Shape differentiation of freeform surfaces using a similarity measure based on an integral of Gaussian curvature

Jing Fu, Sanjay B. Joshi, Timothy W. Simpson

Research output: Contribution to journalArticlepeer-review

22 Scopus citations

Abstract

Freeform surfaces are popularly used to design and model complex 3D objects. These 3D models are stored as computerized models in databases. To facilitate data retrieval and shape matching, a major challenge lies in defining and computing the level of similarity between two or more freeform surfaces. In order to explore the useful 3D information associated with the surfaces, an integrated approach based on the integral of Gaussian curvature is proposed to develop the measures of similarity of freeform surfaces. Specifically, the integral of Gaussian curvature is mapped into the 2D space, and a shape-based measure is developed using statistical methods to compute the level of similarity. For smooth surfaces, a fast approximation algorithm is developed to calculate the curvature of individual subregions. In cases where the target surface has a complex topological structure or a smooth surface is not available, the integral of Gaussian curvature for the discrete surface is first calculated at each vertex, followed by mapping onto a 2D spherical coordinate. The distance measure focuses on the local geometry, which is critical to investigate models with a certain level of resemblance such as products in a family. This proposed approach can be applied to surfaces under various transformations, as well as 3D data from various sources.

Original languageEnglish (US)
Pages (from-to)311-323
Number of pages13
JournalCAD Computer Aided Design
Volume40
Issue number3
DOIs
StatePublished - Mar 2008

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Computer Graphics and Computer-Aided Design
  • Industrial and Manufacturing Engineering

Fingerprint

Dive into the research topics of 'Shape differentiation of freeform surfaces using a similarity measure based on an integral of Gaussian curvature'. Together they form a unique fingerprint.

Cite this