Abstract: To address the problems of large data volume,strong noise,and abundant outliers in the raw structured light rail profile point cloud data,this paper proposes a rail point cloud adaptive simplification method that combines Euclidean clustering with various traditional filtering methods. A clustering segmentation method based on the Euclidean distance of point clouds is proposed to identify and simplify invalid scattered point data. A statistical filter combined with uniform voxel down sampling is used for initial noise reduction. Based on this,Segmentation of noise by Euclidean clusters,an adaptive pass-through filter that automatically acquires the filtering range is used to remove rail bottom adhesion data to ensure the efficiency and accuracy of point cloud registration. The proposed method can effectively simplify invalid data and reduce noise,achieving a point cloud simplification rate of about 94%,while retaining the effective profile features of the original point cloud,which lays the foundation for high-precision identification of point cloud registration and wear points.

Key words: point cloud reduction, point cloud filtering, Euclidean distance, segmentation by cluster, adaptive pass-through filtering