基于车载点云空间分布的城区地物分类

Classification of Urban Ground Objects Based on Spatial Distribution of Point Clouds

城区车载点云数据包含建筑物、树木等不同的地物反射的点的总数,如何对无拓扑、盲目性的激光脚点进行分类成为点云数据后处理的难点。提出一种基于点云分散程度的分类方法:首先构建包围点及其k-邻域点集的最小包围盒,比较其在3个坐标平面的最大投影面积与最小包围盒体积的比值提取面状走势信息;其次比较点集在局部空间的分散程度提取树木点;最后以高程、点云密度作为约束条件进行分类。实例证明,该算法可较好地识别建筑物、树木等地物,具有一定的实用价值。

There is a large amount of reflected points of different ground objects in urban laser scanning,including trees and buildings.It is difficult to separate the points into accurate kinds respectively from these non-related and blind laser points in data post-processing. A newmethod of classification based on dispersion degree is proposed in this paper.Firstly,a minimum bounding box including one point and its k-neighborhood points is constructed. Then by comparing the ratio of the maximum area projected to 3 coordinate planes and the volume of the box,the panel features can be extracted.Secondly,tree points can be extracted though comparing the dispersion of points in the local space.Finally,the elevation and density are used as constrains to the complement of the classification. As the experiment proved,this method has some practical value in classifying ground,building and tree points effectively.