铁轨表面缺陷的视觉检测与识别算法  

Visual Inspection and Classification Algorithm of Rail Surface Defect

作　　者：唐湘娜[1] 王耀南[1] 

机构地区：[1]湖南大学电气与信息工程学院,长沙410082

出　　处：《计算机工程》2013年第3期25-30,共6页Computer Engineering

基　　金：国家自然科学基金资助重点项目(60835004);国家"863"计划基金资助项目(2007AA04Z244)

摘　　要：提出一种铁轨表面缺陷的视觉检测与识别算法。设计铁轨表面缺陷视觉检测与识别系统的总体结构,基于水平投影法提取铁轨表面区域,采用逻辑操作组合检测结果,使用BP神经网络进行缺陷分类。实验结果表明,该算法能准确地检测与识别铁轨表面的疤痕和波纹擦伤这2种缺陷,分类正确率分别达到99%和95%。This paper proposes a visual inspection and classification algorithm of rail surface defect. An overall structure of the visual inspection and classification system for rail surface defects is designed. It extracts the rail surface subarea based on horizontal projection method, combines the inspection result based on logic or operation, and classifies the defect based on BP neural network. Experimental results show that the algorithm is able to detect and identif5＇ the two rail surface defect, scarring and corrugated abrasion accurately, the accuracy of classification is 99% and 95%.

关 键 词：铁轨 表面缺陷检测 水平投影法 灰度补偿 机器视觉 BP神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]