Soccer Video Analysis using Monocular Camera

In this research, we present a technique to track the football players and ball for soccer video analysis using monocular camera. Moreover, we realize football tactics analysis. The camera pans and brings the players and the ball into the view, enabling to record the whole pitch by monocular camera. Therefore, we must track players and ball considering the camera motion and the occlusion.
Fig.1 shows the flow of our proposed method. At the beginning, we estimate pitch area from color information. We use Particle Filter to track each football player. In order to solve occlusion problem, we detect player and resample the center of gravity considering player’s velocity under crowded situations. A soccer ball is localized with labeling algorithm and thresholding. We project player’s point from original image to bird’s eye view.
Fig.2 shows the example of soccer tactics analysis. We enabled to acquire player’s speed and understand football scenes.

© Hirokatsu Kataoka