The correct alignment of real and virtual objects is one of the key technologies in Augmented Reality (AR). The user's point of view must be determined accurately and tracked over time. In this paper, we present an optical tracking system, focusing on real time performance and robustness including fast recovery from unpredictable abrupt motions, as well as stable handling of partial occlusions of tracking targets. The algorithm tracks linear features of landmarks and other objects. For increased performance, we do not extract entire lines but rather use only a few sample points. The 3D motion recovery is done iteratively using robust estimation methods. The system is able to run on standard hardware with high frame rate (20-25 Hz on a O$_2$) and is robust enough to track a hand-held or head-worn camera. We finish with a presentation of AR-applications in the domains of exterior construction, 3D user interfaces and games.     «

The correct alignment of real and virtual objects is one of the key technologies in Augmented Reality (AR). The user's point of view must be determined accurately and tracked over time. In this paper, we present an optical tracking system, focusing on real time performance and robustness including fast recovery from unpredictable abrupt motions, as well as stable handling of partial occlusions of tracking targets. The algorithm tracks linear features of landmarks and other objects. For...     »