A Graph Theoretic Approach for the Identification of Objects Shape Taken from MPEG-7 Database

Objects never occur in isolation, instead, vary with other objects and in particular environment. In order to recognize the objects efficiently which are similar, there is a need for automating this problem. In this paper, we have proposed an approach to identify objects from MPEG-7 database consisting of 69 classes using graph theory. Graph parameters like graph eccentricity, graph diameter, graph radius and graph center values were used to form the feature vector. Back propagation neural network (BPNN) is used as a classifier. Features were reduced based on their performance in identification. Experimental results prove that an average identification accuracy of 91% is attained. The study is extended by combining other feature extraction techniques to train the neural network. This work finds its applications to train the robots in automobile industries to handle the objects.