Point Cloud Grid Reduction Method Based on Feature Parameters

Abstract

A large amount of the 3D point cloud data obtained by 3D laser scanning technology are redundant, which is not conducive to computer storage and calculation and reduces the reconstruction efficiency. Considering the low accuracy of several point cloud reduction method and the lack of details, proposing a point cloud grid reduction method based on feature parameters. Firstly, the point cloud topological relationship is established by calculating the point cloud normal vector and curvature according to the KD-tree, and the product of multiple parameters which are used as the feature parameters; then, the voxel grid algorithm is used to divide the point cloud data and distinguish them according to the average curvature in each grid feature area and non-feature area; finally, different methods are used to simplify the point cloud in different areas. Through the comparative analysis of experiments, this paper shows that the method can improve the reconstruction efficiency, effectively avoid holes, and better retain the detailed features of the point cloud.