Normal estimation algorithm for point cloud using KD-Tree

This paper proposes the normal vector estimation algorithm based on KD-Tree. The speed of searching neighbor field is improved by utilizing KD-Tree data structure. The orientation of normal vectors computed by PCA is ambiguous and confused. In order to solve this problem, the viewpoint of point cloud is set to check and flip over the orientation. For the purpose of obtaining the correct normal vector estimation, the scope of neighbor field is extended to improve the antinoise ability of normal vector estimation. The experiment result proves that the normal vector estimation algorithm is robust.