Robust Hyperspectral Targeting and Reacquisition Software

Hyperspectral imaging (HSI) sensors have the unique ability to identify objects based on the spectral signatures of their surface materials. With these signatures, movements of designated targets could be detected using images collected hours to days apart. Many targets of interest, such as vehicles, have specular (non-Lambertian) surfaces with multiple materials and surface angles. This complexity, combined with possible changing atmospheric/illumination conditions and viewing geometries, can cause the observed target spectral signatures to fluctuate considerably, making detection and/or reacquisition challenging. The goal of this R & D by Spectral Sciences, Inc. and the Rochester Institute of Technology is to develop novel algorithms and software to extract, predict, and account for these signature variations and thus advance the state of the art in hyperspectral detection and identification. The software would incorporate reflectance databases, an embedded 3-d target simulation code for calculating reflectance signature variability, including bidirectional reflectance distribution functions, detection/ID algorithms incorporating signature variability, and an interface to an AFRL data processing system. Phase I would provide a proof of principle for the data processing steps, which would be combined into a working Phase II software system and subsequently a Phase III commercial product for the remote sensing industry.