高速铁路近景影像轨道边缘提取与匹配方法  

Method of Track Edge Extraction and Matching for Close-range Images of High-speed Railway

作　　者：陈强[1] 沈鑫[1] 赵晶晶[1] 常云浩 刘丽瑶[1] 杨莹辉[1] 

机构地区：[1]西南交通大学测绘遥感信息系,四川成都610031

出　　处：《铁道学报》2017年第8期122-128,共7页Journal of the China Railway Society

基　　金：国家自然科学基金(41472255)

摘　　要：近景摄影测量应用于铁路轨道几何平顺性检测具有较大潜力,为获得轨道几何状态参数,需对轨道数字影像进行定向与建模,本文探索一种高速铁路近景影像轨道边缘提取方法并构建轨道边缘线的同名点坐标映射模型。根据轨道数字影像的灰度信息与点位分布特点,利用Canny算法和概率Hough变换提取铁路轨道的内边缘特征,采用ORB算法与KNN算法匹配轨道表面同名点,建立轨道表面同名点的坐标对应关系。利用边缘线与轨道表面几何关系,构建轨道边缘线的多项式映射模型以匹配出边缘同名点。试验结果表明,该方法在轨道边缘识别与同名点匹配方面有较高的可靠性,可为高速铁路轨道几何状态检测提供影像处理技术支持。The application of the close-range photogrammetry technology in the geometric state detection for high-speed railway track has great potential. In order to obtain the geometric parameters of the track, the ori-entation and modeling of close-range digital images for railway tracks are required. A n algorithm of edge detec-tion and matching model for high-speed railway tracks from the close-range images was proposed and the m a p -ping model of the homogeneous point coordinate of the track edge was built. Based on the gray information and point distribution in close-range track images, the inner edge feature of railway track was first detected by using the Canny algorithm and probabilistic H o u g h transform. The homogeneous points on the track surface were matched by using ORB and KNN algorithms to establish the corresponding coordinate relations of homogeneous points on the track surface. By using the geometric relationship between the edge line and the track surface, a polynomial mapping model of track edge line was built to match the homogeneous points on track edge line.The result shows that the algorithm, with a high accuracy in the track edge detection and homogeneous point matching, can provide image processing support for geometric state detection of high-speed railway track.

关 键 词：铁路轨道 边缘提取 同名点匹配 映射模型 

分 类 号：P234.1[天文地球—摄影测量与遥感] U213.2[天文地球—测绘科学与技术]