Next:Synthetic Vision System for Up:Computer Vision Previous:Real-Time Image Segmentation for

Virtualizing Reality

The concept of virtualized reality relates to the problem of transcribing the form of the real world into representations from which synthetic views and/or behaviours can be generated. Much of our work to date has focused on the representation and recovery of 3D shape using a variety of sensor modalities (laser rangefinding systems, stereo imaging, shading analysis) and representations (parametric surfaces, articulated solid geometry, triangulations). Applications have included automatic CAD/CAM model generation from metrological data, environmental modeling for autonomous robotics, morphological characterization of skeletal components for the construction of medical prostheses, and the automatic generation of virtual reality models from sensor-equipped mobile robots.

As part of the VERTEX Centre of Excellence (Institute for Robotics and Intelligent Systems), our current work is focused on both the creation and augmentation of virtual reality from on-line measurements. For example in the VERTEX scenario we are using live data (television + structured light) from an underwater submersible to localize the vehicle within a known environment and to build task-specific descriptions of the surrounding vicinity. Although current work deals primarily with visual models, a long term objective of this research is to incorporate qualitative physical reasoning in order to predict behaviours in response to applied forces and disturbances.

T. Arbel, S. Benoit, P. Simard, G. Soucy, and F.P. Ferrie

Annual Report

Mon Jun 26 21:22:20 GMT 2000