Surface reconstruction from LiDAR point cloud data with a surface growing algorithm

LiDAR generates datasets of sub-randomly distributed points on scanned surfaces, named point cloud, which contains abundant implicit 3D spatial information. Explicit spatial information of scanned surfaces can be retrieved through a process of surface reconstruction. This paper proposes a novel algorithm for surface reconstruction based on the scheme of surface growing. It starts with a seed point and continuously merges adjacent points which are found as extensions of the surface. In order to handle randomly distributed points, LiDAR data is divided into 3D grid for the algorithm to search adjacent points. The point-set in every 3D grid was applied to estimate the normal vector. And there are two factors to conduct the growing process: the angle between two normal vectors of adjacent patches and the distance of the point to the growing surface. Finally, the planar surface is reconstructed by merging the clustered patches. The experimental datasets include point clouds acquired by both ground-based and airborne LiDARs.