Hyperspectral Detector Enhancement Using Auxiliary High-Resolution Imagery

Space Computer Corporation proposes to enhance target detection performance of airborne hyperspectral imaging (HSI) systems by developing new algorithms that exploit the spectral content of HSI and spatial information in panchromatic high-resolution imagery (HRI) normally collected by such systems. Simultaneous generation and analysis of co-registered spatial and spectral features extracted from target cue imagery has the potential to increase the effectiveness of conventional recognition approaches based on either class of discriminants alone. Novel recognition strategies that “fuse” spatial and spectral features extracted from cued image chips may provide additional performance gains through more selective discrimination processes that operate in a joint feature space. In Phase I we plan to develop an effective set of spatial and spectral features and feature extraction techniques, along with related target discrimination metrics and databases, to demonstrate the practical utility of an automated approach to target cue confirmation and false alarm rejection. The work will utilize available HSI and HRI imagery from an existing sensor system, and will emphasize algorithms that are amenable to real-time implementation in Phase II.