A Geometry-Based Interactive Pre-and Post-Processor for Reservoir Simulation and Visualization

Not Available This proposal addresses USAF needs for innovative sub-pixel detection, classification, and identification methods for hyperspectral sensors operating at thermal infrared wavelengths. Development of suitable algorithms for these wavelengths has been hampered by the spectral mixing of the reflected downwelling atmospheric radiance with transmitted emissions and atmospheric path radiance. Spectral Sciences, Inc. proposes to overcome these mixing ambiguities with a novel approach based on an atmospheric path transmittance and radiance correction procedure. Target and background source spectra are separated using subspace projection procedures that include physical constraints. Advantages of this technique include computational efficiency, discrimination against noise and lack of a priori spectral assumptions. In Phase I we will create necessary data bases (tasks1, 2 and 5) to evaluate the approaches to atmospheric corrections (task 3) and the algorithms for sub-pixel target detection and identification (tasks 4 and 6). This work will lead to an improved understanding of the performance limitations of LWIR hyperspectral sensors. In Phase II we will develop and optimize detection and classification algorithms for selected sensors and determine sensor performance specifications. The Phase II software will be delivered to the USAF for evaluation in sensor system design.