Object structure and action requirements: A compatibility model for functional recognition

For an intelligent entity to carry out tasks in the real world, perceived three-dimensional shapes are transformed into objects identified by their functional category which makes explicit the roles or uses of objects in actions. This study describes an approach to the recognition of functions that combines ideas about representations of shapes, concepts, and object categories, with goal requirements of actions. A particular conceptual model of the compatibility between objects and actions is introduced, the outline of the solution is given, and the experimental domain of hand actions and of objects useful in such actions is described; the solution is currently under implementation and computational verifications. the article is organized as follows: first, the computational definition of the problem of functional recognition and a comprehensive theoretical framework for it is given and, second, the relation between primary functions of hand-manipulable objects and auxillary functions is discussed and modeled in the framework of fuzzy sets and possibility theory.