Polynomial-Fitting Temperature and Emissivity Separation in LWIR Hyperspectral Imaging

摘要

In this paper we present a new algorithm for Temperature-Emissivity Separation (TES) in LWIR hyperspectral imagery. The simultaneous retrieval of both physical quantities from the measured radiance represents an ill-posed problem, because the target spectral signature and its temperature are jointly combined into the remotely-sensed signal. Furthermore, the atmospheric downwelling radiance and the surface-emitted radiance are also coupled together through the emissivity, making the estimation even more complicated. The proposed technique solves the indeterminateness exploiting an optimization procedure, by estimating the best temperature that minimizes the atmospherical-residuals features inside the emissivity spectral shape. The temperature is estimated within a small spectral interval where the emissivity is smooth. In order to measure the signature smoothness in several intervals, an erosion-moving average filtering procedure is applied to the ground leaving radiances. Such filtering allows to establish the smoother region where the algorithm produces better results in terms of emissivity polynomial fitting.