基于车载LiDAR点云的地物分类方法的研究

基于点云分类常用的近邻聚类法和物体表面分割等方法,本文提出了一种基于最大网格密度的近邻聚类的方法。该方法首先对原始点云进行低点提取,设置格网的大小,在此基础上对点云数据进行去噪并进行主成分分析,再对点云空间进行均匀格网化,使具有最大密度的格网为聚类中心,加入高程、强度以及法向量等特征对分割后的点云实现了不同地物的分类,提高了运算效率,降低了错分率。

Research on Classification Method of Vehicle LiDAR Point Cloud

Combining the nearest neighbor clustering algorithm and object surface segmentation method,this paper proposes a new method which combines the nearest neighbor clustering algorithm based on the maximum grid density.Firstly,we should have low point extraction of the original point cloud,set the grid size,based on the point cloud data of denoising and the principle components analysis,then uniform grid point cloud space,maximum density of grid clustering center is introduced,add height,strength and the method of vector feature for segmentation of point cloud to achieve the classification of different features,and improve the efficiency of operations and reduces the error rate.