On-board spectral calibration algorithm for an airborne hyperspectral imager and elimination of the effect of the atmospheric underlying surface.

Abstract

As an image-spectrum merging technology, the hyperspectral sensor has become an important part in remote sensing. The spectral calibration results of the hyperspectral sensor measured in the laboratory should be refined when applied to the on-board imaging spectrometer due to variations between the laboratory and actual flight environments. The larger the spectrum offset of the spectral response function (SRF), the worse the retrieval accuracy of the reflective characteristics of the targets. Therefore, an on-board spectral calibration algorithm based on standard diffuse boards with a gradient of reflectivity is proposed in this paper. By constructing the differential function of radiation transfer coefficients, the center wavelength offsets of SRF can be solved. In addition, a back-propagation neural network has been established to eliminate the effect of the atmospheric underlying surface and improve theaccuracy of on-board spectral calibration. The three-sigma confidence interval of the on-board spectral precision is $ \pm 0.23\,\,{\rm nm}$±0.23nm (uncertainty $ \lt {0.05}$<0.05 spectral pixels). The algorithm can be applied to a general hyperspectral sensor.