The Realization Of Visual Servo System For Ping-pang Robot

Ping-Pang robot has been one of research focuses in the filed of robotics, as a platform of highly integrated advanced technology. And the visual servo system is the 'eye'of the robot, which obtains the information of the ball and predicts the trajectory of ball so that the robot controller can hit the ball back successfully. This paper elaborates the study and improvement of the visual servo system on the hardware designs and the algorithm developments for the related imagery processing, the recognition and the trajectory prediction. Firstly, the whole architecture of binocular visual servo system is introduced, including the hardware component and selection, and the software structure and functions. In particular, the solutions with coordinating the hardware and software are elaborated for the synchronous imagery processing. Secondly, the algorithms improvement on the ball detecting, locating and tracking is presented in detail and comparably analyzed. A novel linear and robust self-location algorithm of camera is also presented. Thirdly, a more exact model for the ball bouncing from the table is presented such that the prediction of the velocity change after bouncing has been improved. Regarding the air damping factor and external force acceleration as the unknown parameters, the extended Kalman Filter (EKF) is adopted to improve the recognition of ball trajectory. An adaptive trajectory prediction algorithm is proposed to combat the problem that the approximate ball flying models are different for different people playing the ball. And then, a new Ping-Pong pat posture detecting algorithm based on quadric surface intersection is presented. Lastly, a conclusion of our work is given, together with suggestions on the future work.