基于法向量密度聚类的LiDAR点云屋顶面提取

针对现有算法从LiDAR点云中提取复杂建筑物屋顶面不完整、阈值难以设置的问题,提出一种结合点云空间分布的法向量密度聚类提取屋顶面点云方法。通过构建Delaunay三角网,计算建筑物LiDAR点云的法向量;在分析建筑物点云空间和法向量分布特点的基础上,定义一种邻域关系度量屋顶面点云之间的相似性,并利用提出的算法聚类建筑物点云,得到屋顶面片点云粗提取结果;通过构建屋顶面片缓冲区,经面片处理得到建筑物各屋顶面的完整点云。选取不同复杂程度的建筑物进行实验,结果表明,算法能有效提取复杂建筑物屋顶面点云,具有较好的适应性,并能为建筑物三维重建提供可靠的屋顶面信息。

Roof Extraction Using LiDAR Point Clouds Based on Normal Vector Density-based Clustering

Aiming at the problems of incomplete and thresholds setting difficulty in complex building roof extraction from LiDAR point clouds of the existing algorithms, a normal vector density-based clustering method combined with spatial distribution of point clouds is proposed to extract building roof. Normals of building point clouds are calcu-lated based on delaunay triangulation. A neighbour relationship is defined to measure similarity between roof point clouds based on analysing features of building point clouds' spatial and normal vector distribution, and rough ex-traction results are obtained by using the proposed method to cluster building point clouds. All point clouds of each roof are acquired completely through buffer zones construction and roof surface process. Buildings with different complexity are used, and experimental results show that the proposed method can extract complex building roofs ef-fectively with preferable adaptability.