Open Collections

Augmenting reality, naturally

University of British Columbia

Augmented Reality (AR) is a promising new technology, aimed at enhancing the user's visual perception of the physical world with computer-generated virtual imagery. Virtual objects, such as rendered 3D models, 2D textures, highlights and labels, must appear correctly projected onto live video captured by a mobile camera. To achieve such a synthesis in a realistic manner, it is necessary to recognize what is viewed by the camera, and to accurately localize the camera and the virtual objects in the real world. In this thesis I address the challenge of automated and robust video augmentation in a variety of natural and unprepared settings. I have implemented a system which computes the camera pose by matching natural image features from a current video frame to a previously constructed 3D model of an operating environment. The system provides built-in tools for the construction of the scene model from reference images, the autocalibration of the camera and the interactive insertion of a virtual object into the modelled scene. Invariant natural features replace special markers for achieving successful recognition in images, and enable stable camera tracking in occluded and dynamic settings. Model recognition from arbitrary viewpoints removes the need to manually initialize the tracker. Experimental results demonstrate geometrically consistent augmentation for a wide variety of environments and unconstrained camera motion.

Thesis/Dissertation

application/pdf

2009-12-02

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/16178

Computer Science

University of British Columbia

UBCV

DSpace