基于Yolo v5与Grabcut的架空线路绝缘子可见光图像自动识别与分割方法  

Visible Light Image Automatic Recognition and Segmentation Method for Overhead Power Line Insulators Based on Yolo v5 and Grabcut

作　　者：何锦强 李锐海 李昊 廖永力 龚博 郝艳捧 梁苇 吴建蓉 文屹 HE Jinqiang;LI Ruihai;LI Hao;LIAO Yongli;GONG Bo;HAO Yanpeng;LIANG Wei;WU Jianrong;WEN Yi(Electric Power Research Institute,CSG,Guangzhou 510663,China;Electrical Power Research Institute of Guizhou Power Grid Co.,Ltd.,Guiyang 550000,China;CSG Laboratory of Ice Prevention and Disaster Reduction,Guiyang 550002,China)

机构地区：[1]南方电网科学研究院,广州510663 [2]贵州电网有限责任公司电力科学研究院,贵阳550000 [3]中国南方电网有限责任公司防冰减灾重点实验室,贵阳550002

出　　处：《南方电网技术》2023年第6期128-135,共8页Southern Power System Technology

基　　金：中国南方电网有限责任公司科技项目“中国南方地区电网自然覆冰大数据分析与人工智能应用”(066600KK52190063);国家重点研发计划项目(2021YEE0204200)。

摘　　要：准确识别与分割绝缘子图像是架空线路绝缘子状态感知与缺陷诊断的重要前提。针对绝缘子的可见光图像提出了结合深度学习Yolo(You only look once)v5模型与图论分割Grabcut算法的两阶段图像识别与分割方法。首先采集图像建立数据集并训练Yolo v5,实现绝缘子的识别,然后用识别框坐标确定感兴趣区域,预判前景和背景,基于Grabcut算法实现绝缘子的自适应分割。根据该方法识别与分割架空线路绝缘子的在线监测可见光图像,结果表明,该方法无需分割标注或人工交互,可以高效准确地定位复杂背景下绝缘子的边缘并分割绝缘子,极大地提高绝缘子图像的分析效率。Accurate recognition and segmentation of insulator image is an important prerequisite for state perception and defect diagnosis of overhead power line insulators.In this paper,a two-stage image recognition and segmentation method combining the Yolo(you only look once)v5 and Grabcut is proposed for the visible image of insulators.This paper firstly collects images,establishes data set and trains Yolo v5 to realize insulator recognition.Then the recognition frame coordinates are used to determine the region of interest,predict the foreground and background,and realize the adaptive segmentation of insulators based on the Grabcut method.This method is used to identify and segment the on-line monitoring visible light image of overhead power line insulators.The results show that this method can accurately locate the edge of insulator under complex background and segment insulator without segmentation and annotation or manual interaction,which can efficiently improve the efficiency of insulator image analysis.

关 键 词：绝缘子 识别 分割 可见光图像 自然背景 架空线路 

分 类 号：TM766[电气工程—电力系统及自动化] TP391.41[自动化与计算机技术—计算机应用技术]