| Compare with tranditional vision system, omni-directional vision system has a large field of view, information is more abundant. Omni-directional vision system can provide mobile robots with a full 360°view of their environment, which is mobile robots often require to perform navigational tasks such identifying landmarks, localizing within the environment, and determining free paths in which to move. In this paper, uses RoboCup middle-size soccer robot as the experimental platform, design and research soccer robot's omni-directional vision system.At the start, hardware and software design processes and methods of omni-directional vision system is introduced, and designs a omni-directional vision system based on hyperbolic catadioptric mirror, the field of view of this system can reach 360°×110°.In this paper, the methods to restore the distortion in omni-directional image is study, after analysis imaging principle of the omni-directional vision system based on hyperbolic catadioptric mirror and the perspective imaging principle. And a method is proposed which can restore the distortion of the image of the level of scence in the omni-directional image. Programming with Matlab,the distortion of the image is restored.In the filed of image processing, a method based on BP neural network to classify different color is proposed also. Ultilize this method to segment the vision image based on color. The results of experiment indicate that this method is feasible and can be utilized into Visual image processing of omni-directional vision system of soccer robot.At the end, a self-localization method of soccer robot based on white lines'matching is proposed. This method must construction a field modle image first. Then ultilize the methods which mention above to processing the real-time graphic and find out the white lines. At last, matching these two images based on white lines. The results of experiment indicate that the method with high precision and faster location. |
PDF Full Text Request