基于视觉的输电线路金具锈蚀缺陷检测方法研究进展  

Research progress of vision-based rust defect detection methods for metal fittings in transmission lines

作　　者：刘传洋 吴一全[1] 刘景景[2] Liu Chuanyang;Wu Yiquan;Liu Jingjing(College of Electronic and Information Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China;College of Mechanical and Electrical Engineering,Chizhou University,Chizhou 247000,China)

机构地区：[1]南京航空航天大学电子信息工程学院,南京211106 [2]池州学院机电工程学院,池州247000

出　　处：《仪器仪表学报》2024年第3期286-305,共20页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(61573183);高校自然科学研究(2023AH052358,CZ2022ZRZ07)项目资助。

摘　　要：输电线路金具的表面锈蚀作为常见的缺陷类型,是危害输电线路安全运行的重要隐患之一,如何快速、准确地发现锈蚀的金具设备并进行修复是线路巡检运维工作亟待解决的问题。本文综述了近十年来基于视觉的输电线路金具锈蚀缺陷检测方法的研究进展。首先简述了基于传统图像处理的金具锈蚀缺陷检测流程;然后按照基于传统图像处理、深度学习方法概述了金具设备锈蚀缺陷检测,重点阐述了基于深度卷积神经网络的目标检测和语义分割算法在输电线路金具锈蚀缺陷检测中的应用;随后介绍了基于深度学习的金具锈蚀缺陷检测自建数据集以及性能评价指标;最后指出了基于深度学习的输电线路金具锈蚀缺陷检测方法目前存在的问题,并对未来研究工作进行了展望。As a common defect type,surface rust of metal fittings in transmission lines is one of the important hidden dangers endangering the safe operation of transmission lines.How to quickly and accurately discover and repair rusted metal fittings is an urgent problem to be solved in the work of transmission line inspection.This article reviews the research progress of vision-based rust defect detection methods for metal fittings in the last ten years.Firstly,the rust defect detection process of metal fittings based on traditional image processing is introduced.Then,the rust defect detection of metal fittings is summarized according to traditional image processing and deep learning methods.The application of object detection and semantic segmentation algorithms in rust defect detection of metal fittings is emphasized.Next,the self-built data sets for metal fittings′rust defect detection and performance evaluation indexes are introduced.Finally,the existing problems of rust defect detection methods based on deep learning are pointed out and future research work is prospected.

关 键 词：金具 锈蚀缺陷 图像处理 深度学习 目标检测 语义分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TM930.1[自动化与计算机技术—计算机科学与技术] TH89[电气工程—电力电子与电力传动]