Geometric feature points based optical character recognition

Optical character recognition is an application of pattern recognition which automatically detects and recognizes the optical characters with out human intervention. All the characters are basically made up from three geometric entities, i.e. corners, endings and bifurcations which can be used to identify different characters. In this paper, we present a method for optical character recognition based on basic geometric features. The method uses a crossing number method to extract features from thinned character. The feature vector for each character consists of number of corners, endings and bifurcations. The classification stage recognizes a character by using a simple rule based method. The proposed system is tested using different samples for each character and the results show the validity of the proposed algorithm.