Cost-effective camera pose estimation for basketball analysis using radon transform

We are planning to develop an analyzing system of a basketball game. Hand-held video cameras are assumed to be used, by which multi-viewpoint image sequences are captured. To obtain 3-D trajectories of a ball and players with solving occlusions, a camera pose has to be accurately known. In this study, we propose a method for measuring a camera pose by detecting orthogonal two vanishing points. In the method, the lines on a court are detected with low computing amount by the Radon transform. Our method is verified through real image experiments.