Separating temperature, emissivity and downwelling radiance in thermal infrared pure-pixel hyperspectral images

The identity of a material may be obtained by measuring its emissivity spectrum. Unfortunately, hyperspectral remote sensing instruments measure spectral radiance, which is a nonlinear function of the material's emissivity and temperature, the atmospheric downwelling radiance, and other factors. Therefore, accurate interpretation of hyperspectral data requires estimation and separation of these quantities. By leveraging hyperspectral measurements of the same scene spread across time, this paper develops a new algorithm for estimating temperature, emissivity and downwelling radiance. We show the results of applying this algorithm to real hyperspectral data. The estimated emissivity spectra are in good agreement with laboratory measured spectra, and estimated downwelling radiance agrees well with downwelling products computed using radiative transfer models.