Imagery Manipulation for Simularor Database

Increasing application of visual simulation has led to a need for more efficient production of large high-fidelity databases derived from remote sensing imagery. Realistic databases require both 3-D models of buildings and structures, and high-resolutionterrain imagery. At present, remote imagery must be manually processed both to extract 3-D structures, and to correct the imagery for time-specific objects and shadows. This proposal describes the development, testing, and validation of imagerymanipulation software to automatically process remote sensing imagery to produce 3-D simulation database products. The application uses commercially available data from multiple sensor types. Specific processing steps include (1) innovative extraction ofbuildings using correlated segmentation and feature extraction from both imagery and height map data, (2) elimination of transident objects such as vehicles and aircraft by correlation and segmentation, (3) building shadow elimination by prediction andsegmentation, (4) terrain imagery filling to replace excised regions using statistical interpolation and (5) texture extraction and prediction for buildings viewed at oblique angles. The software creates 3-D- database objects in Open Flight format andprocessed terrain imagery in NITF format.