Research Of Ping-pong Referee Assistant System Based On Digital Image Processing

With the development of computer vision and image processing, Visual tracking has been widely used in national security, industrial production and sports.The traditional Ping-pong referee is manly rely on the people, due to the persistence of vision, the resolution of the eyes is24HZ, namely the response time of the needed is0.042s. Therefore, when the Ping-pong running at a high speed(can reach17m/s), the effective response time for referee is only0.16s. Thus, the referee can only capture to4-5frames, so the possibility of making misjudgment is large. In this paper, camera frame rate can reach60fps, which is2.5times of human reaction speed, so the system can obtain more track information, and improve the correctness and fairness of referee effectively.According to the characters of Ping-pong match, the paper analyzes the skill fields of Ping-pong referee system, and mainly focus on the technology of motion object detection, targets location, track prediction and referee algorithm and so on.The Ping-pong referee system based on image processing is researched and implemented on the basis of reading a great quantity of references and making a study of Ping-pong referee rules. The design is generally consisted of two parts:image processing and expert decision. The main content of image processing as follows:According to the thesis of Zhang’s, camera calibration is taken, and camera’s internal parameters and external parameters are obtained, and can obtain target’s3D information; The method of background difference is used to detect moving targets, combined with updating background image in time to improve the accuracy of detection; Combining the background difference theory with the information of color and shape feature, and use morphology filter to remove the noise, this paper has realized the recognition of ping-pong. And the experiment results prove that this method is effective; The state function model of the flying up phase and flying down phase of ping-pong and the motion model of the bounce process of ball are builded up, based on which the paper can predict the motion trace of the ball by Kalman filter. The predicting results can reduce the search range of ping-pong recognition, and can provide basis for the analysis of characteristics of the track. Data interpolation is used to smooth3D curves data. Referee expert module designed by the theory of expert system, takes the characteristics of ping-pong track as input, uses finite automata theory to describe the ping-pong referee rules, and verifies the correctness of the module through Stateflow.Finally, achievement and deficiency of thesis are summarized, and some ideas are given for the next development.