基于谱对称的形状配准方法

Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-17 Cooperative journals: 《计算机应用研究》

韩丽 李丹 刘书宁 陈思宇 朴京钰 唐棣†

Abstract： Aiming at the issue of symmetric flips in the process of 3D shape registration, this paper developed an efficient shape registration algorithm based on intrinsic symmetric feature detection. Firstly, it constructed intrinsic symmetric point pairs of the model by heat kernel signature (HKS) and geometric constraints. Secondly, based on the spectral embedding space analysis, it extracted the intrinsic symmetric plane of the model and effectively identified the symmetrical properties of the model according to the model surface normal vector, getting intrinsic symmetry point pair. Therefore it presented the consistent spectral symmetry structure of the model. Finally, combining the coherent point drift method, it implemented the shape registration of non-rigid model based on spectral symmetry. The experimental results show that the matching method is efficient and robust to the non-rigid deformable shape matching. Moreover, the structural features in same category models also are effectively identified.

 Hits 2363  Downloads 1349  Comment