Temperature and emissivity retrieval from ground-based thermal infrared hyperspectral imagery of open scene

Abstract

Since the measured radiance in ground-based thermal infrared hyperspectral imagery of open scenes is simultaneously influenced by target surface emission, atmospheric path emission, and environment-reflected radiance, accurately retrieving surface temperature and emissivity for close-range targets remains challenging. To mitigate this coupling effect, this letter proposes a lightweight sky-radiance-assisted retrieval method that leverages the spectral similarity between background sky radiance and a standard atmospheric radiance template. Sky pixels are automatically identified from the hyperspectral image, and their averaged spectrum is used to estimate the ambient radiance directly, eliminating the need for auxiliary atmospheric measurements or external sensors. Based on the estimated ambient radiance, surface temperature and emissivity are subsequently retrieved using the ASTER-TES framework, which is further refined through wavelet-based suppression of residual atmospheric line absorption effects. Validation using HypercamLW data acquired over a short horizontal path demonstrates that, under favourable conditions with relative humidity below 28% and temperature contrast exceeding 6 K, the proposed method achieves root-mean-square errors below 0.5 K for temperature and 0.009 for emissivity. These results indicate that the proposed approach provides a practical and lightweight solution for temperature and emissivity retrieval in close-range open-scene thermal infrared hyperspectral applications.