A new atmospheric correction method based on ANN in hyperspectral remote sensing image

Abstract

Atmospheric correction is a basic task of quantitative remote sensing in order to obtain accurate bio/geophysical products for Earth observation purposes. The effect of atmospheric correction directly determines the subsequent quantitative remote sensing quality of parameters inversion and the accuracy of classification. General flow of traditional atmospheric correction consists of two major steps: atmospheric parameters estimation and surface reflectance retrieval. In this paper, we proposed a new atmospheric correction approach based on artificial neural network (ANN). In order to retrieve atmospheric correction parameters, we employed an error back-propagation feed-forward neural network program. The new approach can build up the relationship of imagery spectra information and atmospheric correction parameters. The result was validated by two Hyperion scenes and one simulated hyperspectral image. Two commercial softwares, FLAASH and ATCOR, were employed to compare with the proposed method. The result demonstrated the proposed method have the highest accuracy and also have good consistency with the two commercial softwares.