Real-time image processing in robot soccer

Abstract

This paper deals with design of fast image processing necessary for robot soccer system of MIROSOT category. The robots are small cubes with color patches on their top. A color camera situated over the playground captures the image of the playground. The purpose of image processing is to acquire the position and orientation of the robots in appropriately short time (the frame rate is 50 fps) from the image. In this paper we present our approach to this problem. Our techniques of image segmentation, objects classification and scene understanding are optimized for speed and robustness. We also present various types of the patches and methods for their recognition.