Point Cloud Reduction Method Based on Curvature Grading and Voxel Filtering

Abstract

In the context of automated assembly engineering based on reverse engineering in the aerospace field, redundant data will make it impossible to complete 3D model reconstruction efficiently and quickly, and point cloud data must be streamlined. A single simplification method cannot accurately retain the feature information of the scanned point cloud. Aiming at this background, a point cloud simplification method combining curvature grading and octree voxel filtering is introduced. The obtained curvature is divided into different levels by using the properties of logarithmic function, and the strong and weak feature points are distinguished. The weak feature points are filtered by octree voxel, and the filtered results are merged with the strong feature points to obtain the final streamlined results. This method is compared with the random sampling method and the point cloud simplification method combining 3D-SIFT feature extraction and voxel filtering. The experimental results show that this method not only greatly reduces the amount of data, but also preserves the local details of the original point cloud data, so as to achieve the purpose of efficiently compressing point cloud data.