Relational fuzzy model: A representation for object identification

An intelligent vision system needs to identify objects that are presented in a scene. Some of the basic problems related to object identification are addressed. A new object representation scheme, the relational fuzzy model (RFM), is described. It inherits the advantages of the approaches of relational models and of recognition by parts. Fuzzy values and hedges are used to describe deformed geons that form the primitive parts of the representation. Based on the RFM representation, the structural complexity of an object and the structural similarity between objects are analyzed and measured. These fuzzy measurements aid in the identification of objects. An identification strategy used by a fuzzy knowledge-based vision system is also briefly described