Table of Content

15 May 2020 Volume 33 Issue 5

  

Image Super-resolution Algorithm Based on SqueezeNet Convolution Neural Network

QIN Xing,GAO Xiaoqi,CHEN Bin

Electronic Science and Technology. 2020, 33(5):  1-8.  doi:10.16180/j.cnki.issn1007-7820.2020.05.001

Abstract ( 441 )   HTML ( 22 )   PDF (1495KB) ( 106 )  

Figures and Tables | References | Related Articles | Metrics

In order to effectively improve the resolution of depth image, the study proposed a convolutional neural network model based on the Fire Module by referring to the classic SqueezeNet network structure. The proposed algorithm implemented mapping and transformation directly from low-resolution images to high-resolution images. As a nonlinear mapping module of the network, Fire Module learned the deep features of the image while reducing the parameters. To avoid interpolation preprocessing, a deconvolution layer was introduced in the output layer of the network to achieve a final 3 times up-sampling and high resolution image output. Experiments showed that the super-resolution image obtained by the deconvolution algorithm of the convolutional neural network model based on Fire Module was richer in detail, and the evaluation of objective index PSNR value and SSIM value was also superior to other algorithms.

A Method of Weld Defect Classification Based on PCA and SVM

WAN Dongyan,XU Zhenying,YANG Qing,WU Mengqi

Electronic Science and Technology. 2020, 33(5):  9-14.  doi:10.16180/j.cnki.issn1007-7820.2020.05.002

Abstract ( 270 )   HTML ( 9 )   PDF (1121KB) ( 36 )  

Figures and Tables | References | Related Articles | Metrics

In order to ensure that the welding structure is in safe working state, a classification method combining PCA and SVM was proposed to classify the weld feature guided wave defect signals. Firstly, based on the sparse reconstructed signal of defect echo, the feature parameter matrix of defect signal was extracted, and the dimensionality of parameter matrix was optimized by PCA to eliminate redundant information. Then, the optimized low-dimensional feature matrix was applied to SVM for classification training, and the classification effect of different principal components was compared. Kernel function and related parameters were selected to improve the accuracy of the classifier. The experimental results showed that the method could effectively classify weld defects.

Research on Transient Temperature Field of Motor Based on Multifield Coupling

SONG He,WU Yaohui,WU Haozhen

Electronic Science and Technology. 2020, 33(5):  15-20.  doi:10.16180/j.cnki.issn1007-7820.2020.05.003

Abstract ( 254 )   HTML ( 9 )   PDF (967KB) ( 36 )  

Figures and Tables | References | Related Articles | Metrics

To solve the transient calculation of temperature field problem caused by unstable heating of the motor under complex working conditions, a two-dimensional coupling model of electromagnetic field and temperature field was established according to the law of unstable heat. Based on the structural characteristics of small type asynchronous motor, the structural parameters, boundary condition parameters, electrical and thermal physical parameters calculation, equivalent and configuration of the electromagnetic field-thermal field two-dimensional model were performed. According to the law of heating variation under complex working conditions, the coupling frequency and heat source coupling mode were designed, and the relevant parameters were calculated and configured. The transient temperature field calculation of the motor under unstable load was realized, and the actual maximum temperature distribution of each part of the motor was investigated. The temperature rise experiment was carried out through the scheme design and construction of the experimental platform. The simulation and experimental results of the transient temperature field of the motor showed that the rationality of parameters equivalence of the two-dimensional model and multi-field coupling method could accurately calculate the full transient temperature field of the motor under complex conditions.

Hyperspectral Image Denoising via Spectral Weighted Low-rank Matrix Approximation

LIU Lu,ZHANG Hongyan,ZHANG Liangpei

Electronic Science and Technology. 2020, 33(5):  21-27.  doi:10.16180/j.cnki.issn1007-7820.2020.05.004

Abstract ( 379 )   HTML ( 7 )   PDF (1167KB) ( 33 )  

Figures and Tables | References | Related Articles | Metrics

HSIs are often contaminated by various types of noise, which degrades the quality of the acquired image and limits the subsequent application. Furthermore, the noise of HSI appears different statics and intensity in different bands. In this paper, a spectral weighted low-rank approximation model was proposed for hyperspectral image denoising. The spectral weighted matrix was introduced to balance the data fidelity of the different bands in consideration of their different noise intensity. To further separate the noise from the clean image, weighted nuclear norm minimization was utilized to depict the patch-wise low-rank structure of the high dimensional HSI. The proposed model was formulated into a linear equality-constrained problem and solved by alternating direction method of multipliers. Experimental results on both simulated and real HSI datasets validated the effectiveness and superiority of the proposed method.

An Image Retrieval Algorithm Based on Improved Hashing Method

LU Chaowen,LI Feifei,CHEN Qiu

Electronic Science and Technology. 2020, 33(5):  28-32.  doi:10.16180/j.cnki.issn1007-7820.2020.05.005

Abstract ( 280 )   HTML ( 8 )   PDF (870KB) ( 40 )  

Figures and Tables | References | Related Articles | Metrics

The coding methods of the traditional visual features adopted in current image retrieval approaches lack sufficient learning ability and have no strong feature expression ability.In addition, due to the high dimensionality of visual features, a large amount of memory is consumed, thus reducing the performance of image retrieval. In this paper, an image retrieval algorithm with end-to-end training based on deep and improved hashing method was proposed and designed. The proposed algorithm combined the high-level features extracted by CNN with Hash function and learned Hash codes with expression ability to perform large-scale image retrieval in low-dimensional Hamming space. The experimental results on two main datasets showed that the retrieval performance of the proposed method was superior to that of some state-of-the-art ones.

Diagonal Gait Simulation Analysis of a Quadruped Robot

CHEN Jiupeng,SAN Hongjun,ZHANG Daoyi,HU Qiongqiong,ZHANG Kaixiang,XIONG Binzhou

Electronic Science and Technology. 2020, 33(5):  33-38.  doi:10.16180/j.cnki.issn1007-7820.2020.05.006

Abstract ( 500 )   HTML ( 17 )   PDF (1004KB) ( 60 )  

Figures and Tables | References | Related Articles | Metrics

The bionic quadruped robot has become the focus of bionics research because of its simple control, strong adaptability to various road conditions, and environmental adaptability. Aiming at the complexity of quadruped robot model and the difficulty of diagonal gait control, a quadruped robot simulation platform was established in the proposed study. The quadruped robot model was built from SolidWorks and its kinematics equation was obtained. The diagonal walking of the robot was realized by the method of ADAMS and Simulink. The simulation showed that the robot had a forward distance of about 0.65 m in 7 s and a small amount of centroid fluctuation in the vertical direction, which proved that the robot could achieve stable operation and laid a solid foundation for robot dynamics and trajectory planning.

Isogeometric Analysis of Complex Planar Porous Model

CHEN Yu,CHEN Long

Electronic Science and Technology. 2020, 33(5):  38-44.  doi:10.16180/j.cnki.issn1007-7820.2020.05.007

Abstract ( 281 )   HTML ( 8 )   PDF (1188KB) ( 43 )  

Figures and Tables | References | Related Articles | Metrics

The NURBS-based geometric analysis method commonly use the same geometric description in the model design and analysis stage, which has many advantages. However, due to the limitations of the NURBS surface, it is difficult to effectively model and analyze the porous model. In order to solve the above problems, a segmentation-splicing method was adopted in this paper. In the modeling process of the porous model, the model was divided into several surface patches which were easy to be built by NURBS by using manual division and automatic division. In the analysis process, the surface patches were spliced and the control points were automatically numbered. Modeling and isogeometric analysis of the porous model was achieved in the case of C₀ continuous. Comparing the analysis results with the results obtained by ANSYS, the results of the two methods were similar, which proved the effectiveness of the new method.

FPGA-Based Real-Time Edge Detection and its Implementation for Deep-Space Images

SHI Weizhong,CAO Weiwei,FAN Yanming,DONG Jiajun,CHEN Shu,XIAO Hao

Electronic Science and Technology. 2020, 33(5):  45-49.  doi:10.16180/j.cnki.issn1007-7820.2020.05.008

Abstract ( 442 )   HTML ( 7 )   PDF (841KB) ( 55 )  

Figures and Tables | References | Related Articles | Metrics

Navigation control of spacecraft is one of the key technologies in deep space exploration. However, As the exploration distance is getting farther, the limitations of traditional ground station control become more obvious. Therefore, autonomous optical navigation has become the development direction of aerospace exploration in the future. The edge detection technology is one of the critical algorithms in optical autonomous navigation. In order to meet the real-time requirement of on-board computing, this study proposed an optimized Canny edge detection algorithm and the circuit architecture of FPGA to optimize the non-maximum suppression and adopt dynamic single threshold of Canny edge detection algorithm. In this way, it could be implemented on the FPGA with less resource occupation and pipeline architecture. Finally, the real-time requirement of optical autonomous navigation was satisfied on the premise of guaranteeing the accuracy of edge extraction. and it had good robustness to complex star targets.

Analysis of Ripple Current in Low-Capacitance Single-Phase Cascaded H-bridge STATCOM

DU Shaotong,DUAN Liangyu,XIE Wenjuan,YANG Wanli

Electronic Science and Technology. 2020, 33(5):  50-57.  doi:10.16180/j.cnki.issn1007-7820.2020.05.009

Abstract ( 253 )   HTML ( 1 )   PDF (1233KB) ( 48 )  

Figures and Tables | References | Related Articles | Metrics

Low-capacitance single-phase cascaded H-bridge STATCOM with thin film capacitor has the advantage of high reliability. But if the capacitance decreased, the DC side ripple voltage would increase as the compensation capacity increases. Besides, the DC side ripple voltage has a direct effect on the current ripple of STATCOM. In order to clarify the effect of the reduction of capacitance on the ripple current of cascaded H-bridge STATCOM, the current ripple of single-phase low-capacitance cascaded H-bridge STATCOM system was studied. Based on the single-phase low-capacitance cascaded H-bridge STATCOM control system, the approach of time series analysis was utilized to establish the current ripple expression of carrier phase shift pulse width modulation. The influence of capacitance on current ripple was investigated by numerical analysis. The analysis results showed that thecapacitor capacitance reduction failed to increase ripple current. The simulation results verified the correctness of the theoretical analysis, which provided an theoretical reference for the design of low-capacitance single-phase cascaded H-bridge STATCOM.

Application of CHMM and AR Model in Evaluation and Prediction of Bearing Performance Degradation

LIU Yimin,LIU Tao,CHEN Qing

Electronic Science and Technology. 2020, 33(5):  58-65.  doi:10.16180/j.cnki.issn1007-7820.2020.05.010

Abstract ( 321 )   HTML ( 2 )   PDF (1544KB) ( 50 )  

Figures and Tables | References | Related Articles | Metrics

The classic fault diagnosis technology can evaluate the operating status of the system in real time, but in practical applications, it is more desirable to predict the occurrence of the fault to guarantee personal and economic security. As a key component of mechanical equipment, the damage of the bearing may cause serious engineering accidents, so the bearing needs to be diagnosed. In this study, a continuous hidden Markov model was introduced, and the log likelihood was used as an evaluation index to evaluate performance degradation. The logarithm likelihood ratio based on the model output was combined with the autoregressive model, and then was used to predict the performance degradation of the bearing. The validity of the method was verified by comparing the two sets of full-life data. The results showed that the bearing performance degradation evaluation method based on continuous hidden Markov model was effective in evaluating the degradation of bearing performance, and the autoregressive model obtained more accurate results in life prediction.

State Estimation for Permanent Magnet Synchronous Motor Based on Adaptive Interpolation Extended Kalman Filter

ZHU Jun,LI Zihao,LIU Bingchen,MENG Xiangbin,ZHANG Zhe

Electronic Science and Technology. 2020, 33(5):  66-71.  doi:10.16180/j.cnki.issn1007-7820.2020.05.011

Abstract ( 355 )   HTML ( 5 )   PDF (1025KB) ( 54 )  

Figures and Tables | References | Related Articles | Metrics

Speed and rotor position estimation are critical for PMSM non-sensing control system. The EKF algorithm has been widely used in the industrial field as a sensorless control technology. However, the EKF algorithm produces truncation errors during system linearization, and accurate estimates cannot be obtained for highly nonlinear models. In order to reduce the error caused by nonlinear problem of EKF algorithm, based on AIEKF, a state estimation for PMSM was proposed in the study. Based on the degree of nonlinearity of the quantized state equation, this method reduced the influence of the error generated in the linearization of EKF algorithm on the estimation accuracy by adding pseudo state value, thereby reducing the errors caused by the linearization process of the system. Finally, simulation showed that the interception error of AIEKW was lower than EKF by 55.6%.

A Method of Laser Center Extraction Based on Gaussian Fitting

ZHANG Tianhao,LI Haolin

Electronic Science and Technology. 2020, 33(5):  72-76.  doi:10.16180/j.cnki.issn1007-7820.2020.05.012

Abstract ( 424 )   HTML ( 7 )   PDF (775KB) ( 84 )  

Figures and Tables | References | Related Articles | Metrics

In laser-assisted vision measurement system, the selection of fitting data in Gauss fitting method affects the accuracy of laser light center extraction. Aiming at the difficulty to determine the number of Gaussian fitting data in laser images under uneven illumination, an adaptive Gaussian fitting method based on random sampling consistency algorithm was proposed in this paper. According to the characteristic of gray-scale variation in background region of laser image, the random sampling consistency algorithm was used to estimate the gray-scale distribution model of background region. The outliers of the model were extracted as Gaussian fitting data. Then, the Gaussian fitting method was used to obtain the sub-pixel coordinates of the center point of the laser light. The experimental results showed that the Gaussian fitting data in the laser image had an average fitting coefficient of 0.934, which improved the accuracy of laser light center extraction compared with the Gaussian fitting method for fixed fitting data length.

Application of S7-1200 PLC in Position Control of Servo Motor

FAN Zhenrui

Electronic Science and Technology. 2020, 33(5):  77-81.  doi:10.16180/j.cnki.issn1007-7820.2020.05.013

Abstract ( 1334 )   HTML ( 43 )   PDF (964KB) ( 143 )  

Figures and Tables | References | Related Articles | Metrics

With the improvement of production technology, the motion control module of servo motor requires more precise position control. The SIMATIC S7-1200PLC designed and developed by Siemens has good modern process performance. It is given by pulse interface. The motor’s operational control functions are supported to meet current process requirements and are fully functional. This paper studied the application of SIEMENS S7-1200 PLC in the position control of Panasonic servo motor. This paper firstly explained the wiring steps of Panasonic servo motor A5 series hardware and PLC system, and then controled the displacement motion of servo motor through software program. Finally, the experiment was used to verify the correctness of the program. The results showed that the accuracy of the displacement operation could reach 0.1 mm, which could meet the performance requirements of position control.

Design of Moving Target Detection System Based on ZYNQ

WANG Chunjiang,LI Peng

Electronic Science and Technology. 2020, 33(5):  82-86.  doi:10.16180/j.cnki.issn1007-7820.2020.05.014

Abstract ( 479 )   HTML ( 18 )   PDF (1072KB) ( 66 )  

Figures and Tables | References | Related Articles | Metrics

ing at the problems of large size and high power consumption of the current moving target detection system platform, this paper designed a moving target detection system based on the Xilinx ZYNQ platform. ViBe algorithm was a foreground detection algorithm, which mainly included three parts: background modeling, foreground detection, and model update. It had better detection effect. The Vivado HLS development tool was used to complete the development of the algorithm part. The algorithm was optimized by adding optimization instructions. After passing the simulation test, it was packaged into a hardware IP core, which greatly shortened the development process. On the ZYNQ platform, the software and hardware cooperation method was adopted, the PL part was used as the algorithm implementation unit, the PS part was used as the control core, and the video image was collected through the OV5640. The data was stored in the DDR using the VDMA IP core, and the result was processed Output to the display via HDMI. Experimental results showed that the system could detect moving targets in real time, and the system was small in size and low in power consumption.