1. Automatic registration algorithm for scattered point clouds based on curvature feature.
- Author
-
Ma Zhongling, Zhou Mingquan, Geng Guohua, Sun Jiaze, and Li Jing
- Abstract
Due to the bottlenecks such as low efficiency in the process of the virtual recovery of the Terra-Cotta Warriors broken fragments, this paper proposed a novel curvature-based point cloud automatic registration method. Firstly, it estimated curvature of each point by using MLS surface, then extracted the feature points based on the maximum local change of curvature. Secondly, it obtained the initial matching points by computing the Hausdorff distance of curvature, and acquired the accurate matching points by the PSO algorithm based on the maximum, minimum curvature similarity function. Finally, estimated the coordinate transform by the quaternion to achieve coarse registration, and improved the registration accuracy by the ICP algorithm. The experiments show the effectiveness and stability of this algorithm. [ABSTRACT FROM AUTHOR]
- Published
- 2015
- Full Text
- View/download PDF