Research on Moving Target Capture and Trajectory Reconstruction of Orthogonal Double Vision System

In a certain space, reproducing the trajectory of a moving target can provide a reliable data source for motion analysis. This paper uses two cameras with the same parameters to construct an orthogonal dual-view system, taking the table tennis movement in the table tennis space as the research object Carry out research on moving target capture and trajectory reproduction. Firstly, an orthogonal dual-view system is constructed, and the camera is calibrated, and a three-dimensional coordinate system based on a ping-pong table is constructed through dual-view; secondly, a HSV and RGB hybrid model is constructed for coarse filtering, and the background is filtered by the Gaussian mixture model. Morphological processing was used to filter the moving target region; finally, the two-dimensional coordinate of the ping-pong ball is obtained by edge extraction and Hough circle transformation. Aiming at the existence of smear in the image, the ellipse fitting method is used to make optimization. After obtaining the coordinates of the ping-pong ball, Kalman filtering is used to predict, supplement and correct the obtained series of coordinating. The experimental results show that this method can eliminate most of the interference, accurately obtain the three-dimensional trajectory of the table tennis ball and make short-term predictions.