An Optimized Partially Overlapping Point Cloud Algorithm

The traditional iterative nearest point algorithm (ICP) is often used in 3D point cloud splicing, but there are still some problems, such as low computational efficiency, and it is very easy to be affected by the initial pose and other factors, poor anti-interference ability, and easy to fall into the local optimal situation. For point clouds with only partial overlap, it is difficult for traditional ICP algorithm to accurately extract feature points and complete registration. Therefore, this paper proposes a rough registration algorithm and an improved fine registration algorithm that can provide excellent initial position. First, we extract the key data we need through clustering to reduce the impact of environment and noise on the registration accuracy. Then, the feature point supply sampling consistency (SAC-IA) algorithm is extracted by the fast point feature histogram FPFH with partial constraints, and the partial overlapping point cloud is spliced by the improved ICP algorithm. Moreover, the accuracy of the data set of the fitted model and the real point cloud is verified respectively, and the feasibility of the proposed strategy is proved by comparing with a variety of registration strategies.