重载铁路货车机械零件与踏面损伤图像检测方法  

Image Detection Method for Machine Element and Tread Damage of Heavy Duty Goods Wagon

作　　者：李朋 郭志远 赵永钢 LI Peng;GUO Zhiyuan;ZHAO Yonggang(Suning Vehicle Maintenance Branch of China Energy Railway Equipment Company Limited,Cangzhou 062350,China;Tianjin Hveic Technology Co.,Ltd.,Tianjin 301799,China)

机构地区：[1]国能铁路装备有限责任公司肃宁车辆维修分公司,河北沧州062350 [2]天津哈威克科技有限公司,天津301799

出　　处：《机械制造与自动化》2024年第4期287-291,共5页Machine Building & Automation

摘　　要：当前重载铁路货车的机械零件与踏面损伤以人工检测为主,受到零件所在区域和角度以及人工主观性的影响,检测准确率有待提升。提出一种重载铁路货车机械零件与踏面损伤图像检测方法。采集重载铁路货车的机械零件与踏面图像,采用同态滤波方法消除图像中存在的噪声,增强图像对比度;通过Canny边缘检测算法对采集的图像展开边缘检测,将机械零件与踏面损伤区域分割出来,获得目标区域;在迁移学习的基础上建立VGG-19网络,将上述目标区域输入VGG-19网络中,获取机械零件与踏面损伤特征,实现机械零件与踏面损伤检测。实验结果表明:所提方法的图像处理效果好,机械零件损伤与踏面损伤检测精度均高于90%,检测时间均少于10 s,具有实用性。The present detection of machine element and tread damage of heavy haul goods wagon is mainly implemented manually,which is affected by the area and angle of the parts and the subjectivity of the manual,and whose detection accuracy needs to be improved.This paper presents an image detection method for damage of machine element and tread of heavy-duty goods wagon.The images of machine element and tread of heavy haul goods wagon are collected,and homomorphic filtering method is applied to eliminate noise in the image and enhance image contrast ratio.With Canny edge detection algorithm,the edge of the collected image is detected,and the damage area of Machine element and tread is segmented to obtain the target area.The VGG-19 network is established on the basis of transfer learning,and the above target areas are input into the VGG-19 network to gain the damage characteristics of machine element and tread,so as to realize the damage detection of machine element and tread.The experimental results show that the image processing effect of the proposed method is favourable,the detection accuracy of mechanical parts damage and tread damage is higher than 90%and the detection time is less than 10 s,which is practical.

关 键 词：迁移学习 重载铁路货车 机械零件与踏面损伤检测 同态滤波 CANNY边缘检测算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]