Doctoral Defense by Sebastian Thomas Lieberknecht
|
- Date: Monday, January 21th, 2013
- Time: 10:30am
- Location: 00.12.019, FMI-Building, Garching, Boltzmannstr. 3
|
Contributions to Real-time Visual Tracking and Live Environment Modeling for Augmented Reality Applications
Abstract:
Seamlessly embedding virtual geometries in a live camera stream is only possible when accurate estimates of the camera poses and a map of the environment are available. To this end, two real-time methods have been developed within the scope of this thesis. The first method extends template tracking to 2.5D and targets convex objects. The second method reconstructs and tracks objects of arbitrary shape based on RGB-D images. Furthermore, a novel methodology for benchmarking visual tracking algorithms is presented. The camera motion is precisely recovered using an external measurement device. The first dataset specifically designed for planar template-based methods was published in 2009. This dataset is still used daily at metaio, it has also been used at several research institutes.