Shadow compensation for computer vision in a robot soccer team

This paper introduces a method to compensate for shadow in a computer vision system for a robot soccer team. For a non-uniform lighting, a simple color classification based on a fixed threshold will fail. The shadow compensation method uses a lighting model to perform the color classification based on the albedo ratio. The system has been tested on the robot soccer team that participates in Robocup 2005. The system performed well with improving percentage of the correct classification.