Segmentation of the 3D point cloud based on GMM and hierarchical clustering for the wrist arm structure from Railway Catenary System

The catenary wrist arm structure in railway scenes serves as the primary support structure that bears the weight of the catenary wire and suspension device. To address the issue of point cloud segmentation for contact wrist arm structures, a 3D point cloud segmentation approach based on Gaussian Mixture Model (GMM) and hierarchical clustering is proposed. Firstly, the 3D point cloud data of the wrist arm is obtained by LiDAR for data preprocessing, and the outliers are removed by statistical filtering to avoid the interference of noise points. Finally, the filtered point cloud data is fitted by GMM to obtain the cluster center, and the hierarchical clustering is used to segment the cluster center, and finally the sub-clusters of each cluster are obtained. Through the experimental verification of real data and simulated data, the $F_{1}$ value of the method on the simulated data set is above 98%, and the $F_{1}$ value on the real data set is above 96%, which has a high accuracy.