SHAPE RECOGNITION METHOD BASED ON THE k-NEAREST NEIGHBOR RULE

In artificial intelligence, each shape is represented by the vector of the characteristic features and represents a point in the d-dimensional space of the descriptors. An unknown shape is rejected or is identified with one or more models (known shapes previously learned). The shape recognition is based on the similarity between the unknown shape and each model. The distance between the associated two vectors is used for estimate the similarity between any two shapes. The paper presents a recognition method based on the k-nearest neighbors rule. This method supposes two stages: model classification and shape identification. Only invariant descriptors are used for the model classification at many levels and this classification is realized only once, in the learning stage of the recognition process. The shape identification supposes the identification of the most similar class of models, the identification of the nearest neighbors and the identification of the most similar model. A shape is identified with a model if that shape is inside the region of the accepted tolerance of the model. The proposed method assures fast and simple shape recognition in robotics.