基于同步标记算法的钢轨表面缺陷检测  

Rail surface defect detection based on synchronized labeling algorithm

作　　者：陈华伟[1] 吴禄慎[1] 袁小翠[1] 

机构地区：[1]南昌大学机电工程学院,江西南昌330031

出　　处：《计算机工程与设计》2016年第1期127-131,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(51365037);江西省教育厅基金项目(GJJ14128)

摘　　要：针对区域标记算法流程长、中间数据结构多、算法稳定性依赖于阈值化处理等问题,以钢轨图像为研究对象,提出一种像素标记同步搜索算法。展开数据结构和程序流程设计,着重解析像素标记过程中等价关系的识别、记录、拼接、排序和连通区域识别输出全过程。该算法在面向钢轨表面缺陷检测的轨缝和裂纹定位及参数提取中得到了验证,其在非实时要求的图像处理中具有较高稳定性。To solve problems of connected components labeling（CCL）algorithm that it has relatively long procedure,needs of appropriate intermediate data structure,and dependence on thresholding for its stability,one synchronous searching pixel labeling algorithm was advanced for rail images.Data structure and programming flow including recording,jointing,sorting of equivalent relation and output of connected components were detailed.The algorithm is validated in the location of joint gap and surface crack and through its parameterization extraction during the rail surface defect detection.High algorithm stability can be achieved in non-real-time environment using this method.

关 键 词：缺陷检测 阈值化 连通区域标记 邻域搜索 等价标记 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]