Scene-based spectral calibration for thermal infrared hyperspectral data

Compared with a given laboratory calibration, systematic shifts on the center wavelength and the Full Width at Half Maximum (FWHM) of each band of a hyperspectral sensor, will emerge as the imagery environment changes. The center wavelength shift and the FWHM variation have influence on the inversion precision of the emissivity and temperature, especially near the atmospheric absorption bands. A technical process of spectral calibration for thermal infrared hyperspectral data was proposed, which was verified through a simulation experiment with the water vapor absorption band at 11.73 m selected as the reference band. The experiment shows when the spectral resolution is 50 nm, the center wavelength shift ranges from -50 nm to 50 nm and the FWHM variation ranges from -25 nm to 25 nm, atmospheric water vapor content has the most influence on the error of the estimated wavelength shift and the FWHM variation. Meanwhile, the error distribution were also fitted using different surface functions, and the error distribution models used to estimate the errors were obtained. When the atmospheric water vapor content was high enough, the estimation error of the spectral center wavelength shift was able to reach within 1 nm. Finally, the proposed approach was applied to spectral calibration of the airborne thermal infrared hyperspectral data obtained by a push-broom hyperspectral thermal-infrared imager. The results show that the center wavelength shift of the thermal infrared hyperspectral imager is 28.4 nm and the FWHM variation of the imager is -18.5 nm.