Fuzzy approach in model-based object recognition

Abstract

A fuzzy logic approach to pattern recognition is proposed along with the corresponding model-based problem solving algorithm suitable for recognition in intelligent robotics, where a good visual orientation is required for space orientation of a working robot. For simplified pattern recognition the angle-of-sight signature is used to represent the features of the object image. The features, defined in this way, are then used for building a reference model base. In addition, the membership function of the reference modes was defined in order to structure the demarcation rule base. Finally using the model base built and the rule-based algorithm proposed, the stored image of the "seen" object is classified as pertaining to the reference one or not. Some simulation results are included.<>