Distinctive representations for the recognition of curved surfaces using outlines and markings

Conference item

Recognising 3D objects from single images presents a range of significant problems, mostly to do with the nature and distinctiveness of the representations that can be recovered. In such special cases as polyhedra, surfaces of revolution, general cones, canal surfaces and algebraic surfaces, the geometry can be recovered with varying degrees of success and of ambiguity. We discuss these volumetric primitives, comparing their utility to that of surface primitives.
For a model based recognition system, representation is not simply concerned with particular geometric primitives, but the entire recognition process. In our view, representations should be motivated by the way quantities that are measurable in an image influence decisions throughout the recognition process.
When some geometric information is available, its potential distinctiveness can often be substantially enhanced by constructing representations that capture surface markings in an appropriate frame on the surface itself.

Authors: Forsyth, D; Zisserman, A; Malik, J

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Springer
• Host title: Object Representation in Computer Vision: International NSF-ARPA Workshop, New York City, NY, USA, December 5 - 7, 1994. Proceedings
• Pages: 89-99
• Series: Lecture Notes in Computer Science
• Series number: 994
• Place of publication: Berlin, Heidelberg
• Publication date: 2005-06-01
• Event title: International NSF-ARPA Workshop on Object Representation in Computer Vision
• Event location: New York City, NY, USA
• Event start date: 1994-12-05
• Event end date: 1994-12-07
• DOI: 10.1007/3-540-60477-4_6
• EISSN: 1611-3349
• ISSN: 0302-9743
• EISBN: 9783540475262
• ISBN: 9783540604778
• Language: English
• Keywords: computer vision; surface representation; surface markings; object recognition; invariant theory