Dealing with speed and robustness issues for video-based registration on a wearable computing platform

We are investigating applications in which a field worker, equipped with a wearable computer, is networked wirelessly with a remote expert. In this paper we present a simple and robust augmented reality registration algorithm that can be used to lock annotations given by the remote expert onto parts of the scene viewed by the field worker through a head mounted see-through display. The algorithm can also be used to construct an image mosaic interface for the remote expert to place annotations regardless of the current viewpoint of the field worker. We also present a networkable desktop-based augmented reality prototype system to test the registration algorithm. A manual recalibration user interface is implemented to deal with registration errors.