Isometric registration of ambiguous and partial data

摘要: This paper introduces a new shape matching algorithm for computing correspondences between 3D surfaces that have undergone (approximately) isometric deformations. The approach makes two main contributions: First, the is, unlike previous work, robust to “topological noise” such as large holes or “false connections”, which is both observed frequently in real-world scanner data. Second, our samples space of feasible solutions uncertainty can be detected explicitly. We employ novel randomized feature order find subsets geodesics verify consistency. shows results real world and synthetic data sets could not handled using deformable algorithms.