In this paper, the deep learning method is used to detect the training videos to obtain the prediction trajectory of football motion. At the same time, we propose a football tracking framework based on Kalman filtering. This paper uses the Kalman filter algorithm to track the football target. We give solutions to the target occlusion problem and the trajectory splitting problem. The study results found that the traditional simple prediction method affected the performance of the competition due to its low accuracy. In this paper, the Kalman filter prediction method is used to improve the accuracy of football trajectory prediction.