结合点特征直方图的点云分类方法

摘要

An approach of point cloud classification based on point feature histogram was proposed. Through building the spatial relations between the center point of the point cloud and its neighborhood points, a mul-tidimensional histogram is formed which describes the geometry attribute in its neighborhood. The point feature histogram is used as the classification feature to divide the point cloud into vegetation, ground, and building by a random forest classifier. Comparing point feature histogram with the classic normal and height difference, the results show that the classification based on point feature histogram has better stability and better accuracy.%针对经典的法向量和邻域高差算子包含的信息量较少,点云分类结果不理想的问题,提出一种基于点特征直方图的点云分类方法.该方法用四参数量化中心点与其邻域点之间的空间关系形成一个用于描述中心点邻域几何属性的多维直方图,并将其作为点云分类的特征;用随机森林法将激光点云分为植被、地面以及建筑物3类,点特征直方图、法向量、邻域高差三者均为几何描述算子,用点特征直方图构建了一个高维信息空间的点的几何特征表达,鲁棒性强.通过与基于法向量和邻域高差的点云分类进行对比实验,验证了点特征直方图在点云分类中保边性强、稳定性好.