Research On Trajectory Tracking Technology Of Table Tennis

With the continuous development of science and technology, the development of robot is booming all the time. Table tennis robot, as a typical entertainment robot,has been studying in recent years in China, it needs greater investment to fill the gaps. The trajectory prediction algorithm is a key research point in table tennis robot.It is an emerging research to take a good use of the video tracking technology into sports training to improve the efficiency of scientific and sports training.Moving objects can make use of video tracking technology for capturing and tracking and to analysis of the state of motion and characteristics for training.However, these algorithms are often based on the specific environment, such as a static background, indoors, no shelter, etc. In order to achieve stability tracking in the video sequence, and analyses the target motion implicit semantic information intelligently and accurately, and ultimately to achieve a common goal of automatic detection and tracking system is still a challenging task.In this paper, with the help of sony camera, table tennis robot, as the object of the study, is used to motion detection, target location, trajectory prediction at different speeds, different angles, different rotation on the ball.First of all, it discusses the background, significance and the status of the research and overviews video object tracking and basic techniques, including the commonly used tracking algorithm and difficult issues and technical requirements;Secondly, it analyses of the force of the ball during the flight, simulates its physical model. Finally, due to the complexity of the experimental background, low light, a variety of experimental facilities and equipment, and doors which have similar color with table tennis, when adapting the frame difference method for tracking table tennis, the moving target can be detected and tracked, however, in the process of doing image difference, a lot of clutter false targets emerge. To solve this problem, using Kalman filter algorithm, and combining Kalman filter and gaussian mixture background model to a new algorithm, and it realizes the ball detecting and tracking, the effect has been significantly improved.