王思敏,周利军,刘博凯,王东阳,黄文勋,王继来

• 1:西南交通大学电气工程学院
• 2:中铁第一勘察设计院集团有限公司

摘要(Abstract)：

电气化铁路接触网覆冰是制约列车安全运行的隐患，准确识别覆冰类型对制定科学的除冰策略具有重要意义。为此，针对接触网覆冰类型的识别开展研究。通过对真实覆冰图像、气候模拟实验平台覆冰图像及等效模型物理覆冰图像进行标注与预处理，构建包含多种典型覆冰类型的接触网覆冰图像数据集，该数据集共包含2 800张图像，涵盖雨凇、雾凇、混合凇、湿雪和无覆冰5种类型；从自适应特征融合和空间感知能力两个方面优化Swin Transformer模型，引入自适应特征融合模块(α＿HybridAttention),对不同特征层次的贡献进行自适应调整，以实现多尺度特征动态整合，提高模型对不同覆冰类型细微差异的识别精度，同时，采用坐标注意力模块(CA),进一步增强模型的空间感知能力，通过关注输入特征图中的关键区域，有效捕捉接触网覆冰的几何特征，提出基于CA-DFM-Swin模型的接触网覆冰类型识别方法；通过消融实验及方法比对试验，验证分析了所提方法的识别效果。结果表明：CA-DFM-Swin模型的分类准确率达到98.51%,F1分数达到98.46%,较基准Swin Transformer分别提升2.29个百分点和3.87个百分点；在混合凇和湿雪等易混淆类别的识别中表现出显著优势，对隧道反光、雪地背景等复杂场景具有较强鲁棒性，验证了所提方法的有效性。

关键词(KeyWords)： 接触网覆冰分类;Swin Transformer;动态特征融合;深度学习

Abstract:

Keywords:

基金项目(Foundation): 国家自然科学基金(U2469212)

作者(Author): 王思敏,周利军,刘博凯,王东阳,黄文勋,王继来

参考文献(References)：

• [1] YANG Z F,YANG Z Q,HUANG G Z,et al.Effect of Ice-covered Catenary on Dynamic Interactions of Pantograph-Catenary System and Limited Speed of Trains [J].Cold Regions Science and Technology,2024,228:104331.
• [2] SONG Y,DUAN F C,WU F P,et al.Assessment of the Current Collection Quality of Pantograph-Catenary with Contact Line Height Variability in Electric Railways [J].IEEE Transactions on Transportation Electrification,2022,8(1):788-798.
• [3] WANG S H,JIANG X L.Progress in Research on Ice Accretions on Overhead Transmission Lines and Its Influence on Mechanical and Insulating Performance [J].Frontiers of Electrical and Electronic Engineering,2012,7:326-336.
• [4] ZHANG Z J,ZHANG H,YUE S,et al.A Review of Icing and Anti-icing Technology for Transmission Lines [J].Energies,2023,16(2):601.
• [5] ROSSI A,JUBAYER C,KOSS H,et al.Combined Effects of Wind and Atmospheric Icing on Overhead Transmission Lines [J].Journal of Wind Engineering and Industrial Aerodynamics,2020,204:104271.
• [6] FAN C J,JIANG X L.Analysis of the Icing Accretion Performance of Conductors and Its Normalized Characterization Method of Icing Degree for Various Ice Types in Natural Environments [J].Energies,2018,11(10):2678.
• [7] KANDELIN N.Icing Factors Affecting Railway Traffic [D].Tampere:Tampere University,2021.
• [8] 吴广宁，王敏，郭裕钧，等.高速铁路接触网融冰机理及关键技术：系统性综述 [J].铁道学报，2025,47(3):40-53.WU Guangning,WANG Min,GUO Yujun,et al.Ice Melting Mechanism and Key Technologies for High-speed Railway Contact Networks:a Systematic Review [J].Journal of the China Railway Society,2025,47(3):40-53.
• [9] 吴磊，徐梦楠，张华鹏，等.电气化铁路接触网冲击振动除冰仿真分析 [J].铁道学报，2025,47(1):47-53.WU Lei,XU Mengnan,ZHANG Huapeng,et al.Simulation Analysis of Impact Vibration De-icing for Overhead Catenary of Electrified Railway [J].Journal of the China Railway Society,2025,47(1):47-53.
• [10] MA G M,LI C R,QUAN J T,et al.A Fiber Bragg Grating Tension and Tilt Sensor Applied to Icing Monitoring on Overhead Transmission Lines [J].IEEE Transactions on Power Delivery,2011,26(4):2163-2170.
• [11] HUANG Z D,WU R R,ZHANG W,et al.Real-time Monitoring Method of Icing on Overhead Transmission Lines Based on Multi-sensor Information Fusion [J].IEEE Access,2024,12:160234-160244.
• [12] ZHANG M,XING Y M,ZHANG Z G,et al.Design and Experiment of FBG-based Icing Monitoring on Overhead Transmission Lines with an Improvement Trial for Windy Weather [J].Sensors,2014,14(12):23954-23969.
• [13] LI H Y,CHEN Y L,ZHANG G J,et al.Transmission Line Ice Coating Prediction Model Based on EEMD Feature Extraction [J].IEEE Access,2019,7:40695-40706.
• [14] HUANG X B,ZHANG F,LI H S,et al.An Online Technology for Measuring Icing Shape on Conductor Based on Vision and Force Sensors [J].IEEE Transactions on Instrumentation and Measurement,2017,66(12):3180-3189.
• [15] ZHANG Y,ZHAN Y F,SHANG J H,et al.A Classification Method of Ice Types on Conductors and Insulators by Image Recognition [J].Electrical Engineering,2025,107(9):11377-11392.
• [16] YANG L,JIANG X L,HAO Y P,et al.Recognition of Natural Ice Types on In-service Glass Insulators Based on Texture Feature Descriptor [J].IEEE Transactions on Dielectrics and Electrical Insulation,2017,24(1):535-542.
• [17] LU J Z,YE Y,XU X J,et al.Application Research of Convolution Neural Network in Image Classification of Icing Monitoring in Power Grid [J].EURASIP Journal on Image and Video Processing,2019,2019:49.
• [18] HU M H,HE J C,ALSABAAN M.Image Identification Method of Ice Thickness on Transmission Line Based on Visual Sensing [J].Mobile Networks and Applications,2023,28(5):1783-1792.
• [19] WENG B J,GAO W,ZHENG W C,et al.Newly Designed Identifying Method for Ice Thickness on High-voltage Transmission Lines via Machine Vision [J].High Voltage,2021,6(5):904-922.
• [20] NUSANTIKA N R,XIAO J,HU X G.Precision Ice Detection on Power Transmission Lines:a Novel Approach with Multi-scale Retinex and Advanced Morphological Edge Detection Monitoring [J].Journal of Imaging,2024,10(11):287.
• [21] 丁宇鸣，刘金朝，徐晓迪，等.基于SVM分类受电弓振动干扰信号识别 [J].铁道学报，2021,43(9):38-45.DING Yuming,LIU Jinzhao,XU Xiaodi,et al.Recognition of Pantograph Vibration Interference Signal Based on SVM Classification [J].Journal of the China Railway Society,2021,43(9):38-45.
• [22] 郭维平，韩志伟，杨成，等.基于深度强化学习的局部像素级接触网图像增强方法研究 [J].铁道学报，2023,45(12):67-73.GUO Weiping,HAN Zhiwei,YANG Cheng,et al.Research on Local Pixel-level Catenary Image Enhancement Method Based on Deep Reinforcement Learning [J].Journal of the China Railway Society,2023,45(12):67-73.
• [23] LIU Z,LIN Y T,CAO Y,et al.Swin Transformer:Hierarchical Vision Transformer Using Shifted Windows [C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV).Piscataway:IEEE,2021.
• [24] CHEN J Q,WANG H Q,LI Y Y,et al.Real-time Asphalt Pavement Ice Detection and Annotation with a Transformer-based Model Framework [J].Engineering Applications of Artificial Intelligence,2025,152:110758.
• [25] LI Y F,GONG C,DENG T,et al.Partial Discharge Pattern Recognition Based on Swin Transformer for Power Cable Fault Diagnosis in Modern Distribution Systems [J].Processes,2025,13(3):852.
• [26] 方春华，钮非凡，张鑫，等.基于Swin Transformer的引流板缺陷超声检测技术及智能识别方法 [J/OL].声学技术，2025-03-26.https://doi.org/10.16300/j.cnki.1000-3630.24111102.FANG Chunhua,NIU Feifan,ZHANG Xin,et al.Ultrasonic Detection Technology and Intelligent Recognition Method of Drainage Board Defects Based on Swin Transformer [J].Technical Acoustics,2025-03-26.https://doi.org/10.16300/j.cnki.1000-3630.24111102.
• [27] XU X Y,ZHAO M,SHI P X,et al.Crack Detection and Comparison Study Based on Faster R-CNN and Mask R-CNN [J].Sensors,2022,22(3):1215.
• [28] KIANFAR A,RAZZAGHI P,ASGARI Z.Integrating Convolutional Layers and Biformer Network with Forward-Forward and Backpropagation Training [J].Scientific Reports,2025,15:7230.
• [29] WEN G,LI S B,LIU F C,et al.YOLOv5s-CA:a Modified YOLOv5s Network with Coordinate Attention for Underwater Target Detection [J].Sensors,2023,23(7):3367.
• [30] HU K,ZHANG Z Z,NIU X R,et al.Retinal Vessel Segmentation of Color Fundus Images Using Multiscale Convolutional Neural Network with an Improved Cross-entropy Loss Function [J].Neurocomputing,2018,309:179-191.
• [31] ZHOU Y F,WANG X,ZHANG M C,et al.MPCE:a Maximum Probability Based Cross Entropy Loss Function for Neural Network Classification [J].IEEE Access,2019,7:146331-146341.