A new method of recognizing partially occluded objects usingneural networks is presented. The neural network consists of asimplified ART-2 and a two-layer feedforward network, and its inputs arethe local features of objects. The network is first trained using a setof local features of known objects, then it can be used to recognizeunknown object(s). Our numerical experiments using this method showencouraging results, especially for recognizing the occluded objects
展开▼