A Robust and Automated Hyperspectral Damage Assessment System Under Varying Illumination Conditions and Viewing Geometry

Some target signatures of interest in drought monitoring, flooding assessment, fire damage assessment, coastal changes, urban changes, etc. may need to be tracked periodically. In a typical change detection application, a hyperspectral image collected in an earlier visit may need to be compared with later images collected using different imagers with different viewing geometries, illumination, ground sampling distance (GSD), spectral sampling, signal-to-noise ratio (SNR), and atmospheric conditions. We propose a novel framework that can deal with all of the above challenges. We first propose to apply techniques such as flat-field to obtain the reflectance signature (fire damage signature, for example) from the target radiance signatures in a given hyperspectral image. The target reflectance signature is then saved in a target reflectance signature library for future use. After that, to detect targets (fire damage, for instance) in new images, we will expand a hyperspectral image processing system developed by the Johns Hopkins University/Applied Physics Lab (JHU/APL).