Objects by Accumulating Up:Computer
Vision System for
Extraction of Features from Remote-Sensed
Imagery for a Search and Rescue Synthetic Vision Database
The generation of synthetic displays for the SVS application described
above relies to a large extent on the availability of annotated remote
sensing databases for the regions over which the aircraft will fly. Correct
localization of features, e.g., location of hydro pylons, transmission
lines, height of tree canopies, mountains, etc., translates directly into
higher degrees of certainty in the planning of flight paths and escape
routes. The sheer volume of data arising from both geographic scale and
the need for timely updates poses a considerable challenge.
This project, part of the GEOID national centre of excellence, is aimed
at the development of new algorithms capable of extracting these features
automatically and to higher degrees of accuracy by exploiting new developments
in image analysis and understanding. We are currently investigating the
use of sequential estimation techniques developed earlier for object recognition
as a basis for recognizing terrestrial features, given registered images
at different wavelengths and from potentially multiple sources. Another
part of this project deals with the problem of actually generating the
synthetic view from the data available in the remote sensing database.
P. Simard, F.P. Ferrie, D, Clausi (Waterloo), J. Elder (York), P.
Mon Jun 26 21:22:20 GMT 2000