Silk Ball Pitching Detection Strategy based on Binocular Vision

Most of the research on intelligent pitching robots only focuses on robot control, ball recognition and tracking. In this paper, silk ball pitching detection strategy is proposed. Color recognition, ellipse fitting and binocular distance measurement are combined to realize recognition and position and detection of silk ball pitching. When silk ball is pitched, ribbons interfere with the shape of the ball, thus affecting the recognition. Color recognition avoids this problem effectively. Firstly, the silk ball and the ball frame are respectively identified through color recognition and ellipse fitting. Then, it is judged whether the silk ball appears in the frame area by the acquired image. In this case, the silk ball judged in the frame area may not be in the frame area actually, maybe in front of the frame area or behind the frame area. Secondly, the three-dimensional coordinates of the silk ball in the frame area are calculated according to the binocular distance measurement principle and coordinate transform correspondingly. Finally, whether the silk ball appears in the frame area is judged through the calculated silk ball's actual position. Through many experiments, a higher accuracy rate is obtained, meanwhile, the rationality and feasibility of the method are also proved.