Tracking Players in Volleyball Matches using Vol-Bot-SORT

The general tracking algorithm is deficient in the face of the problems of long-term occlusion and irregular movement between players in volleyball games. In this paper, on the basis of tracking players based on the YOLOv8 algorithm and the Bot-SORT algorithm, the effective tracking of players is realized by constraining the object of the Kalman filter, integrating the momentum feature and introducing the altitude state as a weak cue to adjust the IOU, and combining with the number of players constraints. Experimental results show that compared with Bot-SORT, our algorithm reduces the number of ID exchanges by 40% and the false detection rate by 97%, which effectively improves the tracking accuracy.