Pre-processing Algorithms for Exploitation of Remotely Sensed Optical Spectral Imagery for Automated Target Recognition/Cueing and Multi-INT Fusion

High-fidelity modeling tools can be used to generate signature predictions for the pre-processing of hyperspectral image data.  The accuracy of signature predictions depends on factors that include sensor calibration error, model approximation error, and sensor noise.  In this project we will develop the SenseMod software product for the recovery of hyperspectral sensor models and for the use of these models in pre-processing algorithms. SenseMod will be based on MODTRAN(TM) forward models, innovative computational methods, and new statistical models.  SenseMod will consider the visible through the short-wavelength infrared and the long-wavelength infrared spectral regions.  The new sensor models will be combined with advanced algorithms to allow the detection and characterization of low-spectral-contrast targets.  The sensor model recovery and pre-processing algorithms will be designed to run in real-time on COTS hardware.  We will assess SenseMod using large sets of imagery acquired by several different hyperspectral sensors.  We will also consider the transition of SenseMod for a variety of applications. BENEFIT: Due to its unique position in the marketplace, SenseMod has great potential for the DoD and Intelligence Community.  The new product will greatly improve the operational utility of airborne and space-based hyperspectral sensors for a range of day/night surveillance, reconnaissance, and targeting applications.  SenseMod also has a high likelihood of success for Commercial/Civil applications in areas such as cartography, forestry, perimeter monitoring, and agricultural and environmental monitoring.  HyperTech is already involved in several programs that will lead to commercialization opportunities for SenseMod.