一种利用三维HOG特征的建筑物点云分类方法

A Point Cloud Classification Method of Buildings Using 3D HOG Features

针对建筑物点云精确分类问题,提出一种利用三维HOG(histogram of oriented gridients)特征的建筑物点云分类方法。首先,通过分析点云的空间分布特征(投影形状、高程差异、密集程度等),生成能够显著表达立面目标多尺度特征的三维格网;然后,引入方向梯度直方图算子,依据建筑物立面的邻域梯度对称性,以及多尺度、多方向的立面特征差异,分析格网三维HOG特征,准确提取初始立面格网;最后,通过方向和距离约束获取完整立面格网,进一步将各立面聚类实现建筑物点云分类。实验结果表明,该方法基于立面特征进行建筑物分类,对于地基平台采集的点云具有较好的分类效果,稳健性较强。

Aiming at the problem of accurate classification of building point cloud,this paper proposes a method of buildings point cloud classification using 3D HOG features.This method analyzes the spatial distribution characteristics of the point cloud(projection shape,elevation difference,density,etc.)to generate a 3D grid that can significantly express the multi-scale features of the elevation target.Then,the method introduces the histogram of oriented gridients,HOG,based on the neighborhood gradient symmetry,the multi-scale and multi-directional feature differences of the building facade,and accurately extracts the initial facade grid.Finally,the method gets the complete facade grid through the direction and distance constraints,further clusters each facade to realize the point cloud classification of buildings.The experimental results show that the proposed method has a better classification effect and robustness for the point cloud collected by the terrestrial platform.