A self-localization method through pose point matching for autonomous soccer robot based on omni-vision

Abstract

A self-localization method through pose point matching based on omni-directional vision for autonomous soccer robot is proposed. An omni-directional camera was mounted on an autonomous soccer robot to take 360deg images of competition ground around the robot. First, the feature points of the white lines were extracted through the calculation of the changes of brightness on the ground. Then, the feature points extracted were corrected and matched with pose points. A relatively accurate self-localization was achieved by calculating the position and angle of the robot. Based on the information of self-localization, path planning, obstacle avoidance and other complex cooperation could be processed. This method has a good adaptability because, instead of using the color information of the goal, it depends on the white lines of ground only. According to test, this method is accurate and effective. The speed of the image processing could be up to 20 fps and the average positioning error was about 20 centimeters. It meets the requirements of Robocup MSL competition.