Vision System Research Of RoboCup Middle Size Robot League

Now the robot football match is getting more and more attention. This paper studies the design and creation of robot's vision sub-system based on the RoboCup mid-size football match and applies it to the match successfully.The vision sub-system is one part of the football robot system. It is the main route the robot gets information about match field. The successful building of this subsystem will enhance the capability of the Robot system dramatically. The installation of the mirror and camera is the basis for the vision system to work properly. Any mistake caused in this period will make it very difficult to correct in later procedure. So this paper analyzes this thesis analyzes the theory of catadioptric vision system and gets the requisite of installation to acquire good image result.The identification of the robot is separated to three sub-procedures including the selection of the predefined color space, the separation the color space and the identification of objects. But because of the existence of unequal lighting condition, the different color of objects and the disturbance of environment, it is frequent to get mistaken recognition. To satisfy the robust, accurate and fast requirement of the match, this paper studies ways to diminish the influence of these noises. Accordance with the different character of these noises, this paper used the method of noise clearance in color space selection procedure to minimize the influence of different color and unequal lighting condition. In color space separation procedure, this paper minimizes the influence of noise with the area change ratio and color spot density change ratio. In this procedure, this paper adds the identification of column accordance with the match rule. In the identification procedure, the change ratio of object position and moving condition is used to distinguish the objects. And some exception conditions are also concerned.Finally, this thesis analyzed several kinds of algorithm which can predict the location of objects and put a fast location method based on them.After the study of algorithm, experiment is taken to verify it. The experiment shows that the whole system is stable while keeps reaction fast. It can satisfy the need of the real match.