Research on Tracking Algorithm for Fast-Moving Target in Sport Video

Abstract

Tracking of player actions from sports video sequence is the hotspot in computer vision technology. The state transfer equation and the observing equation In the target tracking system are often nonlinear and non-gauss and mean shift algorithm cannot track the visual target effectively. The paper analyzes the principle and the shortage of the traditional mean shift algorithm. The reason for its weakness is analyzed too. A new tracking algorithm that combines the particle filtering and mean shift is proposed In order to effectively trace the fast-moving target. It estimates the position by particle filter in the previous frame of the targets. The position of the target is updated by the mean shift algorithm. Experimental comparisons show that it has better fusion performance for tracking the fast-moving players in sport video.