Improved Euclidean Distance Clustering Segmentation Method for Rail Point Cloud Based on Reflection Intensity
Data structuring of railway point cloud is a technical precondition for deepening professional computing and a-nalysis.Point cloud segmentation is the basis of point cloud data structuring.The continuity and smoothness of the space position of the rail,as a wheel-rail running surface,directly affect the safety of train operation.Therefore,in the seg-mentation of track structure,it is necessary to segment the rail first.In response to the difficulty in unifying and defining distance thresholds due to the traversal of cloud data of all scenic spots of routes in traditional Euclidean distance cluste-ring,resulting in too many classifications and difficulty to find,or the low degree of automation caused by manually se-lecting initial points and adjusting parameters,an improved Euclidean distance clustering rail point cloud segmentation method was proposed based on reflection intensity.Based on the analysis of the characteristics of the track structure,the cloth simulation filtering algorithm was used to distinguish ground shape points from ground object points,and simplify railway line point clouds into the track structure point clouds.The concept of extraction rate was proposed by integrating the reflection intensity attribute of point clouds to determine the value range of rail high reflection intensity to carry out pre-segmentation of rail top surface point clouds.Furthermore,with the pre-segmentation point on the rail top surface as the initial point,the distance threshold was calculated according to the diagonal length of rail head height and rail head width of rail section.The Euclidean distance clustering was carried out by Kd-Tree to find the points less than the dis-tance threshold of the rail head,so as to realize the segmentation of the convex collection point cloud of the rail head.The multi-section rail point cloud segmentation experiments show that the precision and recall rates are greater than 90％,proving the feasibility and effectiveness of the method.
万方数据
维普
