Aligning Algorithm of 3D Point Cloud Model Based on Dimensionality Reduction

Abstract

With the rapid improvement of three-dimensional scanner hardware technology, the accuracy of the point cloud is getting higher and higher, so the number of point-clouds is increasing shar ply, which greatly affects the speed and performance of point-cloud registration. Based on feature matching and ICP algorithm, a 3D point-cloud model stitching algorithm by using Kinect sensors scanning was proposed. In this algorithm, the three-dimensional point-clouds were projected to image plane to get the two-dimensional matching feature points. By using the hash index table, the two dimensional matching feature points are correctly projected back into the three-dimensional space. Finally, the transformation matrix is obtained by using three-dimensional matching points and decomposition of SVD. The model obtained by using the transformation matrix in different angles can realize automatic and correct splicing. The experimental results show that the proposed algorithm can achieve efficient and accurate stitching models to verify the accuracy and validity of this algorithm.