Registration Algorithm based on Point Cloud Skeleton

In response to the problems of substantial computation time consumption and slow convergence speed of the existing registration algorithms, an improved regional point cloud clustering algorithm is proposed in this paper, which is applied in point cloud coarse alignment, providing a new idea for point cloud alignment algorithm. The algorithm has the capability to dynamically modify the clustering radius based on the point cloud density across varying regions. By adjusting the clustering radius in real-time according to the point cloud density in different areas, it effectively prevents both the under clustering phenomenon in sparse regions and the overfitting problem in dense regions. Simulation experiments on real tree sample point cloud data demonstrate that the algorithm achieves high registration accuracy and excellent performance.