Registration method based on point cloud feature point extraction

The invention relates to a registration method based on point cloud feature point extraction, and belongs to the technical field of three-dimensional registration. The method comprises the following steps: firstly, performing down-sampling processing on a source point cloud and a target point cloud by using improved voxel filtering; secondly, extracting feature points from the down-sampled point cloud by using an internal shape signature feature point inspection algorithm to obtain a point cloud containing rich geometric feature information; then describing the extracted feature points to obtain feature descriptors; performing coarse registration by using an improved random sampling consistency algorithm to obtain an optimal transformation matrix; and finally, on the basis of the optimal transformation matrix, performing fine registration by using an iterative optimal point algorithm based on a KD tree, obtaining a final transformation matrix through iteration, and converting the source point cloud into a coo