National Research Council of Canada. NRC Institute for Information Technology
Text, Technical Report
One of the main problems in computer vision is to locate known objects in an image and track the objects in a video sequence. In this paper we present a real-time algorithm to resolve the 6 degree-of-freedom pose of a sphere from monocular vision. Our method uses a specially-marked ball as the sphere to be tracked, and is based on standard computer vision techniques followed by applications of 3D geometry. The proposed sphere tracking method is robust under partial occlusions, allowing the sphere to be manipulated by hand. Applications of our work include augmented reality, as an input device for interactive 3d systems, and other computer vision problems that require knowledge of the pose of objects in the scene.
National Research Council of Canada
Report (National Research Council of Canada. Radio and Electrical Engineering Division : ERB), no. 1115 (28 October 2004).