Spectral Information Content Algorithm for Automated Signature Assessment

Passive hyperspectral imaging (HSI) systems can be coupled with exploitation algorithms for target detection, classification, and identification using the spectral signatures of manmade and natural materials. The detectability of a particular target is dependent on the material optical properties, sensor noise characteristics, and environmental conditions. Previously developed algorithms based on statistical methods and information theory enable the assessment of the impact of material phenomenology on detection performance. However, these algorithms are band-variant and un-normalized. Here, we develop the spectral information content (SIC) algorithm to quantify signature strength and to automate sensor selection for detecting a particular target. Furthermore, we illustrate that the SIC algorithm is reasonably correlated with target detection performance.