基于图像识别的输电线路缺陷检测方法  

Transmission Line Defect Detection Method Based on Image Recognition

作　　者：李旭旭[1] 马小敏 唐军 孔庆宇 于军亮 LI Xuxu;MA Xiaomin;TANG Jun;KONG Qingyu;YU Junliang(State Grid Sichuan Electric Power Company,Chengdu 610095,Sichuan Province,China;Electric Power Research Institute,State Grid Sichuan Electric Power Company,Chengdu 610095,Sichuan Province,China;Marketing Service Center,State Grid Sichuan Electric Power Company,Chengdu 610042,Sichuan Province,China;State Grid Smart Grid Research Institute Co.,Ltd.,Changping District,Beijing 102211,China)

机构地区：[1]国网四川省电力公司,四川省成都市610095 [2]国网四川省电力公司电力科学研究院,四川省成都市610095 [3]国网四川省电力公司营销服务中心,四川省成都市610042 [4]国网智能电网研究院有限公司,北京市昌平区102211

出　　处：《电力信息与通信技术》2023年第6期31-36,共6页Electric Power Information and Communication Technology

摘　　要：针对现有输电线路检测存在的缺陷类别少、尚无统一检测参数等问题,文章提出一种基于深度学习的输电线路缺陷检测方法。首先,建立基于Wire_10的传输线路缺陷识别数据集;为进一步降低因背景和照明因素对识别准确度的影响,定义相关数据集,提出一种基于R-CNN的端到端高识别精度的深度学习算法,用于建立具有转移学习和微调的检测模型。最后,实验结果表明,该检测方法能够准确识别Wire_10数据集中的缺陷类别,并对具有复杂背景和不同照明的航空图像具有鲁棒性。To solve the problems of few detection categories and lack of unified comprehensive evaluation index in the existing transmission line detection,this paper proposes a deep learning-based transmission line defect detection and evaluation method.Defects are treated as a category by building a Wire_10 based transmission line dataset.In order to further reduce the target detection of images due to the attachment of nests and foreign objects in the line,which is easily affected by the background and lighting,these two factors are used as variables to define the background dataset and lighting dataset,and a new method based on R-CNN end-to-end high recognition accuracy deep learning algorithm is proposed for building detection models with transfer learning and fine-tuning.The results show that the detection method can accurately identify defect categories in the Wire_10 dataset and is robust to aerial images with complex backgrounds and different illuminations.

关 键 词：输电线路 图象识别 缺陷检测 深度学习 

分 类 号：TN915.853[电子电信—通信与信息系统]