Data reduction based on variable radius sphere

There is an increasing interest for data reduction of point cloud in reverse engineering. It is the base of model reconstruction and registration. After point cloud data is reduced, the speed of reconstruction and registration can be improved. A data reduction algorithm is proposed to preserve the shape of original point cloud. Firstly, normal vector of every point is calculated and outliers are discarded by normal vectors' relation of neighbor points. Then curvedness of each point is calculated. Finally, point cloud is simplified and dominant points are detected based on moving variable spheres. Experiments demonstrate the algorithm can reduce points of point cloud effectively when the geometrical shape of point cloud is preserved.