苏云霞,施志伟,周广丽,费周辰,钱佳成,吴佳成,邵洁
SU Yunxia,SHI Zhiwei,ZHOU Guangli,FEI Zhouchen,QIAN Jiacheng,WU Jiacheng,SHAO Jie

• 1:国网上海市电力公司信息通信公司
• 2:上海电力大学电子与信息工程学院

摘要(Abstract)：

基于图像识别技术的电力架空线路异物检测在复杂天气条件下存在图像质量受水汽噪声干扰、异常天气图像标签不足、测试数据与训练数据分布差异大等问题，导致检测模型性能下降。为此，提出一种基于自学习的域间融合与对抗注意力方法，实现跨天气条件的电力架空线路异物检测。首先，通过深层特征混合策略和成对的注意力对抗模块，实现不同天气条件下电力架空线路异物颜色、边缘等多样信息的融合提取，使模型关注多种天气下异物的共同特征。然后，采用对抗学习策略对已标注图像的检测损失和模型的判别损失进行调整，实现跨天气条件的异物检测。最后，在跨天气多场景电力架空线路异物测试集中进行验证，结果表明：与传统目标检测方法相比，所提方法在雨天和雾天等未标注场景下，也能够保持较高的异物检测精确率。
Foreign object detection for overhead power transmission lines based on image recognition technology faces several challenges under complex weather conditions, such as image quality degradation due to moisture and noise, insufficient labeled images for abnormal weather, and significant distribution differences between test and training datasets. These issues can lead to a decline in the performance of detection models. To address these challenges, a domain adaptation and adversarial attention method, based on self-taught learning, is proposed for foreign object detection across varying weather conditions. First, through a deep feature fusion strategy and paired adversarial attention modules, the method integrates and extracts diverse information, such as color and edges, of foreign objects across varying weather conditions, enabling the model to focus on the common features of foreign objects. Next, an adversarial learning strategy is employed to adjust the detection loss of labeled images and the model's discrimination loss to detect the foreign objects. Finally, the proposed method is validated on a test dataset of foreign objects on overhead power transmission lines in multiple scenarios across varying weather conditions. The results show that, compared to traditional object detection methods, the proposed method maintains high detection accuracy under rainy and foggy weather, as well as other unlabeled scenarios.

关键词(KeyWords)： 天气条件;电力架空线路;异物检测;目标检测;成对注意力
weather condition;overhead power transmission line;foreign object detection;object detection;pairwise attention

Abstract:

Keywords:

基金项目(Foundation): 国家自然科学基金(62202286);; 国网上海市电力公司科技项目(B3090F24001K)

作者(Author): 苏云霞,施志伟,周广丽,费周辰,钱佳成,吴佳成,邵洁
SU Yunxia,SHI Zhiwei,ZHOU Guangli,FEI Zhouchen,QIAN Jiacheng,WU Jiacheng,SHAO Jie

DOI: 10.19585/j.zjdl.202508013

参考文献(References)：

• [1]邓益民，汤智谦，李红兵，等.基于LOD的大规模输电线路场景实时渲染算法优化[J].计算机与现代化，2017(1):115-118.DENG Yimin,TANG Zhiqian,LI Hongbing,et al.Realtime rendering algorithm optimization for large scale transmission lines based on LOD[J].Computer and Modernization,2017(1):115-118.
• [2]段中兴，张雨明，马佳豪.基于改进YOLOv4的电力设备红外图像识别[J].激光与光电子学进展，2022,59(24):2410002.DUAN Zhongxing,ZHANG Yuming,MA Jiahao.Infrared image recognition of power equipment using improved YOLOv4[J].Laser&Optoelectronics Progress,2022,59(24):2410002.
• [3]卞蓓蕾，江炯，刘鹏.输电线路人机协同立体巡检系统设计[J].浙江电力，2022,41(5):61-68.BIAN Beilei,JIANG Jiong,LIU Peng.Design of a threedimensional transmission line inspection system based on human-machine collaboration[J].Zhejiang Electric Power,2022,41(5):61-68.
• [4]熊凯飞，樊绍胜，吴静.基于改进YOLOv4的雾天变电站电力设备识别方法[J].无线电工程，2022,52(8):1504-1512.XIONG Kaifei,FAN Shaosheng,WU Jing. Identification method of substation power equipment in foggy weathers based on improved YOLOv4[J]. Radio Engineering,2022,52(8):1504-1512.
• [5]杨璐雅.基于图像处理的瓷质绝缘子污秽程度检测技术研究[D].西安：西安工程大学，2019.YANG Luya. Research on detection technology of contamination degree of porcelain insulators based on image processing[D].Xi’an:Xi’an Polytechnic University,2019.
• [6]杨宁，高飞，杨洋，等，雨天变电站设备智能视觉检测算法与实时实现技术[J].测控技术，2023,41(11):20-27.YANG Ning,GAO Fei,YANG Yanget,et al.Detection of substation equipment in rainy days and real-time implementation[J].Measurement&Control Technology,2023,41(11):20-27.
• [7]季世超，曲星合，宋庆彬，等.基于改进YOLOv5的轻量化绝缘子缺陷检测算法[J].浙江电力，2023,42(12):64-72.JI Shichao,QU Xinghe,SONG Qingbin,et al. A lightweight insulator defect detection algorithm based on the improved YOLOv5[J].Zhejiang Electric Power,2023,42(12):64-72.
• [8] FAN W,LIAO Y. Wide area measurements based fault detection and location method for transmission lines[J].Protection and Control of Modern Power Systems,2019,4.
• [9]董亚松，侯立群，可变形特征融合网络的设计及在复杂天气电力设备图像处理中的应用[J].电力科学与工程，2024,40(7):1-9.DONG Yasong,HOU Liqun,Design of deformable feature fusion network and its application in image processing of complex weather power equipment[J],Electric Power Science and Engineering,2024,40(7):1-9
• [10]刘健.基于YOLOX的输电线路异物检测算法研究及软件设计[D].徐州：中国矿业大学，2022.LIU Jian.Research and software design of foreign body detection algorithm for transmission lines based on YOLOX[D]. Xuzhou:China University of Mining and Technology,2022.
• [11]徐丹，余南南，张嘉睿，等.基于层级化注意力融合的航拍图像电力线分割模型[J].南方电网技术，2024,18(4):162-170.XU Dan,YU Nanan,ZHANG Jiarui,et al.Power line segmentation model of aerial images based on hierarchical attention fusion[J],Southern Power System Technology,2024,18(4):162-170.
• [12]金波，陈铈，赵青尧，等.基于改进卷积神经网络的输电线路异常检测研究[J].高电压技术，2023,49（增刊1）:68-71.JIN Bo,CHEN Shi,ZHAO Qinyao,et al. Research on transmission line anomaly detection based on improved convolutional neural network[J].High Voltage Engineering,2023,49(S1):68-71.
• [13]王建，吴昊，张博，等.不平衡样本下基于迁移学习-AlexNet的输电线路故障辨识方法[J].电力系统自动化，2022,46(22):182-191.WANG Jian,WU Hao,ZHANG Bo,et al.Fault identification method for transmission line based on transfer learning-AlexNet with imbalanced samples[J]. Automation of Electric Power Systems,2022,46(22):182-191.
• [14]王逸舟.面向目标检测的视觉预训练方法及其在电力场景的应用[D].杭州：浙江大学，2023.WANG Yizhou.Visual pre-training method for target detection and its application in power scene[D].Hangzhou:Zhejiang University,2023.
• [15] GUO N,JIANG M Y,GAO L J,et al.SimAM:a simple,parameter-free attention module for convolutional neural networks[C]//International Conference on Machine Learning.PMLR,2021.
• [16] AUBRY M,RUSSELL B C,SIVIC J. Painting-to-3D model alignment via discriminative visual elements[J].ACM Transactions on Graphics,2014,33(2):1-14.
• [17] WANG Q L,WU B G,ZHU P F,et al.ECA-net:efficient channel attention for deep convolutional neural networks[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).IEEE,2020.
• [18] LIN M,CHEN Q,YAN S.Network In Network[J].arXiv preprint arXiv:1312.4400,2013.
• [19] CHEN L C,PAPANDREOU G,KOKKINOS I,et al.DeepLab:semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2018,40(4):834-848.
• [20] ZHU J Y,PARK T,ISOLA P,et al.Unpaired image-toimage translation using cycle-consistent adversarial networks[C]//2017 IEEE International Conference on Computer Vision(ICCV).October 22-29,2017,Venice,Italy.IEEE,2017:2242-2251.
• [21] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM,2017,60(6):84-90.
• [22] BOCHKOVSKIY A,WANG C Y,LIAO H Y M.YOLOv4:optimal speed and accuracy of object detection[EB/OL]. 2020:2004.10934. https：//arxiv. org/abs/2004.10934v1
• [23] LI C C,DU D W,ZHANG L B,et al. Spatial attention pyramid network for unsupervised domain adaptation[M]//Lecture Notes in Computer Science. Cham:Springer International Publishing,2020:481-497.
• [24]刘春娟，乔泽，闫浩文，等.基于双路径监督的遥感图像语义分割网络[J].北京航空航天大学学报，2025,51(3):732-741.LIU Chunjuan,QIAO Ze,YAN Haowen,et al.Semantic segmentation network of remote sensing images based on dual path supervision[J].Journal of Beijing University of Aeronautics and Astronautics,2025,51(3):732-741.