Research Of Omni-directional Vision System For Soccer Robots In RoboCup Middle-Size

As an important approach of acquiring environmental information,the vision subsystem is critical to the localization of robot. It has strong engineering background and involves a broad range of areas .So it's an important research topic.In this thesis, it is focused on the research of omni-directional vision system with mobile robot being a platform and Robocup middle league being the research topic background. The omni-directional vision system that consists of omni-mirror and camera is chosen to realize image segmentation, objection recognition and robot self-localization and some other function. Also,the system is able to observe the image of the entire area, getting precise location and orientation in the world coordinate system.The main work is as follows: the omni-directional image is collected by the vision system. The HSI color model is not sensitive to light, so we can use it for color calibration, image segmentation by run-length code, and objection recognition to the ball, goal, and the white lines. Two-dimensional interpolation method is discussed to reduce the error of distance calibration which is caused by the omni-directional machining accuracy. The robot self-localization is realized according to the features of image. Some improvement in hough transform to detected lines is mentioned while obtaining the line parameters. In order to further improve the accuracy of localization, Bayesian filtering, particle filter, Kalman filter, UKF-MCL method is discussed separately to solve the impact on the estimated uncertainty of moving model. Finally, by using Visual C + + 6.0, some experiments are done, such as HSI model-based image segmentation, distance calibration, robot self-localization.It has proved that omni-directional vision system and robot localization system can be widely used in the robot self-localization.