基于深度学习和无人机图像的架空线路缺陷巡检综述

架空输电线路巡检是电网运维工作的一项重要内容，运维人员利用无人机进行线路巡视检测已成为电力巡检工作中的重要手段。本研究首先概述了无人机巡检任务中人机协同作业系统以及无人机智能自主作业系统的架构；其次，分析了当前架空输电线路缺陷巡检领域数据集状况以及数据扩增技术；然后，综述了基于深度学习的无人机图像缺陷检测典型方法以及评价指标，并对比总结了各种方法的优缺点；随后，讨论了无人机图像视觉检测方法中图像采集规范、数据集形式、缺陷检测算法专业化应用等对架空线路缺陷检测效果，指出了图像检测指标和类别定义在电力巡检专业化领域中的不足；最后，探讨了基于深度学习的无人机图像缺陷巡检任务的未来发展方向。

Review of overhead transmission line defect inspection based on deep learning and UAV images

The overhead transmission line inspection is an important part of power grid operation and maintenance work, and it has become an important means for the operation and maintenance personnel to use UAV for line inspec-tion and inspection. In this study, the architecture of human-machine collaborative operation system and UAV autonomous operation system in UAV inspection mission were firstly. Secondly, the current status of data set in the field of defect inspection of overhead transmission lines and data amplification technology are analyzed. Then, the typical methods of UAV image defect detection based on deep learning were reviewed, and the ad-vantages and disadvantages of various methods were also compared. Then, the effects of image acquisition specifications, dataset formats, and the application of defect detection algorithms on overhead transmission line defects in UAV image were discussed. The shortcomings of defect detection metrics and category definitions in the specialized field of power line inspection were also pointed out. Finally, the future development direction of UAV image defect inspection task based on deep learning was discussed.