Abstract
In basketball analysis, a ball and players are often hidden by other players, which is called an occlusion, and such a condition is important to be solved. Therefore, we use image sequences taken from multi-views at the same time to avoid the occlusion and to obtain 3D information of a ball and players. When we suppose to record scenes of a game under a good condition by moving hand-held video cameras with adjusting zoom, we have to know the position and the pose of each camera so as to use a multi-view analysis system. In this study, we focus on the pose, i.e. the direction of the camera, and propose a measuring system. Its effectiveness is confirmed through experiments.
