An Improved ICP Point Cloud Registration Algorithm Based on Three-Points Congruent Sets

Abstract

ICP (Iterative Closest Point) is the most widely used point cloud registration algorithm. However, some shortcomings still exist in this algorithm, such as (1) the need to manually determine the initial value of the registration; (2) the low efficiency for large-scale point cloud registration. Therefore, this paper proposes an improved ICP point cloud registration algorithm based on the three-points congruent sets. Firstly, the algorithm narrows the search of corresponding points by extracting 3D-SIFT key points. Then, possible corresponding points are confirmed by the position relationship between the centroid and key points. The optimal transformation matrix can also be determined based on the error function. Finally, the two point clouds are accurately aligned according to the resulted optimal transformation matrix and ICP algorithm. Experimentally, the algorithm is proved to be efficient without manual intervention.