High Throughput Processing for Hyperspectral Imaging Sensors

Physical Sciences Inc. (PSI) proposes to develop a signal processing system to facilitate real-time detection of chemical and biological threats via analysis of hyperspectral infrared imagery produced by portable imaging spectrometers. The processor will support simultaneous detection of up to 16 different target materials using a spectral-matching-based detection algorithm. For datacubes acquired in band-sequential order, the processor will also support real-time image registration to facilitate on-the-move detection from vehicles and aircraft. Target detection decisions will be based on a likelihood ratio test which follows from a parameterized representation of the radiative transfer equation and a parameterized statistical model for background variability. In contrast to adaptive-subspace-based and matched-filter-based target detection, the proposed detection algorithm is not constrained by the presumption of a linear additive signal model or Gaussian noise statistics. Relaxation of these constraints yields superior Receiver Operator Characteristic (ROC) curves relative to adaptive-subspace-based and matched-filter-based detection.