Research On 3D Reconstruction And Trajectory Prediction Of Table Tennis Under Binocular Vision

Because of the multi-disciplinary content of robotics,many scientific research works are carried out with robots as the research object.As a kind of robot,table tennis robot has important research value because it covers many core technologies such as machine vision,3D reconstruction,trajectory prediction,robot intelligent control,and embedded.This dissertation starts with a table tennis robot,and focuses on the 3D reconstruction and trajectory prediction process of table tennis.In this dissertation,we first researched the stereo vision system at high binocular speeds.It lays a foundation for the research of 3D reconstruction and trajectory prediction algorithm.(1)In the 3D reconstruction,its algorithm model constructed by ignoring the small tangential distortion avoids the complex algorithm that needs to iteratively solve the 3D world coordinates of the ping-pong ball.While improving the speed of the 3D reconstruction algorithm,the accuracy also satisfies the follow-up requirements.The prediction requires that the average error of the reconstruction algorithm is within15 mm.(2)In trajectory prediction,in view of the problem that the traditional UKF(Unscented Kalman Filter)algorithm predicts that the trajectory error of the rotating ball will increase with time,a table tennis trajectory prediction algorithm based on SPM-UKF(Simple Physical Model-Unscented Kalman Filter)is proposed,which can better correct the trajectory prediction error after collision.Compared with the traditional UKF algorithm,the post-collision error sum is reduced by about 10 mm.(3)According to the 3D reconstruction and trajectory prediction algorithm proposed above,a table tennis robot system was built and the hitting experiment was completed.