An evidence based Object Recognition System using neural networks

Abstract

An Object Recognition System in an industrial environment needs to be capable of adapting to the changing environment in its task space, such as additon and deletion of objects, quickly and be able to re-adjust the feature space division optimally for the purpose of efficient discrimination. The above feature space division forms a part of any evidence/rule-based recognition system. In conventional digital computer based object recognition systems, rules are generated off-line for each specific set of objects. As is evident, this is unsuitable in an industrial environment where the task is continually changing. Therefore, an evidence based object recognition system based on self-organising n..