Atmospheric Parameter Inversion And Atmospheric Radiative Correction Based On Hyperspectral Image

Atmospheric radiative correction is a basic work of quantitative remote sensing, and the difference of the effect of atmospheric correction directly determines the accuracy of the subsequent quantitative remote sensing classification and parameter inversion. In the past decades, atmospheric correction algorithms have evolved from the earlier empirical line method and flat field method to more recent methods based on rigorous radiative transfer modeling approaches.In this paper, we present a new atmospheric correction algorithms based on spectral information and back-propagation artificial neural network(BP-ANN) used for hyperspectral remote sensing image, which can effectively avoid the error propagation through inversion of atmospheric parameters to radiative transmission model of atmospheric correction.The content of this paper may contains as follows:(1) The inversion principle of aerosol optical thickness was deduced and the comparation of different inversion methods was made in Chapter 3. By comparing the DDV algorithm and V5.2 algorithm, it can be seen that in the dense vegetation area, the accuracy of the DDV algorithm is credible, while in the non dense region using V5.2 algorithm could get better results.(2) The theory of water vapor content retrieval was derived using band-ratio method in Chapter 4. Both the two-channel ratio method and three-channel ratio method were used respectively on the near infrared region of two water vapor absorption regions to complish the inversion comparison. Verified by sensitivity analysis and results comparison, the result of using 1124 nm watervapor absorption band in the two methods both were closer to the ground observations. The accuracy of using three-channel ratio method was better than the two-channel ratio method, and additionally was as good as the APDA method in dense vegetation area.(3) A novel atmospheric correction method was proposed in Chapter 5. The new method could accomplish the atmospheric correction based on atmospheric radiative transfer model to obtain sample data and BP-ANN to generate the inversion model, which was ultilized to obtain atmospheric correction parameters band by band. This new method of atmospheric radiation correction is compared with the common commercial software FLAASH and ATCOR. It can be seen that the accuracy of the proposed atmospheric radiation corrected method is consistent with FLAASH and ATCOR by validation of hyperspectial remote sensing imagesa, and is better than FLAASH and ATCOR validated by simulated data.