Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Learning of rigid point-based marker models for tracking with stereo camera systems

: Schwald, B.; Figueiredo, P.M.

Brunnett, G.; Göbel, M.; Müller, S.:
Virtuelle und Erweiterte Realität : 1. Workshop der GI-Fachgruppe VR/AR
Aachen: Shaker, 2004 (Berichte aus der Informatik)
ISBN: 3-8322-3225-7
Gesellschaft für Informatik, Fachgruppe Virtuelle Realität und Augmented Reality (Workshop) <1, 2004, Chemnitz>
Conference Paper
Fraunhofer IGD ()
tracking; mixed reality; automatic learning

Marker-based tracking systems like infrared stereo camera systems are often used to track rigid bodies that are equipped with a certain number of passive or active markers. Getting the knowledge on the geometrical constellations of these markers and their 3D position, to allow easily identification and tracking of various objects, is the basis for most systems. For simple objects with only few markers, getting this knowledge is not problematic. For tracking systems with a lot of cameras, having different views, also more complicated objects with several markers can be learned in simple procedures. However, typical stereo-camera tracking systems consist often in two cameras and if the necessity arises to track more complicated objects than e.g. shutter glasses or typical VR hand devices, where occlusion of markers can not be avoided, a procedure as presented in this paper becomes necessary. This procedure allows successive learning of rigid point-based models, by moving them in the field of view of the cameras, until they are completely measured. Both, the background to the procedures in the different phases and the results from tests with a reference to a real application are described.