Study On The Optimization Of Atmospheric Correction Process Based On 6S Radiative Transfer Model

Atmospheric correction is one of the most important processes in remote sensing image processing,but on the basis of ensuring the accuracy of correction,the efficiency of atmospheric correction has not been improved effectively due to the large amount of data of remote sensing image and the calculation of processing process.With this in mind,the atmospheric correction method based on the second simulation of the satellite signal in the Solar Spectrum(6S)radiative transfer model is optimized and improved.Firstly,the influence of the correction process of atmospheric correction method based on 6S model and the influence of model input parameters on atmospheric correction coefficient is analyzed,and the main influencing factors of atmospheric correction coefficient are analyzed on the basis of the existing research results,combined with spearman correlation analysis.In view of the scope of atmospheric correction area,the simplified lookup table method and the polyamorphic fitting method are put forward which all based on 6S model.Then the relevant experiments are designed based on the 6S model,and the correction accuracy of the two methods in its scope of application is verified.The precision check results show that when atmospheric correction in small-range areas is carried out,polynomial fitting method can legally replace the traditional lookup table method,which is not only simple to operate,improve efficiency,but also maintain high accuracy.In the largerange areas,the simplified lookup table is optimized for one-dimensional than the traditional lookup table,so the speed of building the table is greatly improved,and the overall atmospheric correction efficiency is directly improved,while the accuracy of atmospheric correction can be guaranteed.Based on the limitation analysis of interpolation method in the traditional tabation method,the paper proposes to replace the traditional lookup table method with the back propagation(BP)algorithm,which has been verified that the BP neural network algorithm has higher precision,simple structure and easy operation.Finally,the two methods are verified based on the gaofen-1(GF-1)satellite data separately.After the two methods corrected the remote sensing image,the surface reflectivity of vegetation decreased significantly in the visible light band,increased in the near-infrared band,the vegetation spectral curve of the image was more in line with the actual characteristics.And the normalized vegetation index(NDVI)changed significantly before it was corrected.The change trend of NDVI is consistent with that of the correction result of the fast line-of-fight atmospheric analysis of spectral hypercubes(FLAASH)model,and the variation degree is similar,which further verifies the atmospheric correction accuracy of the two methods.The experimental results show that the simplified lookup table method and the polyamorphic fitting method based on 6S model improve the correction efficiency while maintaining the correction accuracy within its corresponding applicable range.