Imagery Manipulation for Simulator Database

Increasing application of 3-D simulation requires more efficient production of largehigh-fidelity databases derived from sensor imagery. Realistic databases requireboth 3-D models of buildings and structures, and high-resolution terrain 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 proposaloutlines the development of several techniques for automatic manipulation of remoteimagery to produce 3-D simulation database products. The techniques include(1) innovative extraction of buildings using correlated segmentation and featureextraction from both imagery and height map data, (2) elimination of time-specifictargets by segmentation and template correlation, (3) object shadow eliminationby prediction and segmentation, (4) terrain imagery filling to replace excisedregions using statistical interpolation, (5) texture extraction and predictionfor surfaces viewed at low grazing angles. The proposed work also includesdatabase format conversion to OpenFlight, TIFF, and NITF, and a preliminary designof an imagery manipulation tool to be developed in Phase II.The algorithm development and design activities of Phase I will provide thefoundation for an imagery manipulation tool prototype to be built in Phase II.By automating the 3-D object extraction process, the tool will greatly reduce thecost and production time of large simulation databases. Such a toolwould be valuable for organizations that produce or conduct simulations for tacticaltraining, mission rehearsal, commercial simulation, and urban planning.