Study On Preprocessing Methods Of Airborne Hyperspectral Thermal Infrared Remote-sensing Data

With the development of the thermal airborne spectrographic imagers,the application of hyperspectral thermal infrared remote-sensing imagery is widely expanded.To retrieve the precise temperature and emissivity of the observed targets is the key to its application,and one of the hot and difficult problems of the research on its application as well.This article uses Thermal Airborne Spectrographic Imager(TASI)as the main data resource and some synthetic data as a supplement.The research focuses on the methods of atmospheric compensation and temperature/emissivity separation.The main results and conclusions are summarized as follows:1.Due to the dependence on the quality of the hyperspectral thermal infrared remote-sensing data,the factors of noise,spectral shift,spectral resolution,etc.poor the performance of AAC and ISAC.2.To reduce the uncertainty of the atmospheric profile,which calls MODTRAN to calculate the atmospheric spectrum,we give the theoretical framework of Atmospheric Compensation via Atmospheric Radiative Transfer Model Based on Priori Knowledge of Emissivity Spectrum,which introduces the priori knowledge of natural surfaces' emissivity spectrum as constraints.3.Based on the analysis of the convergence condition of NEM module's iteration,we derive its iterative regression equation,and its analytical solution.Using this analytical solution,the redundancy iterative process of NEM module can be greatly simplified.4.Based on the systematic study of the mechanism of temperature/emissivity separation,we discuss its Bayesian model.In the Bayesian framework,we give the equivalent form of ASTER-TES algorithm and one complex algorithm involving both ASTER-TES and ISSTES.5.Considering the redundancy of natural surfaces' emissivity spectrum,we propose a complete set of temperature/emissivity separation,which puts the sparse representation error under some certain domain as the constraint.6.The preprocessing practice of TASI demonstrates as below: the performance of ISSTES is poor;ASTER-TES and the complex algorithm are of almost the same performance;SR-TES performs best.