A contribution to knowledge representation and interpretation on objects an active perception system

We present in this paper a method for symbolic and structured description with several levels of abstraction. The application area concerns the construction and interpretation knowledge on the 3D objects in a active vision system. This method is largely consecrated to knowledge representation and scenes interpretation tasks, that the system carried out by autonomous robot system vision. The necessary knowledges to this task are formalised. Then, we propound to resolve the main interpretation problem, the control and the identification of objects, by prediction-checking of hypotheses. The representation, which is suggested, is based on the object-oriented paradigm. These objects are described in a redundant way and are considered as communicating processes. The decision for their detection is modelled inside them. With this end in view, an organisation system of perception and scenes interpretation tasks, actually is a prototype state, in order to maintain a coherent representation of a structured and evolutive environment. The released concepts and the proposed system are realised in a LE-LISP object-oriented environment based on the SHIRKA knowledge representation system