摘要
铁路运输是提升国民经济发展的重要支柱,轨道作为铁路运输的重要组成部分,需要更先进的检测方法来检测轨道缺陷。基于此,文章设计了一种基于计算机视觉技术的轨道缺陷检测系统,整个系统将机器视觉作为理论基础,系统对待测轨道图像进行图像滤波、图像分割和图像边缘检测等预处理,并提取图像区域的周长、面积、长宽比、矩形度、复杂度等特征参数,最终输入BP神经网络学习,最终获得缺陷识别结果。通过现场实验验证,该系统能够较好地识别表面断裂、局部凹陷、表面剥离、表面裂纹这四种缺陷,平均分类精度高达86.3%,相比传统人工巡查的方式,具有无法比拟的优势。
Railway transportation is an important pillar to promote the development of national economy.As an im-portant part of railway transportation,track needs more advanced detection methods to detect its defects.Based on this,this paper designs a track defect detection system based on computer vision technology.The whole system takes ma-chine vision as the theoretical basis.The system preproc-esses the track image to be measured,such as image filter,image segmentation,image edge detection,etc.,extracts the characteristic parameters of the image area,such as perim-eter,area,ratio of length to width,rectangularity,complex-ity,etc.,and finally inputs them into BP neural network for learning,and obtains the defect recognition results.Field experiments show that the system can better identify four kinds of defects of surface fracture,local depression,surface peeling and surface crack with an average classification accuracy of 86.3%,which has incomparable advantages compared with traditional manual detection.
作者
王勇
贺锐
蔡志贤
WANG Yong;HE Rui;CAI Zhixian(Guangdong Guangfo Rail Transit Co.,Ltd.,Guangzhou,Guangdong 510380;Guangzhou Metro Group Co.,Ltd.,Guangzhou,Guangdong 510380;Guangzhou CRRC Rail Transit Equipment Co.,Ltd.,Guangzhou,Guangdong 510380)
出处
《工程技术研究》
2022年第23期13-16,共4页
Engineering and Technological Research
关键词
铁路运输
轨道检测
深度学习
机器视觉
railway transportation
track detection
deep learning
machine vision
