Simplification of scattered point cloud based on feature extraction

Simplification of scattered point cloud is one of the key preprocessing technologies in reverse engineering. Most simplification algorithms always lose geometric feature excessively in the process. On the basis of feature extraction, a new algorithm is proposed for the simplification of scattered point cloud with unit normal vectors. First, points in point cloud are distributed into uniform cubes. Next, bounding spheres are constructed with their centers at each point; accordingly K-nearest neighbors are searched in the relevant sphere. Later, a specified function is defined to measure the curvature of each point so that feature points can be extracted. Finally, feature points and non-feature points are simplified according to the radius of bounding sphere and the threshold of normal vectors´ inner product. The experiments show that the proposed algorithm has the advantages of fast speed and high reservation of the geometric feature of point cloud.