Intelligent hybrid hierarchical architecture based object recognition system for robust robot vision

Abstract

Judging from the fact that the human being depends on the vision information for more than 80% of information received from the outside, it is analogized that the robot’s vision information occupies a large part in the robot operation. Accordingly, based on the Vision System, the Robot Vision of Object Recognition, Face Recognition and Digit Recognition have been largely researched. In this paper, a method of modeling an object to hierarchical structure based on the object’s global features and local features is proposed. Also, the object recognition system of inferring the object through a synthetic approach combining the Statistical approach and Syntactic approach is proposed. As the approach method, the database is implemented by extracting the local features including a corner, arc and lines, and global features including the object’s area, Eccentricity and Elongatedness. By preprocessing in the inputted image, the object’s candidate area is extracted from a noise, background and unnecessary factors. The extracted candidate area’s image features are modeled and the object is recognized using synthetic approach. Based on it, the system which can recognize the object in various environments is implemented, and the performance is validated.