A Novel Robust Color Tag for a Highly Efficient Soccer Robot Vision System

Soccer robot league is a competition on the cooperation between multi-robots in a highly dynamic environment. Getting accurate and precise information on time in a computationally constrained environment is challenging. In the small size league, each robot has a unique colored tag on the top for identification. Traditional color tag maximizes the colored pattern to give a stable recognition result, however, it might cause recognition confusion in the competitive game. This paper proposes a novel rotation-invariant tag to avoid the confusion without extra recognition constrains. A highly efficient vision system is constructed, each module is optimized to improve the recognition efficiency and robustness. The system focuses on the small size league but capable to be transferred to other leagues. The experiments evaluate the time consumption of each part and the whole system. The results demonstrate that recognizing the robot through our novel identity tag by the highly optimized system has achieved the real-time standard.