基于点云深度映射颜色的导轨表面损伤识别

Damage Identification of Guide Surface Based on Point Cloud Data Depth Mapping Color

针对高速滑动电接触导轨,研究了表面微小损伤快速检测及分类识别方法。基于激光扫描原理,构建了三维测量系统,用于获取导轨表面形貌的三维点云信息,并给出了一种基于点云深度映射颜色的方法,用于导轨表面微小损伤的检测。将三维点云数据经过去噪、滤波平滑、数据精简等预处理之后,根据所设定的深度基准平面,构建点云深度映射颜色模型,将点云深度信息映射为红绿蓝(RGB)信息,采用一维最大熵法设定最优颜色阈值,实现损伤区域的准确提取;采用二叉树模式识别方法,建立损伤分类模型,实现导轨表面微小损伤的识别与分类。结果表明,损失质量小于1 g的微小损伤检出率达98%以上、微小质量损失检测精度可达毫克级;凹坑与划痕两大类损伤识别率达80%以上。

Aiming at the high speed sliding electrical contact rail, the surface micro damage detection and identification method are studied. Based on the principle of laser scanning, a three-dimensional measurement system is constructed for acquisition of point cloud information for rail surface, meanwhile a method of depth mapping color based on point cloud is proposed for the detection of rail surface micro damage. After the threedimensional point cloud data being denoised, smoothed and reduced, according to the datum plane set, point cloud depth mapping color model is constructed, the point cloud depth information can be mapped into red, green, blue（RGB） information, and the optimum color threshold is set by using the one-dimensional maximum entropy method to realize the accurate extraction of the damage area. Binary tree pattern recognition method is used to establish damage classification model and realize the identification and classification of rail surface micro damage. The results show that detection rate of the micro damage of which mass loss is less than 1 gram is more than ninety-eight percent, detection accuracy of minor mass loss is milligram and the damage identification rate of pits and scratches can reach above eighty percent.