Knowing the locations of the players and the ball on a ground field is important for soccer game analysis. given an image sequence, we address three main problems: 1) ground field extraction, 2) player and ball tracking and team identification and 3) absolute player positioning, The region of ground field is extracted on the basis of color information, within which all the other processing is restricted. Players are tracked by template matching and Kalman filtering. Occlusion reasoning is done by color histogram back-projection. To find the location of a player, a field model is constructed and a transformation between the input image and the field model is computed using feature points. Otherwise, an image-based mosaicking technique is applied. Using this image-to-model transformation, the absolute positions and the trajectories of players on the field model are determined. We tested our method on real image sequences and the experimetnal results are given.
展开▼
