Research in Model-Based Change Detection and Site Model Updating
Final rept. Jul 93-Jul 96
UNIVERSITY OF SOUTHERN CALIFORNIA LOS ANGELES INST FOR ROBOTICS AND INTELLIGENT SYSTEMS
Pagination or Media Count:
USCs major research project deals with change detection and site model updating. A system was developed to validate building models and detect changes, i.e. in the dimensions of a building. The system uses the site model to predict the location and position of structures in the image. The comparison between model and image features gives an indication of the changes to the structures. Updating the model by adding descriptions of new structures requires site construction techniques. CMU has developed two systems for site model construction with focus on building structures. These systems implement perceptual grouping techniques to detect and describe the buildings in the scene by means of 3-D models. One system uses a single image to derive a model with the aid of 3-D clues, such as shadows and visible walls. Results from more than one image are combined to reinforce detection confidence and attempt to complete partial detection. This system also incorporates facilities for user interaction that help construct high quality site models with minimal interaction. The second system uses multiple images at all levels of processing and uses shadows and visible wall information to hypothesize and verify buildings and construct 3-D models. Results from RADIUS model board are presented.