Automatically Identifying Planar and Spherical Targets from MLS Point Clouds

To address the challenges of automatic planar and spherical target extraction in high-speed rail scenarios, this article proposes a novel hierarchical target detection and automatic localization method. First, a combination of point cloud filtering and height constraint methods is used to segment candidate target points and eliminate interference from ground points and power line infrastructure points. Next, a Triangular Irregular Network (TIN) is constructed using the candidate target point cloud, and candidate target points can be quickly segmented into different regions using the TIN region growing method based on edge constraints. Third, a pattern recognition method based on tripod characteristics is designed to automatically identify candidate target regions. Finally, target types are distinguished based on the surface roughness differences of the candidate target point cloud, and the center position of the target point cloud is determined using a Rotating Minimum Enclosing Circle (RMEC) method. The validation conducted on point clouds in high-speed railway scenarios demonstrates that this method can swiftly accomplish automatic detection and positioning of planar and spherical targets within point clouds comprising tens of millions of points. When compared to manually calibrated positions, the positioning error is confined to within the millimeter range.