Course filters for shape matching

Jonathan R. Corney, Heather Rea, Doug Clark, John Pritchard, Michael Breaks, Roddy MacLeod

Research output: Contribution to journalArticlepeer-review


The collaborative reuse of engineering data is one way that e-commerce can significantly reduce the development costs of new products. However the proliferation of Web-based catalogs for standard components (such as brochure-ware for nuts and washers) only hints at the possible productivity gains. The research reported here is motivated by the belief that shape matching technology is the key to enabling a much deeper form of Internet-based collaborative commerce. This article describes the coarse "shape filters" that support a 3D, Internet-based search engine, known as ShapeSifter, which aims to locate parts, already in production, that have a shape similar to a desired 'newly designed' part. The research vision is that once component models are on the Web, and indexed by, say, the ShapeSifter (or some similar system), a designer could query the search engine by uploading a 3D model of the part required. The search engine would then analyse the shape characteristics of the target model and perform a similarity match on the contents of its database. The challenge of the research is to identify shape metrics that produce effective characterizations of 3D models for similarity comparison purposes. In this context, the work reported focuses on the use of three novel convex-hull-based indices to carry out a preliminary coarse filtering of candidates prior to more detailed analysis (such as the construction of multidimensional feature vectors). The article describes the crucial role played by two databases of benchmark objects.
Original languageEnglish
Pages (from-to)65-74
Number of pages10
JournalIEEE Computer Graphics and Applications
Issue number3
Publication statusPublished - May 2002


  • Shape Matching
  • 3D Similarity Search
  • Part-sourcing
  • Geometric Feature Matching
  • Convex-hull


Dive into the research topics of 'Course filters for shape matching'. Together they form a unique fingerprint.

Cite this