Table of Content

15 April 2026 Volume 39 Issue 4

  

Multi-Scale Channel Attention Based on Feature Fusion Method for Skin Disease Segmentation

XU Kuncai, DENG Zhan, LIU Xuan, ZHANG Ning, LU Jiadong

Electronic Science and Technology. 2026, 39(4):  1-7.  doi:10.16180/j.cnki.issn1007-7820.2026.04.001

Abstract ( 122 )   HTML ( 16 )   PDF (3369KB) ( 88 )  

Figures and Tables | References | Related Articles | Metrics

In view of the problem of low image segmentation accuracy caused by factors such as irregular shapes, blurred edges, and foreign body occlusion of melanoma, this study proposes a MSFFNet(Multi Scale Feature Fusion Network). In the encoding stage, a pyramid split attention module is introduced to expand the receptive field and capture feature information at different scales. The channel attention mechanism is used to recalibrate the weights of different channels, and then point-by-point multiplication fusion is performed with the original feature space. The Dice Loss function is adopted for end-to-end optimization to alleviate the negative impact of class imbalance in samples and further improve the segmentation performance of the network model. The segmentation performance of the model is verified on the ISIC2018 skin disease image dataset. Experimental results show that the Dice and IoU(Intersection over Union) of the proposed method are 89.40% and 82.27% respectively. Compared with mainstream segmentation algorithms, the proposed segmentation method is more similar to the results of manual segmentation by doctors.

Research on Road Traffic Sign Recognition in Complex Scenario

HE Qianwei, ZHANG Xuanxiong

Electronic Science and Technology. 2026, 39(4):  8-18.  doi:10.16180/j.cnki.issn1007-7820.2026.04.002

Abstract ( 96 )   HTML ( 7 )   PDF (4607KB) ( 69 )  

Figures and Tables | References | Related Articles | Metrics

In view of the problems of low recognition accuracy and high missed detection rate of traffic signs under complex environmental conditions, an improved traffic sign recognition algorithm YOLO-Traffic based on YOLOv8(You Only Look Once version 8) is proposed. The multi-scale information extraction ability of the network is enhanced through scale sequence feature fusion and triple feature coding. The local fine-grained features of traffic signs are fully extracted by adding a small target detection layer and refining the local feature mapping. The CA(Coordinate Attention) attention mechanism is introduced into the backbone network to enhance the model's ability to focus on key regions. The new metric NWD(Normalized Wasserstein Distance) is adopted to replace the CIoU(Complete Intersection over Union) in the regression loss function of the detection head, strengthening the detection ability for small targets. The experimental results show that the mAP@0.5(mean Average Precision) of the original model is 90.4%, the mAP@0.5:0.95 is 63.2%, and the model size is 6.3 MB. The mAP@0.5 of the improved model is 95.5%, the mAP@0.5:0.95 is 67.5%, and the model size is 5.2 MB. Compared with the original model, the volume of the improved model is reduced by 17.5%. The improved algorithm reduces the volume of model parameters while enhancing detection accuracy, and can meet the requirements of various complex road conditions and lightweight in practical application scenarios.

Optimization of Dynamic Matrix Control Based on Cauchy Opposite Vulture Fusion Algorithm

ZHUANG Haohan, WANG Yagang

Electronic Science and Technology. 2026, 39(4):  19-27.  doi:10.16180/j.cnki.issn1007-7820.2026.04.003

Abstract ( 66 )   HTML ( 4 )   PDF (1521KB) ( 51 )  

Figures and Tables | References | Related Articles | Metrics

In view of the problem that controller parameters are difficult to select for complex systems with large time delays and large inertia, this study proposes a Cauchy-reverse vulture fusion algorithm to optimize the parameter selection of DMC (Dynamic Matrix Control). By introducing Cauchy mutation, lens opposition-based learning, and the exploration mechanism of the eagle algorithm into the framework of the African vulture algorithm, the search range is expanded and the late-stage exploration ability is enhanced, thereby improving the global optimization capability of the algorithm. The effectiveness of the proposed algorithm in improving optimization accuracy is verified through multiple test functions. The proposed algorithm is applied to the DMC control systems of desulfurization towers and distillation tower models. The control horizon and prediction horizon parameters are optimized through iteration, and set-point tracking control simulation and anti-interference tests are carried out. The simulation test results show that the optimized DMC controller can achieve fast and accurate set-point tracking, significantly reducing the overshoot and adjustment time.

Lightweight PCB Defect Detection Algorithm Based on Dynamic Snake Convolution

XIN Changming, WANG Bo

Electronic Science and Technology. 2026, 39(4):  28-34.  doi:10.16180/j.cnki.issn1007-7820.2026.04.004

Abstract ( 81 )   HTML ( 3 )   PDF (1324KB) ( 59 )  

Figures and Tables | References | Related Articles | Metrics

In view of the problems of insufficient detection accuracy and large number of parameters in the existing PCB(Printed Circuit Board) defect detection algorithms, a high-precision lightweight detection algorithm YOLO-DLN(YOLO-Dynamic Lightweight Network) is proposed based on YOLOv8(You Only Look Once version 8). Dynamic serpentine convolution is introduced into the backbone network to capture more fine structural features of the PCB surface, and the extracted features are processed by the lightweight detection head LSCDHead(Lightweight Shared Convolutional Detection Head). Features of different scales share the same set of convolution to reduce the number of model parameters and perform group normalization operations. The NWD(Normalized Wasserstein Distance) loss function with low sensitivity to the target scale is utilized to evaluate the bounding box similarity, further improving the detection accuracy of the proposed method. In the experiment, YOLO-DLN increases mAP50(mean Average Precision50) and MAP50:95 by 4.4% and 4.6% respectively while using 88.1% of the parameters of the original model. The experimental results show that YOLO-DLN can maintain a high detection accuracy while significantly reducing the number of model parameters, and is suitable for PCB defect detection in resource-constrained environments.

An Implementation of Spectral Noise Shaping for LC3plus on FPGA

BAO Guoan, WANG Faxiang

Electronic Science and Technology. 2026, 39(4):  35-41.  doi:10.16180/j.cnki.issn1007-7820.2026.04.005

Abstract ( 68 )   HTML ( 7 )   PDF (3052KB) ( 55 )  

Figures and Tables | References | Related Articles | Metrics

SNS(Spectral Noise Shaping) in the LC3plus(Low Complexity Communication Codec plus) protocol is one of the important means to improve audio compression quality. However, this part involves many complex mathematical operations, which tends to consume a large number of clock cycles when executed in a CPU(Central Processing Unit) or microcontroller. To address this issue, this study designs an SNS hardware IP(Intellectual Property) core based on FPGA(Field Programmable Gate Array). A hardware architecture is proposed to coordinate the interaction between various internal data of SNS. The hardware overhead is reduced by reusing a basic operation module. Functional simulation and on-board testing are conducted on the SNS hardware module. The test results show that the SNS hardware module complies with the LC3plus protocol standard, can perform noise shaping on spectral data, and has low hardware overhead. Compared with the execution of software source code by STM32 series microcontrollers, the data processing speed of the proposed SNS hardware module is significantly improved.

Research on Online Monitoring and Early Warning of Fire Hydrant Status Based on YOLOv8

ZHANG Yuanrong, ZHANG Xuanxiong, DENG Chenxin, SHEN Tuo

Electronic Science and Technology. 2026, 39(4):  42-49.  doi:10.16180/j.cnki.issn1007-7820.2026.04.006

Abstract ( 81 )   HTML ( 5 )   PDF (2862KB) ( 49 )  

Figures and Tables | References | Related Articles | Metrics

Fire hydrants play a crucial role in the railway fire protection system. However, relevant monitoring devices are only suitable for single specific tasks, and their monitoring accuracy is easily affected by the railway environment, imposing high requirements on the performance of hardware devices. In view of this problem, this study proposes a research method for online monitoring and early warning of fire hydrant status based on YOLOv8(You Only Look Once version 8). CPU(Central Processing Unit)-GPU(Graphics Processing Unit) heterogeneous computing and modified activation functions are adopted to ensure that the device can achieve acceleration under the same accuracy, with relatively low requirements for edge hardware performance. The CPU-GPU heterogeneous strategy is used to optimize the image detection frame rate, and the H-Swish function is applied to improve the computational efficiency of the activation function, enhancing the model's fitting and expression capabilities. Experimental results show that compared with traditional methods, the proposed method accelerates the detection speed by 1.85 times, increases the detection frame rate by approximately 16.5%, and achieves an accuracy of 98.1%, meeting the requirements of practical applications.

Design of Non-Sensor Control System for Rotor Spinning Direct Drive High-Speed Brushless Motor

ZHANG Yiheng, ZHANG Jianxin

Electronic Science and Technology. 2026, 39(4):  50-57.  doi:10.16180/j.cnki.issn1007-7820.2026.04.007

Abstract ( 81 )   HTML ( 9 )   PDF (2272KB) ( 48 )  

Figures and Tables | References | Related Articles | Metrics

In view of the problems of difficult installation of motor sensors and unstable rotational speed in the single-spindle high-speed drive system of traditional rotary cup spinning machines, this study proposes a sensorless control algorithm that adopts high-speed brushless DC motor sliding mode adaptive observation+PID(Proportional Integral Derivative) control, combined with the motor vector control mode to achieve precise motor control. By improving the structure of the sliding mode observer, the slow approach rate and motor buffing problems of the traditional sliding mode observation algorithm are reduced, achieving rapid and accurate estimation of the rotor position and enhancing the robustness of the system. The drive and control system of the rotary cup spinning motor is developed by adopting embedded control technology. The control algorithm is transplanted and implemented, which solves the problems of low speed, large speed buffering and difficult sensor installation in the traditional rotary cup spinning motor system. Through experimental tests, the goal of high stable rotational speed and sensorless control of the high-speed brushless motor system for rotary cup spinning is achieved. The chattering rate at a high motor speed of 100 000 rpm is within 2%.

A Bolt Loosening Detection Technology Based on RFID Tags

CHEN Zhengyi, JIA Jiangming, WAN Changjiang, HOU Liangmei, HE Leiying

Electronic Science and Technology. 2026, 39(4):  58-65.  doi:10.16180/j.cnki.issn1007-7820.2026.04.008

Abstract ( 76 )   HTML ( 4 )   PDF (4008KB) ( 42 )  

Figures and Tables | References | Related Articles | Metrics

In the field of bolt loosening detection, manual inspection has problems such as low efficiency and insufficient accuracy. Detection relying on conventional sensors uses cables for power supply and data collection, resulting in high installation and maintenance costs. To address the above issues, this study designs an ultra-high frequency radio frequency identification devices tag for bolt loosening detection based on radio frequency identification technology. When the bolt loosens, the tag antenna installed on the nut and the connected component deforms, which causes a change in the resonant frequency of the antenna. This further leads to a variation in the backscattered signal strength of the tag. The wireless monitoring of the bolt loosening state is realized by remotely reading the change in the backscattered signal strength. Experimental results show that as the bolt loosening angle increases, the backscattered signal strength of the tag shows a linear downward trend, with a correlation coefficient as high as 0.978 74, which confirms that the proposed detection scheme can achieve accurate identification of the bolt loosening state within the loosening angle range of 0° to 8°.

Radiation Effects of Strained Si MOS Devices

LI Jiajun, HAO Minru

Electronic Science and Technology. 2026, 39(4):  66-70.  doi:10.16180/j.cnki.issn1007-7820.2026.04.009

Abstract ( 55 )   HTML ( 1 )   PDF (1863KB) ( 22 )  

Figures and Tables | References | Related Articles | Metrics

With the rapid development of aerospace industry, all kinds of radiation particles in the radiation environment have become one of the key problems that restrict the reliability of electronic components in spacecraft. The single event effect of devices is becoming more and more serious. In view of the degradation of strained MOSFET(Metal Oxide Semiconductor Field Effect Transistor) devices caused by single particle effect in irradiation environment, Sentaurus TCAD software is used to simulate the impact process of 50 nm strain silicon MOS device by heavy ions to verify the existence of funnel electric field and the influence of bipolar amplification effect on the drain transient current, and to study the single particle irradiation characteristics of MOS device by stress, different injection positions and different bias parameters. The results show that the peak value of the transient current increases with the increase of drain bias and LET(Linear Energy Transfer), and decreases with the increase of substrate doping concentration in the sensitive region of the device.

A Small Sample Bearing Fault Diagnosis Method Based on Improved Gaussian Prototype Network

MA Xiangchun, DONG Baoli

Electronic Science and Technology. 2026, 39(4):  71-78.  doi:10.16180/j.cnki.issn1007-7820.2026.04.010

Abstract ( 69 )   HTML ( 4 )   PDF (1650KB) ( 41 )  

Figures and Tables | References | Related Articles | Metrics

In view of the problem of low fault diagnosis accuracy caused by insufficient fault sample data for rotating machinery bearings in practical applications, this study proposes an improved Gaussian prototype network for few-shot fault diagnosis. Vibration signals are converted into time-frequency images using continuous wavelet transform, and a convolutional attention module is introduced to optimize the residual network structure, which serves as the feature extraction network of the Gaussian prototype network. The FCM(Fuzzy C-Means Clustering,) algorithm is incorporated to optimize the Gaussian prototype network's ability to distinguish subtle fault categories through a soft classification strategy. Few-shot experiments under varying working conditions are conducted using the bearing datasets from Case western reserve university and the university of Paderborn. The experimental results show that the accuracy of the improved Gaussian prototype network reaches 91.17% and 89.95%, respectively, indicating that the proposed method achieves a significant improvement in fault recognition accuracy when compared with other models.

Research on Motion Control Strategies for High-Precision Laser Processing Numerical Control Systems

LUO Xiaoqing

Electronic Science and Technology. 2026, 39(4):  79-84.  doi:10.16180/j.cnki.issn1007-7820.2026.04.011

Abstract ( 56 )   HTML ( 4 )   PDF (1220KB) ( 48 )  

Figures and Tables | References | Related Articles | Metrics

In order to improve the motion trajectory accuracy of laser machining computer numerical control system, an improved arc trajectory interpolation algorithm combined with hardware architecture is proposed. The principle of digital integral arc interpolation algorithm, which calculates the intermediate value according to the arc of the motion path trajectory, is studied. An improved two-stage interpolation algorithm based on the combination of arc coarse interpolation algorithm and fine interpolation algorithm based on time division method is proposed. The hardware system architecture including embedded microprocessor and stepper motor controller is designed. The embedded microprocessor is used to implement the coarse interpolation algorithm, while the stepper motor controller is used to implement the fine interpolation algorithm. The experimental results show that compared with the traditional algorithm, the accuracy of the improved two-stage interpolation algorithm is increased by 26%, the error is reduced by 28%, and the speed is increased by 36%, which proves that the proposed motion control strategy has certain engineering significance.