<p>Aiming at the problem that it is difficult to capture the trajectory of the more specialised movements in badminton, which leads to the blurring of the tracking path, a dynamic path tracking method for badminton based on multi-feature fusion is proposed. By constructing the trajectory observation model, extracting the target features and distance features of badminton movement, calculating the state parameters of the tracking trajectory under each moment, and completing the trajectory tracking by virtue of the results of real-time state distribution of frame number continuity. And through the machine vision, we analyse the flight characteristics of badminton and the force trajectory model, predict its landing point, and complete the path tracking of badminton movement. The experimental results show that this method has better trajectory fusion effect, stronger stability of tracking curve, can accurately track the dynamic path of badminton in real time, and can provide theoretical basis for badminton running state analysis and prediction.</p>