Automated Extraction of Rail Point Clouds by Multi-Scale Dimensional Features From MLS Data

Abstract

Existing railway measurement methods are very difficult to adapt to high-density operating conditions in terms of safety and efficiency. A mobile laser scanning (MLS) system using LiDAR begins to be applied in railway measurement in complex scenarios and helps to detect the condition of rails. This paper presents a novel method to extract rails from MLS point cloud data. After the point cloud preprocessing, a spatial distance threshold is first set to obtain the point cloud of the ballast bed part. Then, the multi-scale features of the point cloud dimension are defined, and the dimensional features of each point cloud at different scales are calculated as feature inputs using principal component analysis. Finally, a classifier is constructed using linear discriminant analysis to extract the rail point clouds. The results of several scenarios in existing lines show that our method is highly adaptable and the extracted rail completeness and accuracy are above 94%, which is very promising for automatic extraction of rail point clouds from the point cloud data collected by MLS. The gauges measured on the extracted rail point clouds also validate the method as a promising solution for extracting 3D rails from MLS point clouds.