Design and Realization of Group ID Image Segmentation for RoboCup Small Size League

ID image segmentation is a part that is the most subject to light in the RoboCup small size league vision subsystems and with the largest quantity of calculation in the system. While most teams use YUV color space for image segmentation, we use HSV color space in our paper. Experiments prove that this color space is less subject to light. Secondly, while region segmentation based on window search is the prevailing way used by all teams, we use run-length coding for region segmentation. Experiments show the run-length coding satisfies the real-timeliness of the system and provides good calculation stability.