Analysis Of Influence Factor On The On-orbit Alternative Spectral Calibration Of Hyperspectral Ground Objects Oriented

High-precision remote sensing quantitative research can’t depart from the support of hyperspectral remote sensing data, and delicate portrayal of ground objects spectral characteristics is closely related to the spectral calibration accuracy of hyperspectral data, factors such as the rigid body displacement after the satellite launch lead to spectrum drift. Thus, it is essential to conduct on-orbit spectral calibration of hyperspectral sensor.There is no research about the effect of underlying surface types and spectral matching methods on on-orbit alternative spectral calibration accuracy based on atmospheric transmittance method. In this paper, the oxygen absorption band near 762 nm was studied, spectral data of five underlying surface types and two matching methods were chosen to be analyzed,and spectral calibration errors of different underlying surface types and matching methods were compared. Results showed that calibration errors based on the spectral angle matching method and the euclidean distance method are almost the same. Vegetation and the olive green gloss paint are not suitable to calibration.Compared to the nadir observation, limb observation is less affected by surface reflection effects. The observation made spectral calibration method constrained by less influence factors under the imaging mode, which is easier to control calibration accuracy. the tangent heights and surface types need to be considered in the process of the sensor scans, experiments show the calibration accuracy is relatively stable when the tangent height in 6 ~ 11 km, the tangent height range is fit for limb spectral calibration. Because of the wide coverage of limb observation, underlying surface types in large-scale field-of-view should be considered in order to acquire higher calibration accuracy. Results indicate that vegetation and manmade objects because of its spectral curve of the complexity and uncertainty, the calibration accuracy is low, and even partially match error to 1 nm; snow, withered vegetation and sand of the accuracy is better, given the error in 0.4 nm or less, and it confirms the conclusions based on the nadir observation.