Physics-based synthetic spectra from n and k optical constants for exploitation of hyperspectral images

Hyperspectral imaging (HSI) data cubes are typically evaluated via methods whereby the processed spectral data are matched via some algorithm to libraries of hundreds/thousands of reflectance or emittance spectra of solids, in which the reference data often consist of pure chemicals of just one morphology and are often recorded for a specific application. We introduce an approach that can conveniently account for many different sample forms. Specifically, we demonstrate that different sample morphologies can be detected with high efficacy using “synthetic” spectra generated from physical constants and fundamental laws of optics. Experimental HSI data of thin-layer analyte chemical samples were evaluated using physics-based synthetic infrared spectra calculated from the optical n and k vectors in combination with absorption and reflection models. Besides the outdoor HSI data, hemispherical reflectance spectra were also collected in the laboratory using the same sample planchets. When evaluating the HSI standoff data using the synthetic spectra from two such models, the analysis found chemical identification to be nearly perfect. Both optical models could not only detect the target analyte chemicals but were also able to estimate a layer thickness that was found to be in quite good agreement (few micrometer level or better).