A fast and feature-retained simplification method for point cloud

Abstract

Abstract In order to solve the problem that the speed of feature-based point cloud simplification methods is slow, a fast and feature-retained point cloud simplification method based on voxel convex hull is proposed in this paper, which avoids the calculation of a large number of features and has good simplification effect. In this method, point cloud is quickly voxelized, and then the voxels are divided into two types according to the number of points in each voxel: less-point voxel and more-point voxel. Less-point voxels are simplified directly using uniform sampling method. More-point voxels are simplified by using the voxel convex hull method. The convex hull volume and directed projection area of points within each more-point voxel are calculated as voxels’ feature values, and then the point cloud is divided into groups according to the feature values and points in different groups are simplified separately. Finally, the simplified point cloud is obtained by fusing the simplification results of less-point voxels and more-point voxels. The reduction ratio of point cloud is controllable and the parameters are easy to be set in this method. Experiments show that the simplification time of this method is more than 50% lower than that of the simplification method based on features and clustering. The retention of point cloud features of this method is obviously higher than that of traditional methods such as uniform sampling method, and the average curvature is improved by about 10%. Surface reconstruction experiment shows this method does well in sharp features retaining. Moreover, the reduction ratio of this method is controllable, and the parameters are flexible and easy to be set.