Broad Area Colloquium For AI-Geometry-Graphics-Robotics-Vision
(CS 528)
Visual localization and modeling of urban environments
Jana Kosecka
Department of Computer Science
George Mason University
April 23, 2007, 4:15PM
TCSeq 200
http://graphics.stanford.edu/ba-colloquium/
Abstract
Recent advances in techniques for capturing large scale models of
urban environments, give rise to many novel applications and motivate
development of various systems for enhacing navigational capabilities
of humans as well as autonomous vehicles. For outdoors environment a
prototype system for image-based localization will be described. Given
a database of views of city street scenes tagged by GPS locations, the
system computes the GPS location of a novel query view. The presented
approach is capable of dealing with heavily contaminated and ambiguous
data inherent to urban environments, using novel robust estimation
techniques capable of dealing with large number of outliers. I will
also briefly discuss a set of tools and techniques for 3D modeling of
axis aligned building models from a sparse set of views, capable of
capturing finer geometric details encoded in line segments. These
types models can be used as a front end to alternative rendering
styles or more efficient triangulations.