A new method for image recognition of binocular vision system by tennis intelligent robot

Aim to let the intelligent robot could quick, accurately identify and locate tennis, a new method of tennis recognition is proposed in this paper based on color and shape. First, more independent goals image are removed through color filter method and image filter method with taking advantage of Matlab processing functions. Secondly, corner pixels were extracted to complete stereovision matching in order to obtain tridimensional information of position and shape under the extreme constraints. Sufficient experiments of tennis recognition show that the algorithm which is based on single tennis characteristic has poor accuracy with parallax positioning accuracy of 76%, distance accuracy of 71% and color accuracy of 67%. However, accuracy of the algorithm which bases on three characteristics reaches 91%. The results show that the algorithm has higher robustness and tennis position can be located and distinguished more effectively, work efficiency is therefore enhanced for the intelligent robot.