基于深度学习的输电线路视觉检测研究综述  

Research Overview on Visual Detection of Transmission Lines Based on Deep Learning

作　　者：赵振兵[1] 齐鸿雨 聂礼强 ZHAO Zhenbing;QI Hongyu;NIE Liqiang(School of Electrical and Electronic Engineering,North China Electric Power University,Baoding,Hebei 071003,China;School of Computer Science and Technology,Shandong University,Qingdao,Shandong 266237,China)

机构地区：[1]华北电力大学电气与电子工程学院,河北保定071003 [2]山东大学计算机科学与技术学院,山东青岛266237

出　　处：《广东电力》2019年第9期11-23,共13页Guangdong Electric Power

基　　金：国家自然科学基金项目(61871182);北京市自然科学基金项目(4192055);河北省自然科学基金项目(F2017502016);中央高校基本科研业务费专项资金项目(2018MS095、2018MS094);模式识别国家重点实验室开放课题基金项目(201900051)

摘　　要：保障输电线路的可靠性是能源互联网和智能电网建设的重要内容,基于深度学习的智能化输电线路视觉巡检技术具有安全、高效、便捷等特点,对保障输电网的稳定运行有重要意义。为此,首先总结国内外深度学习视觉检测方法以及输电线路视觉检测方法研究现状;其次,描述无人机巡检、在线监测、激光雷达巡检、高分辨率光学卫星巡检等4种输电线路巡检方式,分析不同方式的差异与优劣,同时讨论了深度学习在4种方式中应用的关键问题;最后,探讨了深度学习在输电线路视觉检测中应用的未来发展方向。It is an important part of construction of energy internet and smart grid guaranteeing reliability of transmission lines.The visual inspection technology for intelligent transmission lines based on deep learning is safe,high efficient,convenient and fast,which is of great significance to ensure stable operation of the power transmission networks.This paper firstly summarizes research status of visual detection methods based on deep learning and visual inspection methods for transmission lines at home and abroad.Secondly it describes four inspection ways for transmission lines including UAV inspection,online monitoring,radar inspection and high resolution optical satellite inspection.It also analyzes differences,advantages and disadvantages of different ways and discusses key problems of application of deep learning in these four ways.Finally,it discusses future development direction of deep learning in visual detection of transmission lines.

关 键 词：输电线路 视觉检测 深度学习 深度卷积神经网络 智能巡检 

分 类 号：TM726.3[电气工程—电力系统及自动化] TP181[自动化与计算机技术—控制理论与控制工程]