Vision Using Streaming Up:Active
Visual Modeling for CAD/CAM and
Much of the research structured under this Project focuses on the intermediate
level of visual sensing as it relates to applications in advanced manufacturing
and mobile robotics. Funded as part of the Institute for Robotics and Intelligent
Systems (a network of centers of excellence), specific topics include the
inference of piecewise-continuous surfaces from discrete data sets, theories
and representations for 3D shape, recovery of shape from motion and stereo,
data and sensor fusion, object and scene modeling, strategies for autonomous
exploration, and the recognition of known objects in unstructured environments.
A key objective consists of achieving an automatic transformation of sets
of unconnected points in 3D space into surface mesh representations which
are properly formatted for interfacing with commercial CAD/CAM systems.
The Project also targets the development of tools for reliably inferring
scene models from sensor data. While the CAD application seeks object descriptions
with an accuracy of a metrological-like quality, the processing strategies
are equally applicable to the fusion of sparse data from multiple low cost
sensors into a virtual sensing device well suited for mobile robotics.
Figure 5: The Autonomous
F. Ferrie (Project leader), G. Dudek, M.D. Levine, S.W. Zucker (McGill),
P. Cohen (Polytechnique), D. Laurendeau (Laval)
Mon Jun 26 21:22:20 GMT 2000