Event recognition using object-motion context

Context based recognition system, which can improve the accuracy of scene recognition if robots cannot obtain adequate recognition features in the camera, has attracted researchers recently. However, there are few researches which focus on part-whole supplement system based on context which can describe the scene using detail parts in the scene and also reduce the ambiguity of each detail part using sentences in reverse. In this paper, we constructed the mutual supplement model focused on the ontological relationship between object and motion. We will take desk work in this research as an example of the experiment and show the usefulness of proposed model which can increase the recognition probability values of object and motion.