This paper presents an approach to match correspondences on 3D meshes, which is an important step for the design automation of customized freeform objects. For a given template model with a set of anchor points defined (knots of semantic features), we identify the corresponding points on the target model by minimizing the sum of differences by a series of transformation regardless of their differences in postures, scales and/or positions. The basic idea of our algorithm is to transform the target model to the template model iteratively. Once the correspondences between the surface points on the target model and the template are determined, we have essentially found the semantic features on the target model. We achieve this goal by four major transformations: 1) Multi-Dimensional Scaling (MDS), 2) Orientation Alignment, 3) Anchor Points Based Matching, and 4) Matching Refinement. The proposed method has been tested on a series of real human bodies to re-locate feature anchor points that are defined on the template model.

This content is only available via PDF.
You do not currently have access to this content.