The UMass RADIUS Project - Year 2.
Technical rept. Oct 93-Oct 94,
MASSACHUSETTS UNIV AMHERST DEPT OF COMPUTER SCIENCE
Pagination or Media Count:
A set of image understanding IU algorithms for automated site model acquisition and extension are being developed at the University of Massachusetts. In 1995, a building extraction system was implemented to acquire flat roofed, rectilinear building models from multiple, monocular images. This system hypothesizes building rooftops from a single image, then searches for supporting evidence in other views and determines the precise 3-D shape and location of each building via multi-image triangulation. Projective mapping of image intensity information onto these polyedral building models results in a realistic site model that can be rendered using virtual fly-through graphics. To perform model extension, a prior site model is registered to new images, and building model acquisition procedures are focused on previously unmodeled areas. In an operational scenario, this process would be repeated as new images become available, gradually accumulating evidence over time to make the site model data base more complete and accurate. Model-to-image registration techniques also are presented that can be used to automatically determine model-based local corrections to the resected camera parameters provided with each image.
- Optical Detection and Detectors