改进GA-BP算法的棒式绝缘子表面缺陷识别  

Surface defect recognition of bar insulator based on improved GA-BP algorithm

作　　者：顾桂梅[1,2] 陈国翠 GU Guimei;CHEN Guocui(School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China;Gansu Provincial Engineering Research Center for Artificial Intelligence and Graphic&Image Processing,Lanzhou 730070,China)

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070 [2]甘肃省人工智能与图形图像处理工程研究中心,甘肃兰州730070

出　　处：《铁道科学与工程学报》2022年第2期546-553,共8页Journal of Railway Science and Engineering

基　　金：甘肃省科技计划项目(18JR3RA104,20JR10RA216);甘肃省教育厅科技项目(2017D-08)。

摘　　要：针对棒式绝缘子及表面缺陷的精确识别技术问题,提出改进遗传算法(Genetic Algorithm)优化BP神经网络算法(GA-BP)和绝缘子水平/垂直方向像素统计的算法,分别用于绝缘子及其表面缺陷精确识别。采用GA算法对BP网络的初始权重进行优化,并用鲍威尔-比厄法(Powell-Beale)更新GA-BP算法中的权值和阈值,高精度识别绝缘子。利用相机标定,实现绝缘子长度直径比、伞裙面积的世界与像素值间的转换。基于像素统计结果,实现绝缘子表面缺陷的精确识别。仿真结果表明,与GA-BP识别算法和以伞裙距离为自爆判断标准的算法相比,改进后的识别率分别提高7.00%和20.50%,且改进GA-BP算法的处理时间为0.015 s/张。为接触网检测监测装置(4C)零部件的准确识别奠定了基础。By aiming at the problem of accurate identification technology of rod insulators and surface defects,improved genetic algorithm,optimized BP neural network algorithm(GA-BP algorithm),and insulator horizontal/vertical direction pixel statistics algorithm were proposed to identify accurately insulators and surface defects,respectively.First,the initial weight of BP network was optimized by GA algorithm,and the weight and threshold of GA-BP algorithm were updated by the Powell-Beale method,which can identify the insulators with high accuracy.Thenthe conversion between the insulator length-diameter ratio,the world value of the umbrella skirt area,and the pixel value was realized by using the camera calibration.Finally,the accurate identification of insulator surface defects was realized by using the result of pixel statistics.The simulation results show that the recognition rate of the improved algorithm is increased by 7.00%and 20.50%,respectively,as compared with the GA-BP algorithm and the algorithm based on the distance of the umbrella skirt.The processing time of the improved GA-BP algorithm is 0.015 s/piece.This lays a foundation for the components of the catenary detection and monitoring device(4C).

关 键 词：棒式绝缘子表面缺陷识别 改进GA-BP算法 相机标定 图像像素统计 

分 类 号：U225.43[交通运输工程—道路与铁道工程]