Research On Point Cloud Processing Method In Geometric Parameter Measurement Of Subway Track

With the rapid development of rail transit in China,the total length of operating lines continues to grow,ranking first in the world,and the subway system is still the mainstream.Whether the geometric state parameters of the track meet the design values is an important factor to ensure the safe operation of the subway.The traditional measurement method of track geometric parameters has the problems of low efficiency,high labor intensity and cumbersome measurement process.The laser scanning technology can effectively obtain the real three-dimensional structure information of the subway track,and has unparalleled advantages in the measurement of geometric parameters,which can realize the rapid and all-weather measurement of the geometric parameters of the subway track.Taking the real three-dimensional point cloud data of the track of Chongqing Metro Line 1 as the research object,this thesis systematically studied the technical process of measuring the curvature change rate,gauge change rate and horizontal irregularity of the subway track.The point cloud at the top of the rail is obtained by reasonable preprocessing and segmentation of the point cloud,and the corresponding measurement points are selected to calculate the geometric parameter values.The specific research work includes the following three aspects:（1）Rail point cloud data acquisition and preprocessing stage.A set of simple subway track point cloud acquisition equipment is designed,and its overall design scheme and key hardware selection are introduced.Aiming at the problem of a large number of redundant data points in the track point cloud,the point cloud simplification method based on intensity value to grayscale image and the point cloud hierarchical simplification method based on minimum distance and angle deviation are selected to simplify it.In order to filter out some noise points in the track point cloud,the conditional filtering method based on Euclidean distance is used to filter the simplified point cloud for the first time.The kd-tree method is used to establish the spatial topological relationship of the point cloud neighborhood.Then it is filtered twice based on the bilateral filtering method.Then,the spatial topological relationship of the point cloud neighborhood is established again,and the final filtering is carried out by using the designed point cloud filtering method based on density analysis and spatial coordinates.Finally,the purpose of removing most of the invalid redundant points and noise points is achieved.（2）The segmentation and extraction stage of rail top point cloud.Based on Kmeans++clustering algorithm,the point cloud of track area is preliminarily extracted,and the number of clusters is determined by elbow method and contour coefficient method.On this basis,a voxel grid-normal angle method is designed to segment the point cloud on both sides of the tunnel wall.Then the improved RANSAC algorithm is used to segment and extract the point cloud of the left and right rails,and the timeliness of the RANSAC algorithm before and after the improvement is compared.Finally,the segmentation and extraction of the rail top point cloud is completed based on the adaptive Euclidean clustering algorithm.（3）The geometric parameter measurement stage of the track.The plane projection and coordinate system rotation of the point cloud at the top of the rail are used to obtain a two-dimensional point for measuring the geometric parameters of the track.The least square method and the designed normal vector-least square method are used to fit the six-order polynomial curves of the left and right rails and the center lines of the left and right rails respectively.The measurement points are selected according to the fixed chord length,and the geometric parameter values at each measurement point are calculated.Finally,the error analysis of the measurement results is carried out and a software for measuring the geometric parameters of the subway track based on the three-dimensional point cloud data is designed based on MATLAB GUI.The average absolute errors of the software for the measurement of subway track curvature,gauge and horizontal irregularity are 6.0683e-5m^-1,1.1594e-3m and 1.9878e-3m,respectively.