Computer and Modernization

Dam Deformation Prediction Based on EMD-GAELM-ARIMA Algorithm

XU Xiao-yao, ZHANG Peng-fei, JIANG Jian

2020, 0(07): 1-5. doi:10.3969/j.issn.1006-2475.2020.07.001

Asbtract ( 284 )

PDF (1260KB) ( 243 )

References | Related Articles | Metrics

In view of the fact that it is difficult for statistical models to make good predictions of nonlinear and non-stationary dam deformation, artificial intelligence algorithms are induced. The empirical mode decomposition method (EMD), genetic algorithm (GA) optimized extreme learning machine (ELM), and ARIMA error correction model were used to construct a dam deformation prediction model. First this paper uses EMD to decompose and reconstruct the monitoring data to stabilize it and obtain eigenmode functions and residual sequences with physical significance; then uses GAELM to analyze and predict the decomposition results; finally, uses ARIMA model to correct errors. Taking a concrete rockfill dam as an example, the dam deformation prediction model constructed by the optimization algorithm is used to analyze and predict it. The analysis results show that the EMD-GAELM-ARIMA model algorithm has higher prediction accuracy than the traditional single algorithm. It is feasible in dam deformation prediction.

Agricultural Information Collaborative Filtering Recommendation Algorithm Based on User Behavior and News Timeliness

XU Jian-peng, XU Xiang, WANG Hui, WU Qiong, WANG Jie

2020, 0(07): 6-10. doi:10.3969/j.issn.1006-2475.2020.07.002

Asbtract ( 225 )

PDF (680KB) ( 220 )

References | Related Articles | Metrics

Agricultural information has strong timeliness and periodicity. Traditional behavior-based recommendation algorithms can mine farmers’ interests but cannot reflect the information needs of farmers at different time periods. At the same time, farmers generally use an anonymous webpage to browse agricultural news directly. Explicit feedback data is very scarce. Traditional collaborative filtering recommendation algorithms need to face cold start problems. This paper proposes a collaborative filtering recommendation algorithm based on user behavior and news timeliness, which integrates the multi-dimensional factors such as user’s implicit and explicit feedback data, and considers the classification characteristics and periodicity of agricultural information. It improves the pertinence and timeliness of agricultural news recommendation according to the periodic attention change of different agricultural classification information and the heat coefficient. The experimental results show that the proposed algorithm can effectively improve the accuracy of agricultural information recommendation.

Application of Markowitz Mean-variance Theory in Portfolio Optimization of Energy Futures

WANG Yi-fan

2020, 0(07): 11-15. doi:10.3969/j.issn.1006-2475.2020.07.003

Asbtract ( 297 )

PDF (1400KB) ( 144 )

References | Related Articles | Metrics

This paper explores the investment strategy of energy futures from the perspective of individual investors. Based on Markowitz mean-variance model and 4 kinds of energy futures, this paper uses MATLAB software to calculate, constructs the dynamic optimal investment ratio strategy from the nth day to the nth+252nd day, and takes the nth + 1st day and the nth+253rd day as the test interval, and tests the interval rolling k times to prove the effectiveness of the investment ratio strategy and the stability of the dynamic strategy. The effectiveness of capital strategy proves that it is a feasible way for investors to invest in energy futures. At the same time, it is proved that Markowitz mean-variance theory can be applied to a small number of energy futures portfolio.

An Average Residual Energy Based Clustering Routing Algorithm for Wireless Sensor Networks

YAN Li-juan, ZHANG Yan-hu

2020, 0(07): 16-20. doi:10.3969/j.issn.1006-2475.2020.07.004

Asbtract ( 159 )

PDF (1851KB) ( 151 )

References | Related Articles | Metrics

The problem of node energy limitation in wireless sensor networks has a serious impact on network performance and network life. From the perspective of energy optimization, aiming at the unbalanced energy consumption caused by clustering mechanism of LEACH protocol clustering algorithm, a new improved algorithm is proposed, which takes the average residual energy as the main parameter, selects the appropriate cluster head, and obtains the optimal cluster head position and the number of cluster heads from the base station based on the understanding of the whole network nodes. When selecting a new cluster head, it is important to consider whether the residual energy of the node is larger than the global average residual energy, and the distance between the node and all the selected cluster heads is greater than the set value. MATLAB software is used for simulation experiment. The improved algorithm can effectively avoid the premature death of a cluster head node due to excessive energy consumption. It can further balance the energy consumption of the network as a whole, increase the network throughput and extend the network life.

Data Center Dynamic Priority Multipath Scheduling Algorithm Based on SDN

XIAO Jun-bi, CHENG Peng, TAN Li-zhuang, MENG Xiang-ze

2020, 0(07): 21-26. doi:10.3969/j.issn.1006-2475.2020.07.005

Asbtract ( 234 )

PDF (965KB) ( 156 )

References | Related Articles | Metrics

With the development of cloud computing technology and distributed services, the “East-West” elephant flow in the data center has surged. This part of the elephant flow is prone to collisions due to improper scheduling, causing link congestion. This paper proposes a Dynamic Priority Multipath Scheduling algorithm (DPMS) based on Software-Defined Network (SDN). The algorithm develops an elephant flow and mouse flow scheduling model based on the characteristics of data center traffic, makes full use of redundant links between network nodes to improve resource utilization. Combined with the group table, the communication mode between the controller and the switch in the SDN architecture is optimized, and the packet processing delay is reduced. The experimental results show that DPMS improves network throughput and link utilization, reduces average flow completion time, and improves overall network performance compared with ECMP and Hedera scheduling strategies.

Handwritten Signature Identification Based on Wavelet Transform and CPN Network

JIA Jian-zhong

2020, 0(07): 27-31. doi:10.3969/j.issn.1006-2475.2020.07.006

Asbtract ( 196 )

PDF (773KB) ( 312 )

References | Related Articles | Metrics

As one of the important technologies in the field of biometric authentication, handwritten signature authentication technology has a wide application prospect. In order to improve the accuracy of handwritten signature verification, a method combining wavelet transform and CPN neural network is proposed. First, we take some preprocessing measures such as filtering and denoising, binarization, thinning, and normalization to the signature sample image, then the text image is decomposed by DB3 wavelet and the decomposed high pass coefficient matrix is extracted and treated as the features, then the CPN neural network classifier is used to train 7500 times for each training sample. Finally, the trained classifier is used to classify and identify the samples. On an experimental data set consisting of 36 identification experiment groups, the sample recognition accuracy of the method reached 93.48%. Comparative tests of various methods were used, the results show that the signature feature extraction of this paper is comprehensive and the recognition effect is better than the linear classifiers.

Security Risk Assessmenton of Attack Graph and HMM Industrial Control Network

CUI Wen-di, DUAN Peng-fei, ZHU Hong-qiang, LIU Na

2020, 0(07): 32-37. doi:10.3969/j.issn.1006-2475.2020.07.007

Asbtract ( 178 )

PDF (1015KB) ( 185 )

References | Related Articles | Metrics

In order to evaluate the network security risk of industrial control system and realize the effective defense of industrial control system, a risk assessment method based on attack graph and HMM is proposed to describe the network security status according to the change of attack behavior. Firstly, the industrial control network attack graph model is established, and the network attack is transformed into the network state migration problem. The network node association (NNC) is introduced to study the association of the industrial control network nodes, and further analyze the network security risks. Then the HMM establishes the relationship between network observation and attack state, and introduces the CVSS evaluation system to evaluate the security status of the industrial control system. Finally, a case study is carried out with the centralized control system of thermal power plant as the experimental background. The analysis results show that the method can comprehensively analyze the safety hazards of industrial control systems and provide a basis for safety management personnel to take effective preventive measures.

Research on Internet Robustness Based on Node Intentional Attacks

YANG Quan, DING Lin

2020, 0(07): 38-42. doi:10.3969/j.issn.1006-2475.2020.07.008

Asbtract ( 226 )

PDF (825KB) ( 114 )

References | Related Articles | Metrics

The robustness of real networks to deliberate attacks has always been an important issue in network science. By considering the actual Internet, the power function of node degree is used to define the initial load of nodes, and the cascading model under local load redistribution is constructed. The effects of two different attack strategies on network robustness are compared, and the influence of important network parameters on network robustness under intentional attack is studied. Through the numerical simulation experiment, the following conclusions are drawn: 1) When the initial load parameter is greater than a certain threshold, attacking high load nodes does harm to the network more than attacking low load nodes， but when the initial load parameter is less than the threshold, attacking low load nodes can destroy the network more effectively； 2) The smaller the initial load parameter, the larger the capacity parameter of a node, the stronger the robustness of the network. The research results of this paper can provide reference for the control and defense of cascading failures on the Internet.

Design of Distributed Resource Sharing Model Based on Blockchain

LI Xiao-yu, DUAN Peng-fei

2020, 0(07): 43-49. doi:10.3969/j.issn.1006-2475.2020.07.009

Asbtract ( 249 )

PDF (1215KB) ( 286 )

References | Related Articles | Metrics

Computer resource sharing provides an effective way to solve the imbalance of computer resource distribution and computing power demand, it also provides a way to solve the computer resource waste in the network. However, the traditional computer resource sharing system generally adopts the way of task allocation, users can hardly use the shared resource in a direct way. In addition, the task allocation and scheduling are managed by the central node. Once the central node is damaged, the resource sharing system will not function properly. In view of the above problems, this paper proposes a distributed resource sharing model. By delivering a private chain to bulid the underlying structure of the resource sharing model, we can directly provide users shared resources in P2P way. This paper analyzes the key implementation mechanism of resource sharing model, and builds a distributed computing network to truly realize computer resource sharing. We conduct experiments on CPU and memory utilization of the resource sharing model during normal operation and the results validate the effectiveness of the proposed resource sharing model in improving resource utilization and computing ability.

Substation Inspection System Based on HoloLens

WU Di, PAN Jian-qiao, YU Fang-zhao, PAN Bai-lang, GONG Yan-feng

2020, 0(07): 50-54. doi:10.3969/j.issn.1006-2475.2020.07.010

Asbtract ( 217 )

PDF (4194KB) ( 158 )

References | Related Articles | Metrics

In order to realize the intelligent inspection of substation, the design method of inspection system of substation based on HoloLens is proposed, the overall design framework of inspection system of substation is analyzed, the function modules of the system are constructed. The wearable AR / MR equipment is used to conduct the human-computer interaction and remote information collection in the inspection process of substation, so as to obtain the engineering file configuration of inspection system of substation. Vision sensing and bioelectric signal measurement sensing technology is used to sample and analyze the inspection image information of substation, and realize the design of visual information acquisition module. The near-field wireless communication is used to design the inspection network of substation, and the multimedia virtual reality enhancement technology is applied to 3D display the inspection output of substation, and the virtual scene generator, helmet display, the aligner of virtual scene and real scene are designed. In the process of substation inspection, users can observe and track the line of sight, and optimize the design of substation inspection system based on wearable HoloLens intelligent products. The test results show that the designed substation inspection system can accurately obtain information, the safety factor of substation inspection is high, it has good human-computer interaction and good practical application value.

JPS Jump Parallel Vertex Sampling Method Based on Online Social Network

ZHAO Qian-wen

2020, 0(07): 55-60. doi:10.3969/j.issn.1006-2475.2020.07.011

Asbtract ( 178 )

PDF (1180KB) ( 149 )

References | Related Articles | Metrics

In view of problems that the existing online social networks (OSNs) sampling methods cannot be effectively applied to low connectivity social networks, and the average degree of sample vertices seriously deviates from the original social network, vertex over sampling, based on the Metropolis-Hasting Random Walk (MHRW) sampling method, a Jump unbiased Parallel vertex Sampling (JPS) method is proposed by introducing double jump strategy, parallel mechanism and vertex buffer. The online social network data set is modeled as a social graph with vertices and edges for simulation sampling, and the sample vertex attribute graph is drawn by using Python / Matplotlib drawing library. The experimental results show that the sampling method is more effective for social graph with different connectivity, which improves the update rate of vertices in the sampling process, reduces the average deviation of sample vertices and can converge more quickly.

Short Text Sentiment Analysis Based on Self-attention and Capsule Network

XU Long

2020, 0(07): 61-64. doi:10.3969/j.issn.1006-2475.2020.07.012

Asbtract ( 267 )

PDF (760KB) ( 238 )

References | Related Articles | Metrics

Sentiment analysis of short texts is a challenging task. Aiming at the shortcomings of traditional convolutional neural networks and recurrent neural networks that can not fully obtain the semantic information contained in texts, this paper proposed a model that used the multi-head self-attention layer as the feature extractor and used the capsule network as the classification layer. The model can extract rich text information and has strong expressive ability. Experimental results on Chinese text showed that compared with the traditional deep learning method, the proposed model improved the accuracy of sentiment analysis. In the small dataset and cross-domain migration, compared with traditional method, the accuracy was greatly improved.

Multi-robot Scheduling Method in Intelligent Warehouse

WANG Zhen-ting, CHEN Yong-fu, LIU Tian

2020, 0(07): 65-70. doi:10.3969/j.issn.1006-2475.2020.07.013

Asbtract ( 355 )

PDF (1051KB) ( 334 )

References | Related Articles | Metrics

In recent years, the traditional storage system has been unable to meet the increasing demand and has gradually turned to intelligent storage. Aiming at the scheduling problem of robots in intelligent warehouse and optimizing the turning times, distance cost and the maximum task waiting time, a scheduling algorithm for both task assignment and path planning is proposed. To ensure that the tasks assigned to each robot are not repeated, tasks are assigned with genetic algorithm and tasks are assigned for multiple mobile robots. Then Q-learning algorithm is used to carry out path planning for tasks assigned by the robot. The path is constrained according to the account of turns and the cost of the distance, and the penalty value is set for the turning of the path and the feasible action in each step. Finally, a path with less turning times and shorter travel is formed. The effectiveness of the proposed algorithm is verified by comparing it with other algorithms.

E-government Text Similarity Evaluation Model Based on Do-Bi-LSTM Model

LI Fan, BAI Shang-wang, DANG Wei-chao, PAN Li-hu

2020, 0(07): 71-75. doi:10.3969/j.issn.1006-2475.2020.07.014

Asbtract ( 169 )

PDF (777KB) ( 177 )

References | Related Articles | Metrics

In view of the inefficiency of manual approval texts in current government systems, this paper introduces text similarity into e-government. In the current network model based on text similarity, there is a huge matrix of generated word vectors, which requires a lot of time to train, and only uses the context of the context to generate word vectors, ignoring the relationship between the word order and semantics of the document. In order to improve efficiency and reduce training cost, this paper proposes a Do-Bi-LSTM text similarity calculation method, which first converts the text in the training data set into a vector through the Doc2vec language model. This method adds a text vector on the basis of the word vector, so can capture the interrelationship between sentences and between paragraphs. Then the obtained vector is trained as the input of the Bi-LSTM network model. Finally, compared with the LSTM network model and the traditional deep network model, the experiment shows that the accuracy of the method is greatly improved and feasible.

Voice Service Quality Inspection Based on LSTM Network

WU Peng, GUO Xiao-yun, CHEN Peng, WANG Zong-wei, CAO Lu, JIN Peng

2020, 0(07): 76-79.

Asbtract ( 119 )

PDF (735KB) ( 131 )

References | Related Articles | Metrics

To solve the problem of low quality and low information processing capability of 95598 customer service center voice service quality inspection, a recommendation technology based on LSTM network was proposed. Combining the indicators of traditional sampling methods with indicators related to deep learning methods, the LSTM network was used to explore the inner relationship among various indicators in space and time. Instead of random sampling method, a representative recommendation quality inspection was used and combined with the characteristics of low problem probability in voice service to improve model performance. The experimental results illustrate that the improved model can significantly improve the quality inspection efficiency and directivity.

A Short Text Classification Method Based on Emotional Features

ZHOU Ling, ZHANG Ying-jun, PAN Li-hu

2020, 0(07): 80-84. doi:10.3969/j.issn.1006-2475.2020.07.016

Asbtract ( 185 )

PDF (962KB) ( 198 )

References | Related Articles | Metrics

With the emergence of a large amount of short texts, using short text classification technology to mine a large amount of effective information in short text has become a hot topic of research. For the feature selection method in the current classification process, which only considers the word frequency, and the short text is short in length and sparse keywords, the paper proposes a short text classification method based on emotional features, combined with TF-IDF, the weight of the feature words is modified with the 〖JP2〗sentiment dictionary, which can effectively improve the weight of the feature words with distinguishing ability, and avoid the problem of low accuracy caused by traditional methods which do not consider emotion but only word frequency. Using the Chinese corpus of teacher Tan Songbo for short text classification, through comparative experiments, the effectiveness of the method is verified.

Study on Association Model Between Cardiovascular Disease Complications and Frailty

FENG Yun-xia, HAN Zheng-liang, XUE Rong-rong, SONG Bo

2020, 0(07): 85-89. doi:10.3969/j.issn.1006-2475.2020.07.017

Asbtract ( 144 )

PDF (790KB) ( 101 )

References | Related Articles | Metrics

In the study of cardiovascular disease, its complications have a certain correlation with the frailty. Studying the association pattern between cardiovascular disease complications and frailty is conducive to the formulation of clinical decisions and the prevention of complications. However, the correlation between complications and frailty is complicated, which leads to the inefficiency and inaccuracy of the classical Apriori algorithm. This paper proposes an improved algorithm called HI-Apriori. The HI-Apriori algorithm introduces the Hash table and the improved rate into Apriori. Firstly, the improved rate is used to prun the unreliable itemsets in the candidate set, and then the Hash table is stored to store the reliable item set. According to experimental data, the HI-Apriori algorithm improves the operational efficiency by at least 60%, and from the association rules, the above-mentioned frailty can increase the probability of cardiovascular disease in patients with diabetes.

Student Specific Behavior Recognition Based on Improved YOLOv3 Network

WANG Chun-hui, WANG Quan-min

2020, 0(07): 90-96. doi:10.3969/j.issn.1006-2475.2020.07.018

Asbtract ( 175 )

PDF (3759KB) ( 158 )

References | Related Articles | Metrics

In order to improve the detection accuracy of convolutional neural networks in student behavior recognition applications, this paper uses K-means clustering to cluster the unique data sets to obtain more adaptive anchor box, and proposes a YOLOv3 network based on improved loss function. The network model dynamically transforms the original squared loss function weights, focusing on the calculation of the loss of continuous variables. The new loss function can effectively reduce the influence of the gradient disappearance of the sigmoid function, making the model converge more quickly. The experimental results show that the deep convolutional neural network based on the improved loss function has improved the recognition of the three poses of “lookup”, “lookdown” and “talk”.

Video Action Recognition in Complex Background Based on Deep Learning

PAN Chen-ting, TAN Xiao-yang,

2020, 0(07): 97-103. doi:10.3969/j.issn.1006-2475.2020.07.019

Asbtract ( 200 )

PDF (2033KB) ( 165 )

References | Related Articles | Metrics

Recognizing human actions in videos has broad application prospects and great potential economic value. However, the accuracy of video action recognition is affected by a number of factors such as swaying, background changes, camera shaking and moving shadows. To reduce the influence of such complex background, we proposed non-local temporal segment networks (NLTSNet). The NLTSNet is based on the temporal segment network but is enhanced with non-local modules over the ResNet so as to capture the non-local spatial and temporal information contained in the video clips. To furthermore improve the network’s robustness against stationary cluttered background, we integrate the optical flow into the non-local module. Finally, we adopt a learnable ensemble network to fuse the prediction results from both the appearance and temporal modality. Extensive experimental results on the TDAP dataset show that our new method can recognize human actions with more accuracy in a complex background compared with several state of the art methods, without increasing the time complexity.

Accurate Extraction of Buildings from Remote Sensing Images Based on Improved Markov Random Field

ZHU Qia, WANG Jian, LIU Xing-yu, ZHOU Zai-wen, MA Zi-wen, GAO Xian-jun

2020, 0(07): 104-110. doi:10.3969/j.issn.1006-2475.2020.07.020

Asbtract ( 149 )

PDF (3232KB) ( 148 )

References | Related Articles | Metrics

With the rapid development and wide application of remote sensing images, the extraction of buildings from remote sensing images can extract building information timely and accurately, which has important research significance in some applications such as rapid map updating and city management. At present, some problems such as image blurring and the improper classification of buildings exist in the gray scale image of feature map. The gray scale images need to be transformed into binary images before it can be used for the subsequent work. In order to improve classification accuracy, based on the preliminary extraction of neural network, this paper first uses the OTSU method to segment the gray scale images and uses morphological methods to process gray scale images. And it improves the Markov Random Field method and proposes a new method that can dynamically estimate the prior parameter β according to the local neighborhood characteristics of the images, introducing the original image characteristics into the Markov Random Field, further segmenting the results which is produced by the segmentation by the OTSU method and correcting the jagged boundaries of building edges in the images to improve classification accuracy. The experimental results demonstrate that the method can effectively reduce the building areas which are classified wrongly in the gray scale images extracted by neural network.

Particle Size Detection of Sandstone Images Based on Full Convolutional Network

ZHU Da-qing, CAO Guo

2020, 0(07): 111-116. doi:10.3969/j.issn.1006-2475.2020.07.021

Asbtract ( 236 )

PDF (6383KB) ( 196 )

References | Related Articles | Metrics

In order to segment the tightly adhering sandstone and obtain the particle size of sandstone accurately, a particle size measurement method based on two-stage deep learning is proposed. This method uses image processing technology to preprocess the sandstone image, and then uses the first-stage segmentation model to segment the sandstone objects. After morphological processing of segmented objects, as many sandstone objects are connected closely, the second-stage separation model is adopted to separate the sandstone objects, then the result graph of segmented and separated is obtained. Finally, the longest diameters of the sandstone objects are calculated and the average particle size of the sandstone image is obtained. Experiments show that this algorithm can segment the closely connected sandstone objects quickly and accurately, and improve the speed and accuracy of sandstone particle size calculation.

Image Inpainting Technology Based on Total Generalized Variation Regularization

CHEN Ying-pin, KE Su-ling, HUANG Hui-ying, WU Ri-sheng, WANG Ling-zhi

2020, 0(07): 117-120. doi:10.3969/j.issn.1006-2475.2020.07.022

Asbtract ( 187 )

PDF (3016KB) ( 115 )

References | Related Articles | Metrics

Image inpainting is a hot spot in digital image processing and computer graphics. To better recover the image, a new image inpainting model based on total generalized variation is proposed. Then, the first order primal dual method is employed to solve the proposed model. The experiments are carried out to verify the presented scheme based on the value of similarity index and the peak signal to noise ratio. The experimental results show that the proposed scheme is capable of achieving good image recovery quality.

Surface Quality Visual Inspection of Turbine Shell Based on Robotic Arm

ZHANG Hui, YU Hou-yun, LI Ke-bin

2020, 0(07): 121-126. doi:10.3969/j.issn.1006-2475.2020.07.023

Asbtract ( 189 )

PDF (3356KB) ( 163 )

References | Related Articles | Metrics

In order to solve the problems of low accuracy and efficiency, and high difficulty to realize in the surface quality detection of turbine shell parts, a method of surface quality detection based on robotic arm and machine vision is proposed. Firstly, the robotic arm is used to drive the vision system to the detecting work station to collect the surface image of the part. Then the image is processed by image filtering, detection region extraction, feature extraction and segmentation. Finally, the defects such as bumps and pits on the surface of the turbine shell are detected. The field test results show that the average detection time of the single work station is less than 2 s, and the missing rate of the detection is only 0.4%, which proves that the accuracy and efficiency of the system meet the requirements of engineering application.

Table of Content