基于数字图像处理和特征提取的钢轨表面缺陷识别方法  

Rail Defect Identification Using Digital Image Processing and Characteristics Extraction

作　　者：王平[1] 刘泽[1] 王嵬[1] 张广伟[1] 孙秀芳[2] 

机构地区：[1]北京交通大学电子信息工程学院,北京100044 [2]北京联合大学机电学院,北京100020

出　　处：《现代科学仪器》2012年第2期24-28,共5页Modern Scientific Instruments

基　　金：国家自然基金资助项目(61050001);北京市属高等学校人才强教计划资助项目(12210993106)

摘　　要：设计了一种基于图像特征分析的钢轨表面缺陷识别方法,运用快速数字图像处理技术可实现钢轨表面裂缝和剥离掉块缺陷的提取,该识别方法通过对采样图像使用高低帽变换进行图像增强,再进行高斯平滑消除毛刺,最后采用阈值处理及边界抑制来提取出缺陷信息。研究了钢轨伤损评价及信息表达方法,可将缺陷信息以曲线的形式表示出来。为验证所设计方法的有效性,设计了模拟实验装置,用工业线阵CCD摄像机获取钢轨样品运动时检测的图像,进行缺陷识别处理,实验表明,系统输出曲线能够明确反映损伤的程度及其位置信息。A rail surface defect identification method based on image characterize analysis is designed in this paper.The defect extraction of surface cracks and peeling pieces can be realized by using rapid digital image processing technology.During the identification process,top-hat and bottom-hat transform of sampled images is used for image enhancement;then Gaussian smoothing is adopted to eliminate small burrs;finally defect information is extracted through threshold processing and boundary inhibition.Rail defect evaluation and information representation methods are studied to present the defect information in the form of curves.In order to verify the validity of the method designed,an experimental system is developed,which uses industrial linear CCD to capture the images of a rail sample detected in motion and then identifies defects of the sample.Experiments showed that the information of defect level and location can be clearly expressed by the system’s output curves.

关 键 词：钢轨探伤 高帽变换 低帽变换 图像增强 阈值处理 缺陷识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]