cold Flutter Masters. “Fighting the Landlord” is a typical cooperative game based on incomplete information. In this paper, a 7-layer convolutional neural network DDZ-CNN is designed to train the network with nearly 300,000 pieces of data based on the self-gaming of “Fighting the Landlord” based on Monte Carlo tree to learn the “Fighting the Landlord” strategy. In the training process, the training data are down sampled by a weight-based method to overcome the problem of uneven distribution, and the network can converge quickly. Finally, the trained model is combated with intelligent MCTS model and real person, and a good winning rate is obtained, which verifies the effectiveness and feasibility of the algorithm in this paper.
    An Electronic Device Container Quality Detection Method Based on Cascade R-CNN
    WU Shui-ming, ZHU Yan, WANG Fang, JING Dong-sheng
    2020, 0(11):  33-38. 
    Asbtract ( 128 )   PDF (2087KB) ( 94 )  
    References | Related Articles | Metrics
    The production of electronic device container is a process with high requirements for safety, efficiency and integrity, which must be paid attention to by major enterprises. But in the actual production and packaging process, the stains on the container, the foreign matters in the container, and the appearance abnormalities are inevitable. These problems need to be solved urgently. At present, the main detection methods to solve these problems are manual detection and traditional machine vision. The disadvantages of manual detection are high accuracy and low efficiency. Traditional machine vision detection methods are of high efficiency and low accuracy, which are difficult to meet the requirements of high-speed automatic production line. Therefore, this paper proposes an electronic device container quality inspection method based on cascade R-CNN. In view of the actual process of the container data oriented improvement network, we add the samples that are difficult to distinguish from Focal Loss detection, use deformable convolution to extract features more efficiently, train the strong robustness model with multi-scale training method, and apply it to the multi-category detection of electronic device containers. The experimental results show that the improved model based on cascade R-CNN has high accuracy and strong robustness.
    Crop Leaf Diseases Recognition: A Generative Adversarial Network Based Approach
    XIONG Fang-kang, LU Ling, CAO Ting-rong, PENG Li-jun
    2020, 0(11):  39-46. 
    Asbtract ( 189 )   PDF (2763KB) ( 180 )  
    References | Related Articles | Metrics
    When people use deep neural networks to classify images, they usually need a large number of training samples. However, it is difficult to obtain enough samples to ensure the training of neural network in practice. In order to solve this problem, this paper proposes an identification method based on generative adversarial network. The main idea is to train a sample generation model after modification of the existing GAN network model, then use neural network to identify the data set generated by the generation model, and finally use transfer learning method to fine-tune the neural network with real data. In order to verify the effectiveness of this method, five crop leaves (500 pieces per sample) are used for validation, the identification accuracy of plant leaves is more than 90%. The experimental results show that this method can improve the identification accuracy of the blade with a small number of samples and has strong universality.
    RDF Stream Data Query Based on Apache Flink
    ZHENG Tao, LIU Meng-chi, FENG Jia-mei
    2020, 0(11):  47-55. 
    Asbtract ( 250 )   PDF (1691KB) ( 146 )  
    References | Related Articles | Metrics
    At present, mature RDF Stream Processing (RSP) systems lack parallel processing characteristics due to the centralized design. Therefore, when querying and processing a large amount of incoming RDF stream data, high throughput and low latency cannot be achieved. In order to improve the query performance, this paper researches the RSP query process and Flink stream calculation structure, designs four logical operators: source, filter, multi-way partition join and project, and designs a Multi-Stream Join (MSJ) algorithm that is used to generate a logical query plan of a directed acyclic graph with parallelism. Finally, a big data stream processing platform called Apache Flink is used to implement the logical operator and logical query plan. The real data set SRBench and simulated data set LUBMs are used for experimental verification. The results show that compared with the most mature systems C-SPARQL and CQELS, the throughput of a single machine increases by 10 times, the throughput of a cluster of 5 machines increases by 28 times, and the latency reaches the millisecond level; in terms of query performance, the throughput is improved and the latency is reduced when processing a large amount of RDF stream data.
    Construction of Meteorological System Data Service Platform Based on Distributed Technology
    LEI Ming, ZHAO Yu-juan, JIANG Han-sheng, WU Guo-liang, LIANG Jian
    2020, 0(11):  56-59. 
    Asbtract ( 137 )   PDF (1765KB) ( 116 )  
    References | Related Articles | Metrics
    The lack of unified planning for the construction and design of the existing meteorological system has resulted in a large number of systems, scattered, disordered and poor data management, weak overall safety protection capability and poor intensification. Using distributed technology, this paper constructs a hybrid model big data service center based on “distributed relational database+transactional database+columnar database+table system+distributed file system”, which stores, manages and serves data, and seamlessly connects with provincial CIMISS system through MUSIC interface, which effectively realizes data sharing and highly integrated process about, and greatly improve the ability of data impact. Under the same conditions, the performance of automatic station data query in 10 years has increased by more than 22 times, and with the increase of time dimension, the advantage of this improvement will be greater.
    Entity Extraction Method of Chinese Electronic Medical Record Based on CNN-BGRU-CRF
    FENG Yun-xia, YI Peng, HANG Zheng-liang, SONG Bo
    2020, 0(11):  60-64. 
    Asbtract ( 169 )   PDF (927KB) ( 110 )  
    References | Related Articles | Metrics
    To solve the problem that traditional methods are too dependent on dictionaries and word segmentation tools in entity extraction of Chinese Electronic Medical Records and cannot make full use of contextual features, this paper proposes a Chinese EMR entity extraction model based on the combination of word embedded convolution (CNN), bidirectional gated loop unit (BGRU) and conditional random field (CRF). In the first place, the word embedding method is used to extract the potential word features, and then the attention mechanism is used to highlight the specific information while using the joint method of word features. At last, the final result is obtained by rationality constraint. This model makes full use of word features to avoid the influence of wrong word segmentation on entity extraction and to reduce the process of artificial feature construction, improve the efficiency of entity extraction. The experimental results show that the F value of the model performs better than the traditional Bi-LSTM-CRF model in entity extraction of diagnosis name, symptom name and treatment type. 
    A Combined Prediction Model of Civil Aviation Passenger Reservation Based on ARIMA and LSTM
    ZHAO Xuan
    2020, 0(11):  65-69. 
    Asbtract ( 229 )   PDF (918KB) ( 158 )  
    References | Related Articles | Metrics
    In this paper, a combined prediction model is proposed for the prediction of passenger reservation in civil aviation. First of all, according to the characteristics of reservation, this paper designs an algorithm dealing with missing values and noise so as to correct the reservation data effectively. Then, for improving prediction accuracy, a combined prediction model which is composed of the ARIMA and LSTM algorithm is constructed. The actual passenger reservation data of an airline company is used for the prediction. The experiments show that the values of MAE and RMSE of the combined model are smaller than single model before combination. The prediction result is also more accurate and applicable.
    Sensor Network Deployment Algorithm for 3D Surface with Multiple Mobile Nodes
    CAI Wen-peng, JU Shi-guang
    2020, 0(11):  70-76. 
    Asbtract ( 106 )   PDF (1514KB) ( 88 )  
    References | Related Articles | Metrics
    At present, most of the researches on sensor network deployment are mainly focused on two-dimensional(2D) planes and three-dimensional(3D) full-space areas. However, many real-world application fields are a complex 3D surface. The existing coverage methods cannot achieve good results. This paper studies the deployment method of 3D surface sensor networks, and proposes a 3D surface multi-mobile node sensor network deployment algorithm. A hybrid sensor network composed of static nodes and dynamic nodes is used to estimate the position and area of the coverage hole from the static nodes. The coverage holes are sequentially repaired by mobile nodes. Simulation results show that the final network coverage rate of the algorithm reaches 99%, which is 6 percentage points higher than the 3DGA algorithm and 8.5 percentage points higher than the Delaunay algorithm, while reducing overall network energy consumption.
    Improved Path Search Algorithm Based on Layering and Reinforcement Learning
    WANG Hai-hong, LIU Li
    2020, 0(11):  77-82. 
    Asbtract ( 134 )   PDF (1086KB) ( 119 )  
    References | Related Articles | Metrics
    The problem of path search in a complex network is a difficult point in network optimization. Existing path search algorithms mainly have the following problems: Firstly, they can only focus on one of solution efficiency and solution accuracy; secondly, they are not adaptable to complex networks with dynamic changes, and the solution effect is not good. So this paper proposes an improved path search algorithm based on double layering and optimized Q-Learning. For the problem that the solution time increases sharply with the increase of scale, a dual-layered strategy of dividing the network by combining k-core and modularity is proposed to reduce the network size reasonably and effectively. In the subnet solution, the reinforcement learning mechanism is introduced to dynamically sense the network. For the problem of slow convergence of the algorithm, the adaptive learning factor and memory factor are added to optimize the update formula and improve the convergence speed. Finally, under different power-law exponents (2 to 3) and complex networks of different sizes, the proposed algorithm is compared with Dijkstra algorithm, A* algorithm and Qrouting algorithm. The results show that the algorithm can effectively improve the solution efficiency while ensuring a good solution accuracy.
    Improved CamShift UAV Target Tracking Algorithm
    LI Rui, SHANG Jia-he
    2020, 0(11):  83-88. 
    Asbtract ( 147 )   PDF (3962KB) ( 111 )  
    References | Related Articles | Metrics
    At present, there are still some problems in the application of UAV video target tracking algorithms. For example, the tracking effect is not good when the illumination is uneven, the target rotates, and the target is blocked. Therefore, this paper proposes a CamShift tracking algorithm combining HLBP feature matching with Kalman filtering. First, the target features are extracted by the HLBP algorithm to obtain more accurate texture features, and then the interference caused by the change in illumination and the target rotation to the feature extraction is reduced. Second, the degree of occlusion of the target is judged by Bhattacharyya distance. Finally, the Kalman filter algorithm is used in the prediction of the target position, which can effectively solve the problem of poor tracking effect when the target is blocked. The experimental results show that in the practical application of UAV target tracking, the improved algorithm can effectively reduce the impact of external interference on the tracking effect, and the tracking accuracy is improved.

    Visualization of 3D Virtual Campus Based on osgEarth
    WANG Xiao-yu, SUN Ka
    2020, 0(11):  89-93. 
    Asbtract ( 231 )   PDF (4898KB) ( 118 )  
    References | Related Articles | Metrics
    Aiming at the complex campus environment and difficult simulation, this article uses osgEarth three-dimensional geographic information system to realize the construction and function development of virtual campus. Taking the Qianhu campus of Nanchang Hangkong University as an example, MultiGen Creator with high modeling accuracy is selected as the modeling tool to create the 3D scene model, and osgEarth with analysis function is used as the simulation platform for the secondary development of specific functions. By building a digital earth, using remote sensing image data, elevation data and 3D models, the visualization and function construction of the 3D virtual scene of the Qianhu campus of Nanchang Hangkong University is realized. The technology plays an important role in school admissions, employment, publicity, education, foreign exchange and other aspects.
    Three-dimensional Reconstruction of Plants Based on Structure from Motion Method
    SU Tao, SHAO Ying-xin, LI Chun-bin, WU Jing, CHANG Xiu-hong, YANG Fan
    2020, 0(11):  94-99. 
    Asbtract ( 190 )   PDF (1489KB) ( 129 )  
    References | Related Articles | Metrics
    In order to explore the effect of three-dimensional reconstruction model of plants based on Structure From Motion (SFM) method and provide a research case for plant 3D reconstruction work, this paper uses setcreasea pallida as research object, selects 35, 75, and 105 sequence images for comparative analysis of 3D reconstruction based on the sequence image acquisition platform. At the same time, the accuracy of the plant three-dimensional reconstruction model is evaluated from the aspect of plant phenotypic parameters. The results show that the reconstruction of 75 image sequences is the best; the height Relative Error(RE) of the images calculated by different image sequences is less than 2.5%, and coefficient of determination(R2) is greater than 0.998; the RE of the different image sequences extracted from leaf length and leaf width are less than 2.89%,  R2 are greater than 0.958. Therefore, the number of sequential images is related to the effect of the reconstruction model, but the two are not positively correlated; the number of sequence images has little effect on the error of reconstruction of blade length and width. SFM method can be applied to the three-dimensional reconstruction of plants with complex structure, which can achieve better reconstruction results.
    Internet of Things Access Control System Based on Blockchain
    SUN Guang-cheng, LI Hong-zhe, LI Sai-fei, ZHANG Xiao-wei
    2020, 0(11):  100-108. 
    Asbtract ( 231 )   PDF (2908KB) ( 163 )  
    References | Related Articles | Metrics
    With the widespread application of 5G technology and the rapid development of IoT technology, it is foreseeable that the number of IoT devices and the scale of connections will further expand. The Internet of Things can effectively use the existing network infrastructure to connect the physical world and the Internet world to achieve data sharing between devices. However, the large-scale and complexity of its network structure brings potential security risks to the Internet of Things system. Access control technology can be used to protect the security of the device and the data on the device. The traditional access control model and implementation are more complex and centralized. Therefore, this article expects to build a new type of blockchain-based Internet of Things access control system: Based on the RBAC model and the ABAC model, the DARBAC model is proposed, and the access control system is built in conjunction with the blockchain technology to overcome the single-point failure problem of centralized entities. It has the characteristics of scalability, lightweight, and fine-grained. Experimental test results show that the system has good concurrency performance, can be effectively deployed and implemented in Internet of Things system and can achieve the expected results.
    CBR-based Emergency Response System and Method for Internet of Vehicle Security
    LIAO Zu-qi, LI Fei, ZHANG Peng-fei
    2020, 0(11):  109-116. 
    Asbtract ( 140 )   PDF (1522KB) ( 87 )  
    References | Related Articles | Metrics
    With the rapid development of the Internet of vehicles, its security issues are increasingly prominent. At present, the main research direction of the academic community is the security of specific terminals, and the Internet of vehicles is not as a whole to carry out security research, so the research in this direction is in an urgent moment. To solve this problem, this paper proposes a network security emergency response method for the Internet of vehicles based on set of experience knowledge structure (SOEKS) and case-based reasoning (CBR) to realize the automatic and fast processing of specific security events. The paper designs the knowledge representation method for the data sources and cases of the Internet of vehicles, and designs a double similarity matching algorithm based on the nearest neighbor algorithm to quickly match the security events, so as to get a fast and more accurate response. Finally, the paper implements the emergency response system, verifies that the system can get accurate emergency response plan of current events from historical data, and confirms the feasibility and effectiveness of the method proposed in this paper.
    Risk Assessment Model of Mobile Secure Payment Based on Improved AHP Algorithm
    CHEN Xiao-wei
    2020, 0(11):  117-121. 
    Asbtract ( 127 )   PDF (729KB) ( 202 )  
    References | Related Articles | Metrics
    With the development of mobile Internet and Internet of things, mobile payment is more and more deeply integrated into people’s daily life. Mobile payment is an indispensable convenience service, but from the perspective of security, it has certain risks. Starting from the uniqueness of mobile payment, this paper establishes a hierarchical evaluation target system by improving AHP, and determines the weight of the factors that affect the security risk of mobile payment. In the construction of AHP algorithm judgment matrix, three-scale method is used to replace the traditional nine-scale method, which is compared with the traditional nine-scale method. The results show that the improved method overcomes the problem that experts are difficult to master the judgment scale in the traditional analytic hierarchy process. The adoption of evaluation index structure and improved AHP can provide effective security payment risk assessment for mobile payment, improve the efficiency and accuracy of security risk assessment for mobile payment, and has strong practicability.
    Interference and Load Aware Routing Metric for Multi-gateway Wireless Mesh Networks
    XU Yin-long, YOU Chun
    2020, 0(11):  122-126. 
    Asbtract ( 148 )   PDF (765KB) ( 111 )  
    References | Related Articles | Metrics
    A routing metric applied to multi-gateway wireless Mesh networks is proposed. This metric first establishes a multi-gateway network model, obtains network topology information and assigns different weights to network nodes accordingly. Then, the distribution characteristics of the link SINR are derived through the network topology information and the fading characteristics of the wireless link to calculate the outage probability and interrupt rate of the link. Finally, this metric obtains the number of Mesh router node cache packets, and calculates the gateway capacity, designs order-aware interfere and load aware routing metrics for client and Internet business. When there are client business and Internet business requests, this metric provides route through routing metric based on routing protocols for different business. This paper considers network interference factors and load distribution factors, can effectively reduce network interference, achieve network load balancing, and improve network throughput.