Table of Content

31 March 2021, Volume 39 Issue 2

Previous Issue    Next Issue

Cuting-Edge Information Technology of Intelligent Transportation

Short-Term Traffic Flow Prediction Method of Different Periods Based on Improved CNN-LSTM

LI Lei, ZHANG Qingmiao, ZHAO Junhui, NIE Yiwen

2021, 39(2):  185-198.  doi:10.3969/j.issn.0255-8297.2021.02.001

Asbtract ( 2620 )   PDF (1339KB) ( 755 )  

References | Related Articles | Metrics

Aiming at solving the problem that existing prediction models could not fully extract the spatio-temporal features in traffic flow, we proposed an improved convolutional neural network (CNN) with long short-term memory neural network (LSTM) for shortterm traffic flow prediction. First of all, a layered extraction method was used to design the network structure and one-dimensional convolution kernel which enabled automatic extraction of spatial features of traffic flow sequences. Second, the LSTM network modules were optimized to reduce the long-term dependence of network on the data. Finally, the optimization algorithm for rectified adaptive moment estimation (RAdam) was introduced to the end-to-end model training process, which accelerated fitting effects of the weight and improved the accuracy and robustness of network output. Experimental results showed that compared with the prediction model of stacked auto-encoders (SAEs) network, performance of the proposed model was enhanced by 3.55% and 8.82% on weekdays and weekends with model running times reduced by 6.2% and 6.9%, respectively. Compared with the prediction model of long-short term memory-support vector regression (LSTM-SVR), its performance was enhanced by 0.29% and 1.79% with model running times reduced by 9.0% and 9.7%, respectively. Therefore, the proposed model was more applicable to the short-term traffic flow prediction of different time periods.

A Charging Vehicle Scheduling Scheme with Traffic Road Restrictions

ZHONG Ping, CHEN Yuanming, DU Zhicheng, LI Lin, GUI Lin

2021, 39(2):  199-209.  doi:10.3969/j.issn.0255-8297.2021.02.002

Asbtract ( 1836 )   PDF (632KB) ( 96 )  

References | Related Articles | Metrics

Charging scheduling is a very important item for wireless rechargeable sensor networks. Existing researches mainly focus on scheduling charging vehicles to obtain the optimal mobile path. However, these algorithms cannot provide good performance when traffic is restricted. Considering the mobile charging vehicle scheduling problem with traffic road constraints, this paper proposes a mobility constrained charging scheduling scheme (MCCS). To better fit the actual scene, we formalize the problem as an edge coverage problem, and enhance the classical MAENS algorithm by adding a path decomposition operator and a mutation operator. The performance of MCCS is evaluated by extensive simulations. Compared with MAENS, experimental results show that MCCS achieves superior performance in terms of low average energy consumption and high charging stability.

New Location Algorithm Based on Sparse Grid Optimization in C-V2X

XIA Xiaohan, CAI Chao, QIU Jiahui, YANG Jingyuan, ZHANG Xiangyun, XIAO Ran

2021, 39(2):  210-221.  doi:10.3969/j.issn.0255-8297.2021.02.003

Asbtract ( 1806 )   PDF (3427KB) ( 199 )  

References | Related Articles | Metrics

The location algorithm in cellular-V2X (C-V2X) has always been one of the important technical approaches for the development of vehicle-road collaboration and autonomous driving. Currently, in the vehicle-road collaboration scenarios of autonomous driving services, many positioning solutions including base stations and GNSS meet challenges in many aspects such as positioning accuracy, positioning processing delay and deployment cost. In response to these problems, a fingerprint location algorithm is proposed for C-V2X based on statistical information grid (STING) algorithm for grid optimization and extreme gradient boosting decision tree (XGBoost). Compared with traditional fingerprint positioning methods, the positioning accuracy and calculation rate are optimized after grid optimization. The new method is more suitable for vehicle-road collaboration scenarios, and provides an effective positioning method for C-V2X scenarios.

Communication Engineering

Research on Insulator Self Exploding Detection in UAV Inspection Based on Deep Learning

WANG Wanguo, MU Shiyou, LIU Yue, LIU Guangxiu, LANG Fenling

2021, 39(2):  222-231.  doi:10.3969/j.issn.0255-8297.2021.02.004

Asbtract ( 2032 )   PDF (10566KB) ( 322 )  

References | Related Articles | Metrics

Insulator self-exploding detection is an important part of UAV inspection. Accurate, rapid and automatic searching for insulator self-exploding areas can greatly save the workload of UAV inspection data processing and improve inspection accuracy and efficiency. Aiming at the problem of low sample size, small target and low precision in the current insulator self-exploding detection, this paper proposes a deep learning self-exploding detection method for UAV inspection insulators. The method uses a large number of collected insulator samples to train the deep learning detection model, and then uses the computer vision method to detect the self-exploding region in the detected insulator. The method of this paper synthesizes the advantages of deep learning in detecting complex targets and the fact that computer vision does not require a large number of samples and can detect small targets. Experiments show that the detection accuracy of this algorithm can reach 84.8%. It has positive significance and application value for insulator self-exploding detection.

Design of the RT IP Core for Satellite Payload Data Bus

LIU Wenting, WAN Xiaolei, XU Nan, YANG Tong, CHEN Liangliang

2021, 39(2):  232-240.  doi:10.3969/j.issn.0255-8297.2021.02.005

Asbtract ( 2000 )   PDF (2242KB) ( 90 )  

References | Related Articles | Metrics

Based on serial bus protocol, a remote terminal IP (intellectual property) core for satellite payload data bus is proposed. The IP core is composed of bus interface module, frequency divider module, protocol processing module and data collecting module. The IP core design is tested and verified by functional simulation and FPGA (field programmable gate array) test, and has been taped out successfully. Test results indicate that the designed IP core has good performance in capability, reliability and less resource occupancy. It can be used for testing the payload data bus in satellite systems or in verification equipment of satellite systems.

Development of NB-IoT Based Intelligent LED Light Pole Monitoring System

JIN Yan, MAO Minmin, XU Qiuyu, OUYANG Yuling, JU Jiaqi

2021, 39(2):  241-249.  doi:10.3969/j.issn.0255-8297.2021.02.006

Asbtract ( 2018 )   PDF (1590KB) ( 279 )  

References | Related Articles | Metrics

In order to solve the problems of traditional LED light pole in remote controlling, automatic inspection, real-time single light regulation and control, and fault location recognition, an intelligent LED light pole monitoring system based on narrow band internet of things (NB-IoT) is proposed in this paper. In the system, STM32L151 micro controller unit (MCU) is adopted as the processor, together with multiple sensors to realize signal acquisition of street lamp. Through the connection between NB-IoT modules and core network, signals are collected and uploaded to OneNET cloud platform. And with a developed mobile application (APP) and a personal computer (PC) monitoring interface, fault location information of street lamps can be obtained in real time. Experimental results show that the developed system can not only achieve real time monitoring and control of street lights, accurate positioning of faulty street lights, but also realize the individual control of street lights.

Q-learning Based Relay Selection Strategy for Hybrid Satellite-Terrestrial Cooperative Transmission

WANG Xiaoxiao, KONG Huaicong, ZHU Weiping, LIN Min

2021, 39(2):  250-260.  doi:10.3969/j.issn.0255-8297.2021.02.007

Asbtract ( 1892 )   PDF (310KB) ( 108 )  

References | Related Articles | Metrics

Cooperative relay networks can achieve spatial diversity, but their system performances heavily depends on relay selection schemes. To solve this problem, a hybrid satellite-terrestrial cooperative network relay selection strategy based on Q-learning is proposed. First, under the consideration that all the relay nodes employ amplify-and-forward protocol, the end-to-end output signal-to-noise ratio after combining the maximal ratio is derived. Next, the state, action and reward function of Q-learning are set to select the relay node with the greatest cumulative return. Then, in order to traverse all states, Boltzmann selection policy is induced to select action by probability approach, so that the source node can explore all states and find the optimal one. Finally, the optimal transmission power is obtained by using power allocation scheme between the selected relay node and the source node. Simulation results show that, compared with the random relay selection algorithm, the proposed strategy greatly improves the system performance.

Signal and Information Processing

Research on Adaptive Speech Enhancement Method for Microphone Array Based on Convex Combination

ZHAO Yibo, LU Haozhi, LI Shuhui, YAN Tao

2021, 39(2):  261-271.  doi:10.3969/j.issn.0255-8297.2021.02.008

Asbtract ( 1841 )   PDF (750KB) ( 125 )  

References | Related Articles | Metrics

Generalized sidelobe canceller is a widely used microphone array speech enhancement method, which has obvious noise reduction effect on speech signals with Gaussian noise. However, when impulse noise is mixed in the speech signal, the recognizability of the enhanced speech signal becomes significantly worse as using this method. In order to improve the noise reduction effect on speech signals with impulse noise, in this paper, an adaptive speech enhancement method for microphone array based on convex combination is presented. In this method, a single linear adaptive filter is replaced by a convex combined filter composed of linear filter and nonlinear spline filter, and the maximum correntropy criterion is used to update the weight of the convex combination filter and the control point factor in the nonlinear spline filter. Results show that the presented speech enhancement method can effectively filter the Gaussian noise and impulse noise in the speech signal, and the speech enhancement effect is extremely obvious.

Attention Guided 3D ConvNet for Aerial Scene Change Detection

ZHANG Han, QIN Kun, BI Qi, ZHANG Ye, XU Kai

2021, 39(2):  272-280.  doi:10.3969/j.issn.0255-8297.2021.02.009

Asbtract ( 1922 )   PDF (4717KB) ( 225 )  

References | Related Articles | Metrics

With high tolerance to the great amount of noise and precise depiction of image changes in high resolution remote sensing images (HRRSI), scene-level change detection strategy makes it possible to detect changes in HRRSI. In this paper, we propose an attention guided 3D ConvNet for HRRSI change detection. Firstly, we develop a simplified 3D AlexNet to extract convolutional features. Then we add a semantic attention module (SAM) to further extract the discriminative regions which strongly relate to land-cover changes. Finally, the refined features are fed into a classification layer to organize the whole framework in an end-to-end trainable manner. Scenes in different phases are put into the convolutional neural network (CNN) with the result of change detection. In order to evaluate the performance of scene level change detection methods, we create a public semantic level high resolution remote sensing images change detection benchmark. Experimental results on this dataset are obviously better than other related methods, demonstrate the effectiveness of our method, and show the prospect of scene level remote sensing change detection based on deep learning.

Orientation Method for Rail Weld Region Based on Level Set

LIU Xingwu, XIONG Bangshu, LIAO Feng, CHEN Xinyun

2021, 39(2):  281-292.  doi:10.3969/j.issn.0255-8297.2021.02.010

Asbtract ( 1957 )   PDF (11140KB) ( 320 )  

References | Related Articles | Metrics

An orientation method for rail weld region based on level set is proposed to improve the adaptability, stability and accuracy of weld positioning under different illumination conditions. Firstly, in order to separate welds from rail waists, rail heads and background, level set is used to segment the contours in preprocessed weld image. Secondly, area sorting and domain connecting are used in combination to eliminate contour interference and achieve coarse positioning of weld contour. The weld contour is then accurately positioned by using double sorting method. Finally, the rail weld region is automatically positioned by sorting the abscissa of weld contour. Positioning experiments for the weld region of 60kg/m rail are conducted under different illumination conditions. Experiments demonstrate the advantages of strong adaptability, high accuracy and good stability, and prove that the proposed method can be used for automatically detecting the weld misalignment in welded rail site.

Improved Steganography Algorithm Based on J-UNIWARD

WU Qian, WU Jianbin, LIU Zixuan, SONG Mengli

2021, 39(2):  293-301.  doi:10.3969/j.issn.0255-8297.2021.02.011

Asbtract ( 1908 )   PDF (251KB) ( 208 )  

References | Related Articles | Metrics

In order to further improve the security of adaptive steganography algorithms, this article introduces the idea of image blocking, rewrites the distortion function of the original J-UNIWARD algorithm, and changes additive distortion functions to non-additive distortion functions. The implementation process of the algorithm is as follows: The carrier image to be processed is divided into four sub-blocks, and the texture complexity of each sub-block is calculated separately, under the constraint of keeping total embedding amount unchanged. The more complex blocks are preferentially chosen to be embedded. By recalculating the distortion function after each block-embedding, the embedding amount is dynamically adjusted according to the complexity. Then the secret information is adaptively embed into the texture area by STC (syndrome trelliscodes) encoding. Finally, detection performance is analyzed by using DCTR and ccJRM steganalysis techniques. Experimental results show that under the same capacity, the proposed algorithm can significantly improve the anti-stealth analysis ability of the algorithm.

Data Augmentation Method Based on Image Gradient

LIU Zhiyu, ZHANG Shufen, LIU Yang, LUO Changyin, LI Min

2021, 39(2):  302-311.  doi:10.3969/j.issn.0255-8297.2021.02.012

Asbtract ( 1828 )   PDF (2913KB) ( 78 )  

References | Related Articles | Metrics

As used in classification of image recognition, convolutional neural network requires large-scale image data set for training. Due to the limitation of the number of target images to be collected and the conditions of image acquisition equipment, it is difficult to obtain enough image samples by conventional methods because of time-consuming, laborconsuming and money-consuming. In order to solve the insufficiency of image samples, a variety of sample enlargement methods have been proposed. This paper introduces the research background and significance of data augmentation. For the purpose of improving the accuracy of image recognition of convolutional neural network, a data augmentation method based on image gradient is proposed. The image gradient is selected to increase image sample and enlarge image data set by precise clipping method, and the convolutional neural network is trained with the expanded data set. By using Tensorflow deep learning framework and VGG16 network model, and selecting some data sets of PlantVillage, the training set data can be expanded to 6 times of the original. The training set before and after the expansion is trained and compared. Experimental results show that the accuracy rate of the model trained by the training set after data augmentation is increased by 4.18%.

Land-Use Information of Object-Oriented Classification by UAV Image

MA Feihu, XU Fadong, SUN Cuiyu

2021, 39(2):  312-320.  doi:10.3969/j.issn.0255-8297.2021.02.013

Asbtract ( 1940 )   PDF (10409KB) ( 268 )  

References | Related Articles | Metrics

In order to effectively classify the rural land, an object-oriented classification method is selected to extract the land classification information of drone aerial photography images. First, original drone-taking images are preprocessed, then by repeatedly performing segmentation tests on the study area, the optimal segmentation scale of each feature is selected, with which the images are segmented at different levels. And based on feature differences in feature vector, spectrum, shape, etc., the most suitable classification rules are established for the features on the optimal segmentation scale layer. Accordingly, the land use information of each layer can be extracted. Experimental results with 734 samples for accuracy verification show that the overall classification accuracy of multi-scale and multi-level segmentation classification reaches 84.20%, and the kappa coefficient is 0.8062, whereas the overall accuracy of single-scale segmentation classification is only 77.11%, and the kappa coefficient is 0.7214. It shows that the data used in this study and the classification accuracy of the categories inside the region are higher.

A Single Image Super-Resolution Method Based on the Dual Network Model

NI Cui, WANG Peng, ZHANG Guangyuan, LI Kefeng

2021, 39(2):  321-329.  doi:10.3969/j.issn.0255-8297.2021.02.014

Asbtract ( 1812 )   PDF (11394KB) ( 259 )  

References | Related Articles | Metrics

This article mainly improves the efficient sub-pixel convolutional neural network (ESPCN) algorithm in the field of deep learning. By adding residual network knowledge and adjusting original ESPCN structure, a dual network model is proposed for single frame image super-resolution reconstruction method. Experimental results show that this algorithm can effectively improve the accuracy of single-image super-resolution reconstruction and enrich the detailed information after reconstruction.

Computer Science and Applications

Fingerprint Recognition System Based on Editable Blockchain

ZHU Yanyan, LI Sheng, FENG Guorui, ZHANG Xinpeng

2021, 39(2):  330-337.  doi:10.3969/j.issn.0255-8297.2021.02.015

Asbtract ( 1907 )   PDF (2634KB) ( 164 )  

References | Related Articles | Metrics

Fingerprint identity authentication system has been widely used in access control, payment, public security and other fields. Existing systems typically store original fingerprint images or features in a database to identify or authenticate users’ identity. Fingerprint data in the database is at risk of being attacked or tampered with. In order to solve this problem, this paper proposes a fingerprint identification system based on editable blockchain. Firstly, we build a private chain environment, achieve multi-node cluster interconnection, and then calculate the fingerprint hash and store it in the blockchain. In order to facilitate the administrator to update the users in the fingerprint identification system, this paper uses the chameleon hash algorithm to calculate the hash of the constructed private chain block. The administrator who owns the chameleon hash private key can edit the information in the block body to implement deletion or modification of the user fingerprint data without changing the blockchain structure. Experiments show that the proposed system has good real-time performance and high accuracy of fingerprint recognition.

D2D Network Resource Allocation Based on Joint Interference Control and PSO

LIU Yuheng, PENG Yi, FU Xiaoxia, AN Haojie

2021, 39(2):  338-346.  doi:10.3969/j.issn.0255-8297.2021.02.016

Asbtract ( 1775 )   PDF (257KB) ( 99 )  

References | Related Articles | Metrics

Address to interference problems faced by D2D (device-to-device) communication users in the process of multiplexing traditional cellular communication channel resources, a resource allocation algorithm based on particle swarm optimization (PSO) algorithm for D2D communication power matching joint interference control is proposed in this paper, which maximizes the throughput of the whole system while controlling interference. Firstly, security communication are divided into different ranges according to threshold values of user-powers, and only the users in the security communication range have opportunity to join in subsequent power distribution, thus effectively reducing system interference. Secondly, a power matching method based on PSO is adopted to take D2D senders as particle swarm. It maximizes the throughput of the whole system by searching for the optimal power through iteration. Simulation results show that the proposed algorithm can significantly improve the overall throughput of the system and minimize the interference, so as to achieve the optimal communication quality of the whole system.