Top Read Articles

    Published in last 1 year |  In last 2 years |  In last 3 years |  All
    Please wait a minute...
    For Selected: Toggle Thumbnails
    Research Progress of Medical Image Segmentation Method Based on Deep Learning
    LI Zenghui,WANG Wei
    Electronic Science and Technology    2024, 37 (1): 72-80.   DOI: 10.16180/j.cnki.issn1007-7820.2024.01.011
    Abstract787)   HTML47)    PDF(pc) (2075KB)(163)       Save

    Medical image processing technology has developed rapidly with the rise of deep learning. The medical image segmentation technology based on deep learning has become the mainstream method in the segmentation field, which solves the shortcomings of the traditional segmentation method's insufficient segmentation accuracy. This technology has been maturely applied to the segmentation of some pathological images. This study introduces and compares the segmentation methods based on deep learning in recent years, and focuses on the major contributions of U-Net and its improved models in the segmentation field, and summarizes the common medical image modalities and evaluation indicators of segmentation algorithms and commonly used segmentation data sets. Finally, the future development of medical image segmentation technology is prospected.

    Table and Figures | Reference | Related Articles | Metrics
    A Review of Research on Cybersecurity Risk Assessment Methods
    WU Jiacheng, YU Xiao
    Electronic Science and Technology    2024, 37 (3): 10-17.   DOI: 10.16180/j.cnki.issn1007-7820.2024.03.002
    Abstract630)   HTML22)    PDF(pc) (988KB)(119)       Save

    Cybersecurity risk assessment is an important part of building a cyberspace security system, which can effectively protect individuals and organizations from the risk of cybersecurity attacks.This study briefly outlines the theory of cybersecurity risk assessment, emphatically introduces the current mainstream cybersecurity risk assessment methods, and classifies and compares the existing methods according to their different nature, analyzes the advantages, disadvantages and application scope of each method.On this basis, this study summarizes and extracts the factors that have an impact on the cybersecurity assessment results and proposes future research priorities in the field of cybersecurity assessment. The analysis results show that the three factors of correlation, uncertainty of assessment indexes and real-time of assessment process are the main factors affecting the effect of risk assessment, and it provides a reference for the research of risk assessment methods in the future.

    Table and Figures | Reference | Related Articles | Metrics
    Review of Text Classification Research Based on Deep Learning
    WANG Jiawei,YU Xiao
    Electronic Science and Technology    2024, 37 (1): 81-86.   DOI: 10.16180/j.cnki.issn1007-7820.2024.01.012
    Abstract297)   HTML23)    PDF(pc) (769KB)(80)       Save

    Compared with traditional machine learning models, deep learning models attempts to imitate human learning ideas and automatically perform feature extraction from massive data through computers.Text classification is an important application in natural language processing and plays a key role in text information processing.In the past few years, research on text classification has surged and achieved good results.This study briefly introduces text classification methods based on traditional models and deep learning models,and reviews advanced text classification methods, with a focus on models for deep learning.The deep learning methods, research progress and achievements used in text classification in recent years are introduced and summarized, and the development trend of deep learning in the field of text classification and the difficulties are summarized and prospected in this study.

    Table and Figures | Reference | Related Articles | Metrics
    A Survey of Text-to-Image Synthesis Based on Generative Adversarial Network
    LI Yueyang,TONG Guoxiang,ZHAO Yingzhi,LUO Qi
    Electronic Science and Technology    2023, 36 (10): 39-55.   DOI: 10.16180/j.cnki.issn1007-7820.2023.10.006
    Abstract282)   HTML8)    PDF(pc) (5679KB)(62)       Save

    The text-to-image synthesis refers to translating the text description in sentence form into an image with similar semantics to the text. In the early research, the task of image generation is mainly based on keyword or sentence retrieval to align the visual content matched with the text. With the generative adversarial network, the method of text-to-image synthesis has made great progress in visual realism, diversity and semantic similarity. The generative adversarial network generates reasonable and real images through the confrontation between generator and discriminator, and shows strong ability in the fields of image restoration and super-resolution generation. Based on the review and summary of the latest research results in the field of text-to-image synthesis, a new classification method is proposed: Attention enhancement, multi-stage enhancement, scene layout enhancement and universality enhancement. The challenges and future development direction of text-to-image synthesis are also discussed in this study.

    Table and Figures | Reference | Related Articles | Metrics
    Chinese License Plate Detection and Recognition in Unconstrained Scenarios Based on YOLO
    CHEN Ziang,LIU Na,YUAN Ye,LI Qingdu,WAN Lihong
    Electronic Science and Technology    2023, 36 (10): 1-8.   DOI: 10.16180/j.cnki.issn1007-7820.2023.10.001
    Abstract272)   HTML24)    PDF(pc) (2014KB)(126)       Save

    In view of the problems of traditional Chinese license plate recognition methods, such as the requirement of scenes, poor real-time performance, and inability to deploy on edge devices, this study proposes a Chinese license plate detection and recognition method based on YOLO(You Only Look Once) in unconstrained scenes. This method is divided into two modules: license plate detection and license plate character recognition. In the license plate detection part, the improved YOLOv5 model is used to predict four groups of key points for license plate correction based on the prediction of target candidate regions, and the pre-training model trained on the COCO data set is used for training, which reduces the error detection problem caused by the complex environment and has high real-time performance. In the license plate character recognition part, the CRNN(Convolutional Recurrent Neural Network) model is improved, which greatly reduces the parameters and computation of the algorithm, so that it can be successfully deployed in various edge devices. Experimental results show that the proposed method can efficiently detect and recognize license plates in complex environments. The map value of the proposed license plate detection model is 3.0% higher than that of Retina-face in the license plate detection data set. Compared with LPR-Net, the accuracy of license plate character recognition model in license plate recognition data set is improved by 4.2%.

    Table and Figures | Reference | Related Articles | Metrics
    TDOA Sound Source Localization Method Based on Particle Swarm Optimization Algorithm
    ZHANG Dagui,ZHOU Zhifeng,ZHANG Yi,WANG Liduan
    Electronic Science and Technology    2023, 36 (9): 21-28.   DOI: 10.16180/j.cnki.issn1007-7820.2023.09.004
    Abstract256)   HTML14)    PDF(pc) (1009KB)(60)       Save

    In order to solve the problem of 3D coordinate estimation of sound source based on planar microphone array, this study introduces particle swarm optimization algorithm in TDOA(Time Difference of Arrival) sound source localization algorithm. The true value of the delay difference is calculated using the generalized cross-correlation method of the PHAT(Phase Transform) weighting function. Combined with the coordinate position of the microphone, the estimated value of the delay difference between the hypothetical sound source arriving at the microphone is calculated through the geometric relationship. The sum of the squares of the error between the actual value and the estimated value of the design delay is the particle fitness function. The particle swarm optimization algorithm is used to search for the sound source points in the space that conform to the fitness function, so as to realize the sound source position estimation. The simulation results show that the proposed algorithm has better robustness and noise resistance than the spherical interpolation method when the calculation speed is similar to that of the spherical interpolation method.

    Table and Figures | Reference | Related Articles | Metrics
    A Review and Prospect of Research on Situational Awareness Technology in Active Distribution Network
    MAN Yanlu,LIU Min,WANG Kai
    Electronic Science and Technology    2024, 37 (2): 6-13.   DOI: 10.16180/j.cnki.issn1007-7820.2024.02.002
    Abstract246)   HTML14)    PDF(pc) (1036KB)(52)       Save

    With the large-scale access of distributed generation and diversified loads, the traditional distribution network is gradually transformed into an active distribution network,which means that the types of faults in distribution networks are becoming more diverse, and the operating environment, operating conditions and topologies are becoming more and more complex. Therefore, in order to accurately predict the potential risks of the system in the active distribution network, it is necessary to improve the timeliness and accuracy of system operation decisions through sophisticated and efficient situational awareness technology.This study expounds the significance and concept of situational awareness technology in active distribution network, constructs its basic framework, latsly summarizes the research process, research difficulties and future research directions of situational awareness, situation understanding and situation prediction in detail.

    Table and Figures | Reference | Related Articles | Metrics
    Research on Blind Roads and Obstacle Recognition Algorithm Based on Deep Learning
    MA Wenjie, ZHANG Xuanxiong
    Electronic Science and Technology    2024, 37 (3): 75-83.   DOI: 10.16180/j.cnki.issn1007-7820.2024.03.010
    Abstract242)   HTML10)    PDF(pc) (3884KB)(119)       Save

    Blind roads and blind road obstacles are important factors that affect the travel safety of blind people. Existing algorithms only deal with blind road segmentation and blind road obstacle detection separately, with low efficiency and high computational complexity. To solve the above problems, this study proposes a multi-task recognition algorithm based on deep learning. The algorithm extracts public features through the backbone network, after the extracted features are fused through the SPP(Spatial Pyramid Pooling)and FPN(Feature Pyramid Networks)networks, they are respectively passed into the segmentation network and the detection network to complete the tasks of blind road segmentation and blind road obstacle detection. In order to make the blind road segmentation smoother, a correction loss function is introduced. In order to improve the recall rate of obstacle detection, the NMS(Non Maximum Suppression) of the detection network is replaced by Soft-NMS. The experimental results show that the algorithm segmentation part MIoU, MPA reach 93.52%, 95.29%, respectively, and the detection part mAP(mean Average Precision)、mAP@0.5 and mAP@0.75 respectively reach 75.58%、91.58%and 74.82%. Compared with using the SegFormer network for blind road segmentation and the RetinaNet network for blind road obstacle detection, this algorithm not only improves the accuracy, but also improves the speed by 73.72%, and the FPS(Frames Per Secon) reaches 18.52. Compared with other comparative algorithms, this algorithm also has a certain improvement in speed and accuracy.

    Table and Figures | Reference | Related Articles | Metrics
    Short-Term Power Load Forecasting Based on FA-SVR-LSTM Combined Model
    WEN Yanfei,WANG Wanxiong
    Electronic Science and Technology    2023, 36 (9): 1-7.   DOI: 10.16180/j.cnki.issn1007-7820.2023.09.001
    Abstract238)   HTML55)    PDF(pc) (957KB)(59)       Save

    As the basis for maintaining the operation and analysis of the power grid system, short-term power load forecasting provides judgment basis and information for the economic dispatch and safety analysis of the power grid system, and plays an important role in maintaining the normal operation of the power grid system. In this study, the FA(Firefly Algorithm) is used to optimize the penalty factor c, nuclear parameter g of SVR(Support Vector Regression) model and the number of neurons m and learning rate lr of LSTM(Long Short-Term Memory) model. The FA-SVR-LSTM combined prediction model is established using the optimal parameters, and the sample data are predicted. Taking the historical data of power load of Florida as an example, four reference models of LSTM, SVR, FA-SVR and FA-LSTM are established to predict the power load of 360 h in 15 days, and the results are compared with those of FA-SVR-LSTM. The experimental results show that compared with LSTM and SVR model, the prediction accuracy of FA-SVR-LSTM model is improved by 33.184 9% and 30.326 5%, respectively. The evaluation values of MAPE and RMSE are significantly lower than those of the other four models. These results indicate that the prediction effect of FA-SVR-LSTM combined model is significantly improved when compared with other models.

    Table and Figures | Reference | Related Articles | Metrics
    Design of Low Sidelobe W-Band Sparse Array Consists of Horn-Like Antenna
    ZHOU Biao,ZHANG Shuai,ZHANG Dexun,LIN Zhicheng,WANG Jian
    Electronic Science and Technology    2023, 36 (9): 8-14.   DOI: 10.16180/j.cnki.issn1007-7820.2023.09.002
    Abstract234)   HTML12)    PDF(pc) (1500KB)(44)       Save

    In order to solve the design problem of wideband wide-angle low sidelobed phased array antenna caused by factors such as large loss, manufacturing process, T/R component size, number of channels and location, a wideband type horn antenna unit in the frequency range of 80~110 GHz is designed in this study, and the antenna standing wave ratio is less than 1.8. Considering the space limitation of T/R components, a 64-element high-gain, low-sidelobe, wide-band and wide-angle scanning array antenna is designed by optimizing the array position, and a small low-profile air waveguide feeding network is designed for the array. The simulation results show that the designed 64 element sparse array has the advantages of wide bandwidth, wide angle, low sidelobes, high gain, easy fabrication and realization, and the gain is higher than 22.6 dB, the sidelobes level is lower than -6.2 dB, and the active VSWR of each element is less than 2 under the scanning range of ±40°.

    Table and Figures | Reference | Related Articles | Metrics
    Research and Application of Polar Code Encoding and Decoding in 5G Communication
    GAO Jie
    Electronic Science and Technology    2024, 37 (1): 61-65.   DOI: 10.16180/j.cnki.issn1007-7820.2024.01.009
    Abstract220)   HTML12)    PDF(pc) (580KB)(49)       Save

    In order to optimize the implementation complexity and information transmission speed of the fifth generation communication technology, the Polar code based on channel polarization phenomenon gradually receives great attention and research, and forms many industry standards and research conclusions. In recent years, the development and optimization of Polar code encoding and decoding technology promotes the popularity of the fifth generation communication technology, and the rapid application of the new generation communication technology greatly promotes the further iteration of the underlying encoding and decoding technology. By reviewing the research history of Polar code encoding and decoding technology, the research status and context of Polar code encoding and decoding technology in the fifth generation communication are summarized. On this basis, the problems and requirements of current coding and decoding theory are analyzed and discussed in depth, and the research ideas and trends of the underlying coding and decoding theory in the future communication technology are proposed.

    Reference | Related Articles | Metrics
    Design and FPGA Implementation of Dehazing Based on Channel Difference Model and Guided Filtering
    CAO Hongfang,WANG Xiaolei,DU Gaoming,LI Zhenmin,NI Wei
    Electronic Science and Technology    2023, 36 (8): 1-6.   DOI: 10.16180/j.cnki.issn1007-7820.2023.08.001
    Abstract204)   HTML1410)    PDF(pc) (2063KB)(190)       Save

    Computer vision systems are affected by foggy weather, resulting in poor quality images captured. To solve this problem, this study proposes a guided filtering dehazing algorithm based on channel difference model and its FPGA design. The channel difference model is obtained by separating the bright channel and dark channel of foggy image, and the model is used as a guide map for guided filtering to smooth the foggy image. Finally, a high boost filtering operation is performed to obtain a dehazed image. The hardware architecture is designed and implemented on FPGA. The experimental results show that the image scene after dehazing has uniform illumination, high degree of texture information recovery and high color fidelity. For an image of 480×270 size, the integrated frequency of the circuit is 108.448 MHz, the throughput is 323.47 MB·s-1, and the time to complete the entire dehazing is 0.001 2 s. These results indicate that the proposed algorithm and its hardware design can effectively improve image visibility and dehazing speed.

    Table and Figures | Reference | Related Articles | Metrics
    Overview of Gallium Nitride Solid State Power Amplifier Development
    XIE Hongxing,LU Hongmin,LIU Liang,REN Yongda,LI Min,ZHANG Jiahai
    Electronic Science and Technology    2023, 36 (8): 65-71.   DOI: 10.16180/j.cnki.issn1007-7820.2023.08.010
    Abstract199)   HTML8)    PDF(pc) (2327KB)(52)       Save

    Solid-state power amplifiers using semiconductor technology have the advantages of small size and high stability, and have replaced traditional travelling wave tube amplifiers in many microwave applications. Among all types of semiconductor materials, the third-generation semiconductor material GaN(Gallium Nitride) has been widely used in power amplifiers because of its advantages of wide band gap, high electron mobility, and high breakdown voltage. Based on the development of power amplifiers, this study describes the development history of solid-state power amplifiers, summarizes the performance comparison between GaN technology and other semiconductor technologies, and focuses on power amplifiers using GaN HEMT(GaN High Electron Mobility Transistor) technology. This study discusses various types of GaN HEMT power amplifiers, including class A, B, C, D, and E, etc., introduces the efficiency and linearity improvement techniques applied to GaN power amplifiers, including Doherty power amplifiers and envelopes tracking technology, and digital pre-distortion technology, etc., and makes a summary and comparison of related technologies.

    Table and Figures | Reference | Related Articles | Metrics
    Power Load Forecasting Model Based on Expansion Period
    ZHANG Haifang,HE Qinglong,ZHANG Lin
    Electronic Science and Technology    2024, 37 (2): 1-5.   DOI: 10.16180/j.cnki.issn1007-7820.2024.02.001
    Abstract175)   HTML31)    PDF(pc) (756KB)(105)       Save

    In view of the problem that the existing power load forecasting models rely on recent data, which leads to the prediction results deviating from the real situation of the time series, a power load forecasting model based on extended period information is proposed. The pre-processed power load time series is processed according to the same time and different days. On this basis, the ARIMA(Autoregressive Integrated Maving Average Model) model and LSTM(Long Short-Term Memory Network) model are used for modeling and analysis, and three evaluation indicators are used to evaluate the predictive performance of the model. The prediction results show that the three evaluation indexes of the ARIMA model constructed by expanding the period information are lower than those of the traditional ARIMA model, and the corresponding RMSE(Root Mean Square Error), MAE(Mean Absolute Error) and MAPE(Mean Absolute Percentage Error) are 32 434.114 8, 5 828.390 9 and 0.025 2, respectively. The LSTM model of expanding the period information is also lower than the original LSTM model, and the corresponding RMSE, MAE, and MAPE are 13 520.497 4, 9 298.352 6, and 0.091 4,respectively.

    Table and Figures | Reference | Related Articles | Metrics
    Research Progress on Control Technology of Multi-Degree of Freedom Parallel Robot
    LUO Xiaoqing
    Electronic Science and Technology    2023, 36 (11): 89-94.   DOI: 10.16180/j.cnki.issn1007-7820.2023.11.013
    Abstract173)   HTML9)    PDF(pc) (665KB)(56)       Save

    In order to further improve the control accuracy and stability of the multi degree of freedom parallel robot, the control optimization methods based on technical models such as dynamics and redundant drive branches have gradually attracted extensive attention and research in the academic community, and a certain number of research results and conclusions have been proposed. In recent years, with the mathematical expression of dynamics and driving branch model becoming clear and refined, the control accuracy and stability of parallel robots with multi-degree of freedom have been further improved. By reviewing the research history of parallel robot control technology, the current research ideas and status quo of parallel robot control technology are summarized. On this basis, the existing problems of parallel robot control technology are deeply explored, and the research trend and development direction of future parallel robot control technology are proposed.

    Reference | Related Articles | Metrics
    Design of SoC Secure Memory Based on Lightweight Block Cipher
    LIU Wei,ZENG Xiangyi,XIAO Hao
    Electronic Science and Technology    2023, 36 (9): 15-20.   DOI: 10.16180/j.cnki.issn1007-7820.2023.09.003
    Abstract169)   HTML13)    PDF(pc) (976KB)(49)       Save

    In view of the security risks faced by RAM(Random Access Memory) and Flash memory of embedded on-chip systems, this study outlines physical attacks against memory of traditional SoC(System on Chip) chips, and presents a memory controller that supports encryption algorithms. Using the lightweight block cipher algorithm LBlock-s, the cryptographic security analysis shows that the algorithm has a good resistance to differential cryptanalysis. Compared with traditional block cipher algorithms such as AES(Advanced Encryption Standard), the proposed method reduces hardware resource consumption while guaranteeing security and is suitable for all kinds of resource-constrained secure SoC chips. To improve the data throughput, the hardware structure of the algorithm is expanded so that the standard 32 rounds of encryption or decryption take 1 clock cycle. This scheme guarantees that sensitive data cannot be parsed even if acquired by the attacker without consuming more hardware resources and encrypting delay, and effectively avoids physical attack on the security chip.

    Table and Figures | Reference | Related Articles | Metrics
    Network Security Device Design Based on Red-Black Isolation Architecture
    GONG Zhi,LIU Chao,FU Qiang
    Electronic Science and Technology    2024, 37 (2): 76-86.   DOI: 10.16180/j.cnki.issn1007-7820.2024.02.011
    Abstract165)   HTML9)    PDF(pc) (3711KB)(66)       Save

    The data transmission of the heaven and the earth integrated network based on IP(Internet Protocol) technology is vulnerable to illegal attacks. The traditional network security device based on IPSec(Internet Protocol Security) is designed by connecting a single host to both internal and external network processing units, which has the risk of unauthorized users directly accessing the protected intranet through the extranet. A new scheme for a network security device based on a red-black isolation architecture is proposed. The scheme adopts the design concept of red-black partition and VPN technology based on IPSec framework under Linux. It implements the validity verification of the transmitted data based on the "quintuple" security and security rules and the encapsulation and decapsulation transformation of the IPSec ESP protocol in the red zone, and implements the public network sending and receiving of the ESP encapsulated and encrypted data in the black zone. In this scheme, the security service module implements dynamic switching of encryption algorithms and encryption and decryption of ESP encapsulated data according to external instructions, and uses the security service module as a data exchange channel between the red and black zones to achieve isolation between the internal and external networks and effectively ensure Intranet security.The experimental results show that the network security device based on red-black isolation architecture has the advantages of strong anti-attack capability, replaceable encryption algorithm, and encryption rate of 1 024 bytes packet length greater than 50 Mbit·s-1 under 100 megabit bandwidth.

    Table and Figures | Reference | Related Articles | Metrics
    Research on Hybrid Critical-Level Task Scheduling and Semi-Partition Algorithm in Multi-Core Processor
    ZHU Jiawei, MAO Hang, ZHANG Fengdeng
    Electronic Science and Technology    2024, 37 (3): 1-9.   DOI: 10.16180/j.cnki.issn1007-7820.2024.03.001
    Abstract162)   HTML26)    PDF(pc) (1186KB)(136)       Save

    At present, the schedulability analysis of mixed critical level tasks and semi-partition scheduling algorithms in most multiprocessors are focused on single-core utilization. However, due to the high complexity of task scheduling in multi-core systems, the existing research results have some problems, such as unbalanced load of each processor and unsatisfactory task schedulability. To solve this problem, the application scope of Dynamic Demand Boundary Function(DDBF) is extended to multi-core processor system in this study. DDBF is improved based on half-partition scheduling algorithm, and SDDBF(Super Dynamic Demand Boundary Function) is proposed by adding forward job and forward job analysis, which can calculate and utilize resources more accurately. Based on SDDBF, the schedulability analysis method of SDA(Stepper Dispatch Algorithm) and semi-partition algorithm MCWF(Mixed-Criticality Worist First) are proposed. The simulation results show that compared with AMC(Adaptive Mixed Criticality), AMC-MAX and XU algorithms, the schedulability analysis of SDA can be improved by 5%~10%. Compared with WF_MY(Worst First_My) and WF_NEW(Worst First_New) algorithms, MCWF makes the system have better CPU(Central Processing Unit) load balancing performance at any critical level.

    Table and Figures | Reference | Related Articles | Metrics
    Optimal Control Strategy for Bi-Directional Active Full Bridge DC-DC Converter Based on Dual Phase Shift
    LI Yudong,LUO Xinquan,LI Peifeng,HUANG Xin
    Electronic Science and Technology    2023, 36 (12): 87-94.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.012
    Abstract162)   HTML5)    PDF(pc) (3967KB)(58)       Save

    In view of the problem of high backflow power, the hard opening loss of the switching device and poor dynamic performance of bidirectional active full-bridge converter,a control scheme based on dual phase shifting combining backflow power optimization and virtual direct power control is proposed in this study. By analyzing the working principle and working characteristics of the dual phase shift control, the soft switching boundary of zero voltage switching is derived as the constraint condition,with the soft switching as the constraint condition, The blackflow power optimization scheme in the whole power range is obtained by KKT(Karush-Kuhn-Tucker) conditional optimization algorithm.This scheme is combined with virtual direct power method and works simultaneously.When the operation condition changes,it could not only reduced effectively backflow power,but also enhance fast dynamic response.Finally,the MATLAB/Simulink simulation model and a small power converter prototype is built for the comparison experiment to verify the correctness and superiority of the proposed control strategy.

    Table and Figures | Reference | Related Articles | Metrics
    Cooperative Localization of IMMKF and Chan-Taylor Algorithm
    WANG Xinyue,YU Huimin,HU Luning
    Electronic Science and Technology    2023, 36 (12): 1-8.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.001
    Abstract157)   HTML15)    PDF(pc) (1037KB)(66)       Save

    TDOA(Time Difference of Arrival) rangmg method is a typical opproach for UWB(Ultra Wide Band) indoor location.A Chan-Taylor-IMMKF(Interacting Multiple Model Kalman Filter) localization technique is suggested in this study to address the unavoidable random error and inaccurate location of targets with changing motion states. With the addition of the adaptive algorithm IMM, the algorithm is made up of the Chan-Taylor weighting algorithm and the Kalman filter algorithm. The Chan-Taylor weighting procedure is used to acquire the target estimated coordinates for the first time. The coordinate value is then used as the measurement value for the Kalman filter of the adaptive algorithm IMM, and the target coordinates are filtered many times. The target's final estimated coordinates are provided by the final weighting. The experimental results reveal that the filtered Chan-Taylor weighting algorithm outperforms both conventional Kalman filtering and the unfiltered Chan-Taylor weighting algorithm. The algorithm successfully lowers the system's random error and fixes the issue that the conventional Kalman filter cannot track the significant error when the target abruptly changes its motion state,and the mean error standard deviation is controlled within 15 cm.

    Table and Figures | Reference | Related Articles | Metrics
    Research on Electromagnetic Transient Acceleration Simulation Technoloy of New-Type Power System
    NIE Chunfang, HAO Zhenghang, CHEN Zhuo, HE Puxiang
    Electronic Science and Technology    2024, 37 (3): 18-25.   DOI: 10.16180/j.cnki.issn1007-7820.2024.03.003
    Abstract155)   HTML6)    PDF(pc) (1938KB)(58)       Save

    In order to solve the problems of low simulation efficiency and difficulty in simulation in the electromagnetic transient simulation of new power system due to complex system topology, numerous power electronic switching devices and insufficient single-core computing capability of simulator, this study uses the ideal transformer model segmentation algorithm to divide the large-scale new power system model into several subsystems. The decoupling and order reduction of the large system are realized and the computation of the whole system as one state-space system matrix is effectively reduced during simulation. In order to further reduce the computing burden of a single processor, an accelerated simulation platform UREP300 for efficient parallel computing in the bare-metal environment is designed by using multi-core CPU(Central Processing Unit) parallel technology. The segmented model is loaded into UREP300 for accelerated simulation experiment and compared with the offline simulation of the original model based on MATLAB/Simulink. The experimental results show that the acceleration simulation technology combining the ideal transformer model segmentation and multi-core parallel operation can not only guarantee the simulation accuracy but also improve the simulation speed to 586 times of the original, which can significantly improve the simulation efficiency, and is suitable for large-scale new power system simulation work.

    Table and Figures | Reference | Related Articles | Metrics
    Multi-Encoder Transformer for End-to-End Speech Recognition
    PANG Jiangfei, SUN Zhanquan
    Electronic Science and Technology    2024, 37 (4): 1-7.   DOI: 10.16180/j.cnki.issn1007-7820.2024.04.001
    Abstract148)   HTML14)    PDF(pc) (949KB)(88)       Save

    The current widely used Transformer model has a strong ability to capture global dependencies, but it tends to ignore local feature information at shallow layers. To solve this problem, this study proposes a method using multiple encoders to improve the ability of speech feature extraction. An additional convolutional encoder branch is added to strengthen the capture of local feature information, make up for the neglect of local feature information in shallow Transformer, and effectively realize the integration of global and local dependencies of audio feature sequences. In other words, a multi-encoder model based on Transformer is proposed. Experiments on the open-source Chinese Mandarin data set Aishell-1 show that without an external language model, the proposed Transformer-based multi-encoder model has a relative reduction of 4.00% in character error rate when compared with the Transformer model. On the internal non-public Shanghainese dialect data set, the performance improvement of the proposed model is more obvious, and the character error rate is reduced by 48.24% from 19.92% to 10.31%.

    Table and Figures | Reference | Related Articles | Metrics
    A Firmware Security Update Scheme for Embedded Devices
    ZENG Xiangyi,LIU Wei,XIAO Hao
    Electronic Science and Technology    2023, 36 (8): 14-18.   DOI: 10.16180/j.cnki.issn1007-7820.2023.08.003
    Abstract148)   HTML13)    PDF(pc) (1427KB)(41)       Save

    In view of the security problem of embedded device firmware update, this study proposes a multi-check firmware security update scheme based on hash, symmetric and asymmetric encryption algorithms. In this study, the master key pairs, temporary key pairs, shared key and hash chain are designed to protect firmware update from identity authentication, data encryption, integrity check and other aspects, which can effectively prevent illegal users, firmware tampering, firmware data leakage, replay attack, firmware rollback and other attacks. In this study, the concrete implementation of the security update scheme is carried out. The experiment results show that compared with the ISP(In System Programming) and IAP(In Application Pragramming) technologies without any security protection, the scheme achieves the security protection of the whole process of firmware update at the time cost of about 7% and 11%, it provides a safe and reliable update method for embedded device firmware update.

    Table and Figures | Reference | Related Articles | Metrics
    Research on Optimal Charging Strategy of Electric Vehicle Based on Multi-Objective Particle Swarm Optimization
    LI Tingting, LOU Ke, WANG Yuan, XU Huachao
    Electronic Science and Technology    2024, 37 (3): 51-56.   DOI: 10.16180/j.cnki.issn1007-7820.2024.03.007
    Abstract145)   HTML4)    PDF(pc) (1090KB)(70)       Save

    Household electric vehicle charging in residential areas has a strong centrality. Large-scale electric vehicle charging load causes large peak-valley load difference and other problems in the distribution network system. This study proposes a user charging selection control strategy based on Multi-Objective Particle Swarm Optimization(MPSO) algorithm. Through the analysis and prediction of electric vehicle charging load, a multi-objective optimization model is established with the minimum variance of the total system load and scheduling cost as the objective function. Meanwhile, considering the constraints of electric vehicle battery and system power, the MPSO algorithm is used to solve the optimal initial charging time of electric vehicles. The simulation results show that compared with unordered charging of EVs in residential areas, the EV charging strategy proposed in this study can effectively reduce load peak and dispatch cost.

    Table and Figures | Reference | Related Articles | Metrics
    Design of Gigabit Ethernet Port Communication Based on FPGA
    LAN Wei,HAN Yanzhe,HU Xiao
    Electronic Science and Technology    2024, 37 (1): 48-54.   DOI: 10.16180/j.cnki.issn1007-7820.2024.01.007
    Abstract142)   HTML7)    PDF(pc) (1331KB)(50)       Save

    In view of the problem of Ethernet transmission rate and real-time in embedded field, a design of Gigabit Ethernet port communication based on FPGA(Field Programmable Gate Array)is proposed. This study designs the forwarding function of Gigabit Ethernet switch, and implements end-to-end data communication based on label forwarding. Datagrams with label are sent by CPU(Central Processing Unit), and are outputted through Gigabit Ethernet interface, and are sent to FPGA through RGMII(Reduced Gigabit Media Independent Interface)port. FPGA judges output port number field in label through internal logic and removes label, and outputs datagrams to connecting device from corresponding Gigabit Ethernet interface. Peripherals input datagrams through Gigabit Ethernet ports, and send datagrams to FPGA through SGMII(Serial Gigabit Media Independent Interface) protocol. FPGA adds labels through internal logic and outputs them to CPU through polling, so as to realize interworking of connecting devices of multiple Gigabit Ethernet interface. The experimental results reveal the feasibility and effectiveness of the FPGA logic. The transmission rate reaches 1 Gbit·s-1, the datagrams forwarding delay is less than 100 μs, and the packet loss rate is 0%,which indicates that the data transmission stability is high, and the proposed design meets the actual needs of existing projects.

    Table and Figures | Reference | Related Articles | Metrics
    PID Parameter Tuning Based on Improved Honey Badger Optimization Algorithm
    HU Tao,JIANG Quan
    Electronic Science and Technology    2023, 36 (12): 46-54.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.007
    Abstract133)   HTML9)    PDF(pc) (1487KB)(42)       Save

    As a swarm intelligence algorithm simulating the predator-prey behavior of honey badger, honey badger algorithm has many problems, such as easy to fall into local optimal solutions, and the number of iterations required. In view of the shortcomings of honey badger algorithm, a cloud honey badger algorithm (CHBA) combining gravity search algorithm and normal cloud technology is proposed. The density factor of the original honey badger algorithm that controls the individual search range of the honey badger is replaced by the acceleration in the gravitational search algorithm to improve the rationality of the individual search range of the honey badger and accelerate the search iteration speed. The normal cloud algorithm is used to generate a new batch of honey badgers with the expectation of the best position of the honey badger between generations, so as to improve the population diversity and avoid falling into local optimization. At the same time, the generation range of the new honey badger is adaptively adjusted to avoid local optimization. Twenty three benchmark functions are selected to test the proposed algorithm. From the optimization results of single peak, multi peak and fixed dimension multi peak functions, the step response PID(Proportion Integration Differentiation) parameters of first-order delay system, non minimum phase system and first-order minimum delay system are optimized and compared, and the results show that CHBA algorithm has better performance in search efficiency and iteration accuracy.

    Table and Figures | Reference | Related Articles | Metrics
    Research on MHz High Voltage Pulse Power Supply
    DING Kai,RAO Junfeng
    Electronic Science and Technology    2024, 37 (2): 69-75.   DOI: 10.16180/j.cnki.issn1007-7820.2024.02.010
    Abstract131)   HTML6)    PDF(pc) (1597KB)(47)       Save

    In view of the application requirements of high frequency nanosecond pulse such as uniform discharge and biomedical, a high frequency and high voltage pulse power supply based on radio frequency Metal Oxide-Semiconductor Field Effect Transistor(MOSFET) is designed. The control signal of the pulse power supply is provided by Field Programmable Gate Array(FPGA) and transmitted through optical fiber. After amplification by the driver chip, each stage of the discharge tube is triggered synchronously. The drive chip adopts power module to provide different ground potential. The discharge tube adopts Residual Current Device(RCD) absorbing circuit to improve the instantaneous working condition at the opening and closing time. The main analysis and research is the effect of the last stage isolation inductance on the voltage waveform of discharge tube and load voltage waveform, and the Marx circuit is improved. The experimental results show that the power supply can output nanosecond pulses with rising edge of 40 ns, half-height width of 100 ns and voltage amplitude of 1.1 kV at 1 MHz repetition rate on 1 kΩ resistance load. It is verified that the power supply can work reliably in high frequency and high voltage state and meet the design requirements.

    Table and Figures | Reference | Related Articles | Metrics
    Application of Intelligent Inspection Robot Technology for Hydropower Station
    SHEN Hao,ZHAO Yifeng,LI Xiao
    Electronic Science and Technology    2023, 36 (12): 99-102.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.014
    Abstract130)   HTML8)    PDF(pc) (1052KB)(64)       Save

    In view of the high cost of crack and seepage detection in the traditional artificial pumped storage station and the difficulty of ensuring the detection accuracy, a set of inspection robot system based on machine vision is designed in this study. A convolutional neural network which combines cross entropy and dice cost function is constructed, and an evaluation function based on total pixel accuracy, cross parallel ratio and F1-score is established to ensure the accurate detection of common cracks. In order to verify the effectiveness of the designed robot inspection system, convolutional neural network is tested in this study, and its performance is compared with common computer vision methods and manual detection methods. The comparison results show that the neural network constructed in this study has obvious progress in detection accuracy and detection efficiency.

    Table and Figures | Reference | Related Articles | Metrics
    A Bayesian Network Structure Learning Algorithm with Structure Priors
    TONG Zhaojing,LI Jinxiang,QIAO Zhengrui
    Electronic Science and Technology    2023, 36 (11): 1-7.   DOI: 10.16180/j.cnki.issn1007-7820.2023.11.001
    Abstract130)   HTML9)    PDF(pc) (947KB)(44)       Save

    The computational complexity of Bayesian Networks(BN) increases with the increase of the number of nodes, and the optimal structure of BN is still a NP(Non-deterministic Polynomial Time)-hard problem. To optimize the BN structure and improve the computing power of the complex BN structure, the BN structure is optimized through the hybrid learning method of constraints and scores. In the constrained learning, PC (Peter-Clark) algorithm is used to generate the initial network structure to improve the initial score of the network. Score-based learning uses the sparrow search algorithm to find the optimal structure of BN to enhance its scoring search ability in BN. The sparrow search algorithm and PC algorithm are applied to BN to optimize its structure, and the standard BN is used to conduct experiments, which proves the feasibility and effectiveness of the proposed algorithm in BN structure learning. Experiments on networks with different complexities show that the proposed method obtains better BIC scores than other algorithms, and in the test of 2 000 samples on the ASIA network, the error from the standard score is only 0.2.

    Table and Figures | Reference | Related Articles | Metrics
    Scene Recognition Algorithm Based on Deep Transfer Learning and Multi-Scale Feature Fusion
    WANG Qiao,HU Chunyan,LI Feifei
    Electronic Science and Technology    2023, 36 (11): 19-27.   DOI: 10.16180/j.cnki.issn1007-7820.2023.11.004
    Abstract123)   HTML4)    PDF(pc) (2190KB)(52)       Save

    CNN(Convolutional Neural Networks) hase achieved excellent results in the field of scene recognition research, but this method do not fully take into account the particularity of the scene. Due to different scales, viewpoints, and backgrounds, there exists large intra-class variation within the same scene class. On the other hand, the common objects also result in a certain inter-class similarities among heterogeneous scenes as well. Considering that scene images of different scales will affect the size of objects in them, this study proposes a scene recognition algorithm based on deep transfer learning and multi-scale feature fusion. First, the network parameters pre-trained on the Places data set are migrated to the CNN model used in this study using migration learning, and then the network is fine-tuned and retrained to reduce the training cost. Secondly, the multi-scale image blocks obtained from the class activation map are fed into the CNN for feature extraction, and the obtained feature vectors are fused to make the final scene image features more comprehensive. Experiment results carried out on the SUN397 data set show that compared with other CNN-based algorithms, the proposed algorithm significantly improves the accuracy of scene recognition.

    Table and Figures | Reference | Related Articles | Metrics
    Research on Generating News Text Summarization Based on Improved T5 PEGASUS Model
    ZHANG Qi,FAN Yongsheng
    Electronic Science and Technology    2023, 36 (12): 72-78.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.010
    Abstract120)   HTML4)    PDF(pc) (906KB)(43)       Save

    The task of generating news text summarizations aims to solve the problems of wasting time and reading fatigue caused by users' inability to quickly grasp the key points of the content when reading news. At present, the best text summarization model for Chinese is the T5 PEGASUS model, but there are few researches on this model. In this study, the Chinese word segmentation of the T5 PEGASUS model is improved, and the Pkuseg word segmentation method, which is more suitable for news field, is used for processing, and its effectiveness is verified on three public datasets with different news lengths: NLPCC2017, LCSTS and SogouCS. It is found that the Pkuseg method is more suitable for the T5 PEGASUS model. The ROUGE value of T5 Pegasus model generated summaries is positively correlated with the length of news text, and the loss value of training set and the decline speed of loss value are negatively correlated with the length of news text. In the face of a small number of training sets, the model can get a high ROUGE score, so the model has a strong few-shot learning ability.

    Table and Figures | Reference | Related Articles | Metrics
    Small Object Detection Based on Convolution and Self-Attention of Aggregation
    WANG Xiaozhu,YU Lianzhi
    Electronic Science and Technology    2024, 37 (2): 14-22.   DOI: 10.16180/j.cnki.issn1007-7820.2024.02.003
    Abstract117)   HTML19)    PDF(pc) (2213KB)(78)       Save

    Small object detection is a research hotspot in most object detection open datasets. In view of the problem of insufficient detection accuracy of small targets in multi-size detection scenarios, an improved small target detection model based on YOLOv5s(You Only Look Once version 5s) is proposed in this study.A convolution self-attention aggregation residual block is added to the feature extraction network of the detector to improve the feature extraction ability, and a new feature graph is introduced from the shallow network to enhance the feature information of small object. The feature fusion network structure is improved to make full use of the newly introduced shallow features. SIOU Loss is introduced to replace the original GIOU Loss rectangular frame loss function to improve the detection accuracy and training speed.The experimental results show that the detection accuracy of the improved model is 0.012 higher than YOLOv5s on the 2007 and 2012 data sets of PASCAL VOC, and the small object detection accuracy is 0.023 higher than YOLOv5s. The detection accuracy of the imporved model in MS COCO data set is 0.001 higher than YOLOv5s, and the detection accuracy of small objects is 0.009 higher than YOLOv5s.

    Table and Figures | Reference | Related Articles | Metrics
    A Miniaturized Broadband Tri-Polarized Antenna for 5G Mobile Communication
    HAN Guodong,FU Ming,CHEN Xi
    Electronic Science and Technology    2023, 36 (12): 32-38.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.005
    Abstract116)   HTML4)    PDF(pc) (5250KB)(36)       Save

    A miniaturized broadband tri-polarized antenna for 5G mobile communication is proposed in this study. A pair of hollow annular orthogonal dipoles is adopted as the horizontal dual polarization of the antenna, meanwhile a planar rectangular frame equivalent to a monopole is constructed as the vertical polarization of the antenna, which is fed by a microstrip integrated balun and the four-equivalent power dividing microstrip lines, respectively. The proper choice of the feeding method and the ingenious design of the pointed cone planar rectangular frame achieve the broadband characteristics of the antenna, and the integrated structure layout realizes the miniaturization of the antenna. The antenna structure is compact with a volume of 0.27 λmin×0.27 λmin×0.17 λmin, where λmin is the free space wavelength of the lowest frequency. To verify the design, the prototype antenna is manufactured and measured,and the measured results agree well with the simulation. The measurement results show that the antenna impedance bandwidth is 3.00~4.48 GHz(voltage standing wave ratio is less than 2), the relative bandwidth is 39.6%, and the isolation degree between the three ports is greater than 27 dB. In addition, in the entire frequency band, the horizontal polarization gain of the antenna is greater than 6 dB, the vertical polarization gain is greater than 3 dB, and the antenna has stable pattern characteristics and good orientation, indicating that the antenna has greater practical value in 5G small-pitch arrays and narrow spaces.

    Table and Figures | Reference | Related Articles | Metrics
    Binocular Vision Localization Method of Underwater Obstacles Based on Red Channel Prior
    WANG Yuhai,ZHANG Meiyan,CAI Wenyu,XIE Qinan
    Electronic Science and Technology    2023, 36 (8): 19-28.   DOI: 10.16180/j.cnki.issn1007-7820.2023.08.004
    Abstract116)   HTML8)    PDF(pc) (4755KB)(29)       Save

    During the underwater cruise of the autonomous underwater vehicle based on binocular vision, the images acquired by the binocular camera have low contrast and color distortion due to the attenuation effect of the water and the scattering effect of the suspended particles on the light, which leads to the low accuracy of underwater obstacle localization. In view of the above problems, this study adopts the red channel prior restoration algorithm to improve the quality of underwater imaging, obtains the binocular disparity map of obstacles according to the calibration parameters of the binocular camera, and proposes an underwater obstacle localization method based on depth disparity map fusion. The proposed method fuses the depth disparity map and the underwater restoration contour map, performs convex polygon detection on the fused image, obtains the contour of the obstacle, and extracts the effective depth information of the obstacle based on the contour information to realize the spatial positioning of the obstacle. The experimental results of underwater binocular localization show that the method can make the binocular stereo matching more ideal and effectively improve the accuracy of underwater obstacle localization.

    Table and Figures | Reference | Related Articles | Metrics
    Research on Single Leg Trajectory Planning of Bionic Quadruped Robot
    CHEN Jia,SAN Hongjun,CHEN Jiupeng,XU Yangyang
    Electronic Science and Technology    2023, 36 (8): 72-80.   DOI: 10.16180/j.cnki.issn1007-7820.2023.08.011
    Abstract115)   HTML11)    PDF(pc) (4105KB)(49)       Save

    Quadruped robot has become a hot topic in the international community because it can adapt to the complex terrain environment using the way of detached feet.In view of the lack of bionic characteristics of the foot trajectory of the quadruped robot and the inconsistency between the walking effect and that of mammals, this study establishes the three-joint leg mechanism with the help of SolidWorks, obtains the leg kinematic model, and analyzes the leg trajectory motion by referring to the laws of biological movement. Combined with bionic characteristics and based on compound cycloid and quintic polynomial, a foot trajectory of compound cycloid quintic polynomial is planned in this study, which not only meets the law of bionic motion, but also meets the requirements of trajectory planning. In this study, MATLAB and ADAMS co-simulation is used to realize the kinematic simulation of single leg, and the prototype and control system are used to realize the physical motion of single leg. The experimental results show that the single-leg motion of the quadruped robot can realize the planned trajectory curve motion, and the acceleration is 0 at the second second, which is basically consistent with the simulation trajectory effect, which verifies the effectiveness of the proposed algorithm to a certain extent.

    Table and Figures | Reference | Related Articles | Metrics
    A Dynamic-Static Load Balancing Algorithm Based on Improved Genetic Algorithm
    HU Yifei,BAO Ziqun,BAO Xiaoan
    Electronic Science and Technology    2023, 36 (9): 79-85.   DOI: 10.16180/j.cnki.issn1007-7820.2023.09.012
    Abstract115)   HTML7)    PDF(pc) (927KB)(47)       Save

    In view of the problems that current load balancing algorithm affects system efficiency under low load and poor distribution efficiency under high load, based on Nginx server, a dynamic and static load balancing algorithm based on improved genetic algorithm is proposed in this study. The algorithm chooses to use server performance parameters based on CPU performance, memory performance, disk I/O and network bandwidth as server node performance evaluation indexes and static weighted polling algorithm weights under low load, and designs a dynamic load balancing algorithm under high load based on the change of node performance utilization rate as a percentage of the cluster average load utilization rate by introducing operation conversion thresholds and dynamic. By introducing the improved genetic algorithm of operation transition threshold and dynamic triangular function operation probability as the threshold calculation method, the transformation of static algorithm dominant area into dynamic algorithm dominant area is calculated. This study designs comparison experiments to verify that the proposed algorithm has better load balancing effect when compared with weighted polling algorithm, probabilistic meritocracy algorithm and dnfs_conn algorithm in the experimental environment, and has about 15% improvement in the values of average response time and actual concurrent connections when compared with dnfs_conn algorithm.

    Table and Figures | Reference | Related Articles | Metrics
    Optimization of Automotive Permanent Magnet Brush DC Motor Efficiency Based on Maxwell Software
    CHEN Fanfan,SUN Ning
    Electronic Science and Technology    2023, 36 (11): 83-88.   DOI: 10.16180/j.cnki.issn1007-7820.2023.11.012
    Abstract115)   HTML1)    PDF(pc) (1648KB)(30)       Save

    Permanent magnet brush DC motor is widely used in the automotive field. "Carbon sink Plan" requires the further reduction of automotive energy consumption, so the optimization problem of motor efficiency is becoming increasingly prominent. Based on vehicle cooling fan with permanent magnet brush DC motor as the research object, in view of the motor inefficiencies, using the finite element analysis software Maxwell, through simulation and parametric analysis of the rotor lamination length, laminated slots, winding wire diameter, winding pitch, winding circle number influences the performance of motor, combined with the simulation results and actual production conditions, determine the feasible motor efficiency optimization scheme. According to the optimization scheme, the sample is made and measured. The test data show that the motor efficiency is improved by 4.8%, and the optimization effect of motor efficiency is obvious.

    Table and Figures | Reference | Related Articles | Metrics
    Research Status of Mid-Infrared Silicon-Based Optical Waveguides
    FENG Lulu,FENG Song,HU Xiangjian,CHEN Menglin,LIU Yong,WANG Di
    Electronic Science and Technology    2024, 37 (2): 36-45.   DOI: 10.16180/j.cnki.issn1007-7820.2024.02.006
    Abstract115)   HTML6)    PDF(pc) (2256KB)(46)       Save

    As the basic passive device in silicon photonic integrated chip, silicon-based optical waveguide is the channel for optical signal transmission. Because of its good performance and compatibility with CMOS(Complementary Metal Oxide Semiconductor) process, silicon-based optical waveguide has been widely used. Silicon photonic integrated circuits used in telecommunications and data centers have been gradually commercialized.The potential applications of mid-infrared band in free-space communication, sensing, environmental monitoring and other fields have attracted much attention from researchers in recent years.In this study, the research status of mid-infrared silicon-based optical waveguides in recent years is analyzed.The research results of waveguide material platform of SOI(Silicon on Insulator)、GOSI(Ge-on-SOI)、SOS(Si on Sapphire)、GOS(Ge-on-Si)、SGOS(SiGe-on-Si)、SON(Si-on Si3N4)、GON(Ge-on Si3N4) and manufacturing process platform of SOPS(Si on Porous Si), Pedestal, Freestanding, Suspended, LOCOS(Local Oxidation of Silicon), plasma structure are summarized. So far, the propagation loss of most monocrystalline silicon in MIR(Mid-Infrared) platform is about 0.7~3.0 dB·cm-1. The application prospects of different types of waveguides are discussed and compared, which provides a reference for the research and development, application and commercialization of mid-infrared silicon-based optical waveguides.

    Table and Figures | Reference | Related Articles | Metrics
    Classification Method of Steel Surface Defects Based on Multi-Scale Feature Fusion
    TIAN Zhixin,XU Zhen,MAO Jian,LIN Binbin,LIAO Wei
    Electronic Science and Technology    2024, 37 (2): 87-95.   DOI: 10.16180/j.cnki.issn1007-7820.2024.02.012
    Abstract113)   HTML7)    PDF(pc) (2470KB)(56)       Save

    In view of the low detection rate of steel surface defect classification, a surface defect classification method based on texture multi-scale feature fusion is adopted. Gabor filter and gray level co-occurrence matrix are used to establish multi-scale feature vectors of texture images. At the same time, convolution operation is used to extract features of texture images, and hybrid dilated convolution module is introduced to increase receptive field. The two feature vectors are fused to obtain the enhanced fused texture feature vector. The fused features are input into Long Short-Term Memory(LSTM) network in sequence to build a classification model, and the classification results are evaluated using the confusion matrix. The results show that the classification accuracy of this method on the NEU(Northeastern University) data set is 97.5%. The LSTM network, BP(Back Propagation) neural network, SVM(Support Vector Machine), KNN(K-Nearest Neighbor), CART(Classification And Regression Tree) and other classification methods are set up for comparative experiments. The results show that LSTM classification method performs best in multi-scale, and F1 index is the highest. Ablation experiments are conducted with BP network, LSTM network, SVM, KNN, CART, CNN, AlexNet and other methods to verify the universality of this method. This method fully exploits the multi-scale feature information of the texture image, and has important significance for the research of the classification method of steel surface defects.

    Table and Figures | Reference | Related Articles | Metrics
    An Arithmetic Optimization Algorithm Based Fault Section Location Method for Low Voltage Distribution Networks
    WANG Xinyang,WANG Ruiyang,WEI Yunbing
    Electronic Science and Technology    2023, 36 (12): 25-31.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.004
    Abstract112)   HTML4)    PDF(pc) (1215KB)(33)       Save

    In order to improve the accuracy and speed of fault location in low-voltage distribution networks and to ensure the safety of residents' electricity consumption, a method based on arithmetic optimization algorithm is proposed to realise fault section location. The arithmetic optimization algorithm has the advantages of simple structure, fast convergence speed and high accuracy. The IEEE 33-node distribution network model is selected and MATLAB is used to program and simulate the node branches, node switching states and adaptation functions of this model. Simulations are carried out for single-point and multi-point faults, as well as single-multi-point faults with signal distortion, and the simulation results are analysed. The results show that the arithmetic optimization algorithm's feature of separate local and global search is used to perform a local adequate search for the fault section location problem, resulting in an accurate location that can achieve an accuracy of 97%, outperforming the binary particle swarm algorithm, genetic algorithm and improved whale optimisation algorithm.

    Table and Figures | Reference | Related Articles | Metrics
    A Convolutional Neural Network Optimization Method for Fault Diagnosis of Power Transformer
    WANG Xuyang,YI Yingping,LI Tianfeng
    Electronic Science and Technology    2023, 36 (12): 79-86.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.011
    Abstract108)   HTML6)    PDF(pc) (1971KB)(38)       Save

    Traditional fault diagnosis methods have disadvantages such as incomplete coding and overly absolute coding boundaries, which are difficult to meet the actual needs of power grid operation and maintenance. Using the gas generated when a power transformer fails to diagnose the transformer fault is currently a popular research area for smart grid condition detection. However, the frequency of various types of faults in transformers varies greatly, which may result in incomplete fault sample information and insufficient data, and the traditional convolutional neural network models have problems such as unstable training process, low training accuracy and long time. Based on the transformer fault diagnosis technology of one dimensional convolutional neural network, this study proposes a new method of data enhancement while keeping the original data features unchanged, transforming expanded one dimensional data into two dimensional pictures to input into the two dimensional convolutional neural network diagnosis model, and improve the Adam optimization algorithm in the convolutional neural network model architecture. Diagnostic results indicate that the accuracy of network training reaches 96.20%. At the same time, it has higher convergence speed and generalization ability than the traditional one dimensional convolutional neural network fault diagnosis method(92.12%).

    Table and Figures | Reference | Related Articles | Metrics
    Review of the Research on the Core Algorithm of Table Tennis Robot
    SUO Fangfei,JI Yunfeng
    Electronic Science and Technology    2023, 36 (12): 16-24.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.003
    Abstract108)   HTML11)    PDF(pc) (1038KB)(43)       Save

    As a comprehensive research platform integrating machine vision and motion control, the research of table tennis robot has always attracted much attention. The complexity and high real-time of its mechanical mechanism, visual algorithm and motion control are important factors hindering the development of table tennis robot. In order to improve the competitive ability of table tennis robot, it is necessary to solve the problems of table tennis rotation measurement, accurate prediction of movement trajectory and selection of hitting strategy, which is also the focus of the research of table tennis robot in recent years. Based on the research progress in this field in recent years, this study focuses on the research results of table tennis robot vision core algorithm and ball return core algorithm, analyzes the advantages and disadvantages of different algorithms and prospects the future research trend, so as to provide a reference for the research and design of table tennis robot.

    Table and Figures | Reference | Related Articles | Metrics
    Design of Dual-Frequency Integrated Ultra-Wideband Ground Penetrating Radar Antenna
    LIN Xiangyu, ZENG Weihua
    Electronic Science and Technology    2024, 37 (4): 25-29.   DOI: 10.16180/j.cnki.issn1007-7820.2024.04.004
    Abstract107)   HTML6)    PDF(pc) (1869KB)(51)       Save

    In view of the disadvantages of time consuming and low efficiency caused by single frequency band of Ground Penetrating Radar (GPR) products, a dual-frequency integrated composite array antenna with center frequency of 400 MHz and 1 000 MHz and bandwidth of 200~1 500 MHz is designed and implemented, and is applied in Ultra-Wide Band (UWB)GPR system. The array antenna consists of three butterfly antennas with a center frequency of 400 MHz and six butterfly antennas with a center frequency of 1 000 MHz, and has the capability of simultaneous detection of dual frequency bands, which can realize the detection results of two different frequency bands in one survey line, overcome the problem of repeated detection of different frequencies of traditional GPR and enhance the practicability of GPR system. The proposed dual-frequency integrated GPR antenna has the characteristics of ultra-wideband, high gain and narrow beam. Its relative bandwidth is 153%, the maximum peak gain of 17.6 dBi is achieved in the whole bandwidth, and the narrowest half-power beam width is 7.6°. It provides a new antenna scheme for high resolution and high efficiency GPR applications.

    Table and Figures | Reference | Related Articles | Metrics
    Semi-Supervised Medical Image Segmentation Method Based on Meta-Learning and Neural Architecture Search
    YU Zhihong,LI Feifei
    Electronic Science and Technology    2024, 37 (1): 17-23.   DOI: 10.16180/j.cnki.issn1007-7820.2024.01.003
    Abstract106)   HTML6)    PDF(pc) (1999KB)(47)       Save

    Most medical image segmentation methods mainly focus on training and evaluating in the same or similar medical data domain, which need lots of pixel-level annotations. However, these models face challenges in out-of-distribution medical data set, which is known as "domain shift" problem. A fixed U-shaped segmentation structure is usually used to solve this problem, resulting in it not being better adapted to specific partition tasks.A gradient-based meta-learning and neural architecture search method is proposed in this study, which can adjust the segmentation network according to specific tasks to achieve good performance and have good generalization ability. This method mainly uses the specific task to carry out the architecture search module to further improve the segmentation effect, and then uses the gradient-based meta-learning training algorithm to improve the generalization ability.On the public dataset M&Ms, under the 5% label data, its Dice and Hausdorff distance are 79.62% and 15.38%. Under 2% label data, its Dice and Hausdorff distance are 74.03% and 17.05%.Compared with other mainstream methods, the proposed method has better generalization ability.

    Table and Figures | Reference | Related Articles | Metrics
    Visual Detection of Structural Cracks Using Depth Deformable Contour ModelLAI
    Bin ,WANG Sen
    Electronic Science and Technology    2023, 36 (9): 35-40.   DOI: 10.16180/j.cnki.issn1007-7820.2023.09.006
    Abstract106)   HTML4)    PDF(pc) (3171KB)(53)       Save

    At present, the deep learning instance segmentation method for crack detection mainly generates a boundary box through target detection to segment pixel by pixel mask, which will affect the detection effect of structural crack contour, and is accompanied by complex post-processing cost. To solve this problem, this study proposes to use deep snake algorithm model of deep deformable contour to identify and detect structural cracks. The robustness of the model is improved by data enhancement of the structural crack data set. At the same time, the pre training network parameters on the large image data set coco are transferred to the structural crack segmentation model as initialization by transfer learning. The experimental results on the self-made crack image data set show that the trained model can correctly identify the crack object and complete the segmentation of multiple crack targets at the same time. On the premise of the average detection time of 0.12 s, the AP50 reaches 75.4%. The comparison between the proposed methool and other deep learning models and edge detection algorithms also reflect the advantages of Deep Snake algorithm.

    Table and Figures | Reference | Related Articles | Metrics
    Optimization and Implementation of 2-Base Exponential Function Algorithm Based on FPGA
    CHENG Tiantian,SONG Yukun
    Electronic Science and Technology    2023, 36 (9): 66-72.   DOI: 10.16180/j.cnki.issn1007-7820.2023.09.010
    Abstract105)   HTML11)    PDF(pc) (1789KB)(56)       Save

    In view of the problems of small calculation value range and large error in common hardware implementation methods of exponential function, an improved hardware implementation method of base 2 exponential function y=2x combining polynomial and lookup table is proposed. The optimization algorithm adopts the interval division to compress the input x to (-1/512,1/512) and then performs the Taylor series expansion of the exponential function to ensure that the accuracy reaches 10-16 in the double-precision floating-point when the Taylor series is expands to x4. And storage resource consumption is reduced by optimizing intermediate data storage policies. Hardware design implementation and performance test of the improved algorithm on Xilinx XC7K325T FPGA(Field Programmable Gate Array) are performed by using Verilog HDL. The results show that within the value range that double-precision floating-point numbers can represent, the full-domain exponential function calculation can be supported by the designed circuit with less storage consumption, and the calculation accuracy is not less than 10-16.

    Table and Figures | Reference | Related Articles | Metrics
    Super-Resolution Imaging of Laminate Debonding Defects via Deconvolutional Neural Network and Ultrasound Guided Waves
    YUE Shengyao,XU Baiqiang,XU Guidong,XU Chenguang,ZHANG Sai
    Electronic Science and Technology    2023, 36 (8): 7-13.   DOI: 10.16180/j.cnki.issn1007-7820.2023.08.002
    Abstract104)   HTML10)    PDF(pc) (1987KB)(39)       Save

    Traditional ultrasonic guided wave imaging detection methods are difficult to accurately characterize structural damage details. In order to obtain detailed features of the damage, the deconvolutional neural network model via deep learning is proposed to investigate the super-resolution imaging problem of subwavelength debonding defects in laminate plates. Initial imaging results are obtained by finite element simulation with the total focusing method. The labeled 12 550 damage images are trained and tested based on extended database expanded by data enhancement method. The results show that compared with the original full-focus imaging algorithm, the deconvolution neural network model improves the accuracy of damage location by 5%, the imaging accuracy is higher than 91%, and the positioning error is lower than 1.8 mm, indicating that the proposed method can significantly improve the resolution of network imaging results and better display the details of subwavelength damage. The above results show that the proposed method has high detection efficiency and does not require manual experience, and has good application value in engineering practice.

    Table and Figures | Reference | Related Articles | Metrics
    Permanent Magnet Synchronous Motor Control Based on Super-Twisting Sliding Film Observer
    HUANG Chengcheng,JIN Hai,LU Wenqi
    Electronic Science and Technology    2023, 36 (11): 8-13.   DOI: 10.16180/j.cnki.issn1007-7820.2023.11.002
    Abstract104)   HTML5)    PDF(pc) (1217KB)(36)       Save

    In view of the high precision requirement of permanent magnet synchronous motor in some workplaces, an algorithm based on second-order Super-Twisting synovium theory is proposed based on the mathematical model of permanent magnet synchronous motor. In this method, the motor speed and position information are obtained by calculating the observed value of the back electromotive force of the motor, so as to realize the sensorless control of the permanent magnet synchronous motor. According to Lyapunov stability theory, the observer converges. MATLAB/Simulink is used to build a simulation model of the control system to verify the feasibility of the control system. The simulation results show that compared with the traditional synovial film observer, the proposed algorithm can effectively reduce the chattering of synovial film, improve the estimation accuracy and response speed of the system, and enables it to track the rotor position and velocity information better.

    Table and Figures | Reference | Related Articles | Metrics
    Research Progress of SiGe Electro-Optical Modulator
    WANG Di,FENG Song,CHEN Menglin,LIU Yong,HU Xiangjian,FENG Lulu
    Electronic Science and Technology    2024, 37 (2): 46-54.   DOI: 10.16180/j.cnki.issn1007-7820.2024.02.007
    Abstract104)   HTML4)    PDF(pc) (1975KB)(42)       Save

    Photon modulator is the core device in the optical fiber communication system, which mainly modulates the optical signal to realize the conversion of the signal from the electrical domain to the optical domain. With the development of silicon based semiconductor technology, silicon based photonic modulator has gradually become a mainstream silicon photonic device. The realization of GHz bandwidth modulator based on silicon technology also lays a foundation for the development of silicon photonics. As a high performance optical modulator for short distance off optical interconnection, SiGe optical absorption modulator has gotten much attention. This study discusses the development status of high-performance SiGe electro-optic modulator, mainly analyzes the research progress of silicon based photonic modulator at home and abroad,and discusses the electrical modulation structures such as PIN, PN junction, which provides a way to continue to develop high-speed, low loss photonic modulator in the future.

    Table and Figures | Reference | Related Articles | Metrics
    A Novel Dual-band Conformal Antenna On Unmanned Aerial Vehicle
    XU Ping,MA Tianyi,LIU Fei
    Electronic Science and Technology    2023, 36 (12): 95-98.   DOI: 10.16180/j.cnki.issn1007-7820.2023.12.013
    Abstract103)   HTML11)    PDF(pc) (2971KB)(41)       Save

    In order to meet the demands of dual-band and integrated conformal antenna for UAV (Unmanned Aerial Vehicle) swarm communication systems, an integrated wing conformal antenna is designed and fabricated. The antenna is realized by combining a broadband dipole antenna and a low-pass and high-pass LC combiner. The antenna element is realized by three pairs of open branch lines to achieve effective choking characteristics and broadband omnidirectional radiation. The combiner consists of K-m LC filters, which are used to realize dual-band operation with low insertion loss and high rejection. The measured Voltage Standing Wave Ratio(VSWR) of antenna is less than 2, the measured gain is more than 0 dBi, and the isolation is better than 30 dBc between L and S frequency band. By mounting only one wing conformal antenna, a dual-band simultaneous operation can be realized which meets the need of UAV swarm data link system for high communication rate. The proposed antenna is whole arranged in the tail wing of UAV to enhance the integration capability of the UAV swarm, and can be widely used in UAV field.

    Table and Figures | Reference | Related Articles | Metrics
Download

Monthly,Founded in September 1987
Competent Authorities:
Ministry of Education of the People's Republic of China
Sponsored by:Xidian University
Chief Editor:Liao Guisheng
Executive Editor:Wan Liancheng
Editor:Hei Lei
Editor and Publisher:
The Editorial Department of Electronic Science and Technology
Distribution Abroad:
China Intermational Book Trading Corporation
P.O.BOx 399,Beijing 100044,China
Address:
P.O.Box 375,2 Taibai Road(South),Xi'an 710071,China
Tel/Fax:0086-029-88202440
Website:http://www.dianzikeji.org
E-mail:dzkj@mail.xidian.edu.cn
Unit Price:$20.00