Supervision of Perception Tasks for Autonomous Systems: the Ocapi Approach

Abstract

This paper deals with the supervision of perception tasks, as a component of an autonomous system. First, the role of this component in an autonomous system is presented. Then a model of supervision of perception tasks is proposed: needed control mechanims as well as knowledge modelling are detailed. In the next section, we propose a software tool named OCAPI that uses artificial intelligence techniques. It has been designed as an expert system shell, and handles knowledge about image processing procedures including their goals and their proper usage. The knowledge base is highly structured to improve the efficiency of the system. The OCAPI system automates planning (selection of a sequence of programs) and control of ezecution (parameters initialization and adjustment, and results evaluation). An application for the supervision of a stereovisual process for obstacle detection in road scenes is descmbed.