A GEOMETRIC REPRESENTATION FOR FUNCTIONAL RECOGNITION

Functional recognition is a paradigm that represents objects by their functional properties rather than by their geometric structure. The use of this paradigm constrains both the representation of object models and the reasoning necessary to recognize them. Common representation and reasoning techniques for structural recognition, such as CAD-based vision, are unsuited to functional recognition. This paper discusses issues in representatio n and reasoning with functional models, and describe a general reasoning framework that addresses many of these issues.