An optimized hierarchical point cloud registration algorithm

Fuqun Zhao

doi:10.1007/s00530-024-01563-x

2024-12-14

PRE

Fuqun Zhao *

He Huang

Wenxiang Hu

DOI：10.1007/s00530-024-01563-x

原文链接

原文求助

摘要

En 中文

Point cloud registration is one of the important research contents in the fields of computer vision and application. It has been widely used in medical research, digital archaeology, reversible engineering and other fields. To improve the accuracy and convergence speed of the point cloud registration algorithm, an optimized hierarchical point cloud registration algorithm is proposed in the paper. Firstly, BFGS algorithm is used to optimize the normal distribution transform (NDT) algorithm, so that the algorithm iterates continuously along the descending direction of the gradient, so as to avoid the algorithm falling into local extremum and improve the efficiency of coarse registration; then, the main curvature, Gaussian curvature and average curvature of the point cloud are calculated and fused to descript the local feature of the point cloud; finally, the iterative closest point (ICP) algorithm based on fused curvature feature is used for precise registration of point cloud, so as to further improve the registration accuracy. In the experiment, the optimized hierarchical point cloud registration algorithm is used for public point cloud registration, skull point cloud detection and matching. The results show that, compared with existed registration algorithms, the proposed algorithm can effectively improve the accuracy and convergence speed of point cloud registration, and achieve accurate detection and matching of skull point cloud.

Keyword:

Point cloud registration

Normal distribution transformation

Gradient

Iteration closest point

Curvature