A Layered Point Cloud Segmentation Method Based on dbscan

Abstract

The invention relates to a layered point cloud segmentation method based on DBSCAN. Firstly, CSF is used to separate the ground points and non-ground points; the non-ground point segmentation process, first layer the point cloud vertically according to a certain height, and then perform DBSCAN clustering on the projection points of each layer on the XOY plane to obtain each Then project all clustered center points to the XOY plane, use DBSCAN to cluster each object subject, and then judge whether the subject point exists for each subject and each layer, and judge each The number of objects contained in the cluster, and finally perform segmentation processing for clusters with multiple objects. The method of the present invention is aimed at the segmentation of side-view point cloud data, which can ensure the extraction of most subjects in the scene, and has certain robustness, especially in scenes dominated by trees. It has certain significance for point cloud classification and point cloud three-dimensional reconstruction after point cloud segmentation.

Description

A DBSCAN-based Hierarchical Point Cloud Segmentation Method technical field The present invention relates to LiDAR point cloud data information extraction technical field, be specifically related to a kind of layering based on DBSCAN Point cloud segmentation method. Background technique The traditional three-dimensional laser scanning technology is another new breakthrough after the surveying and mapping technology relays the GPS system. The principle of optical ranging, which can quickly, accurately and continuously obtain the three-dimensional coordinates and reflection intensity of a large number of dense points on the surface of the object. Information, currently widely used in forest ecology, urban change detection, urban road detection and planning, and robot environment perception and other fields. However, due to the uneven distribution of point cloud data, there is no semantic information, and even most point cloud data do not contain Color information ha...

AI Technical Summary

Problems solved by technology

[0015] 1) Two parameters, Eps (neighborhood radius) and MinPts (minimum number of points in the neighborhood), need to be selected in advance. However, during segmentation, since the density of different regions may be different, fixed Eps and MinPts will lead to both over-segmentation and under-segmentation ;

[0016] 2) Due to the lack of data in the point cloud data itself, the same object may be clustered into multiple objects;

[0017] 3) DBSCAN is prone to under-segmentation in intersecting scenes

Images