| Surface reflectance is a critical physical variable that affects the energy budget in land-atmosphere interactions,feature recognition and classification,and climate change research.At present,the publicly available Landsat-8 and Sentinel-2 surface reflectance products have high accuracy,but the accuracy of domestic satellite surface reflectance has not been paid enough attention to,no standard surface reflectance products can be directly provided to users.In addition,due to different sensors and differences in imaging time,imaging conditions,and atmospheric conditions,the spectral values of multi-source and multi-temporal satellite remote sensing data are not comparable.When multi-source sensors are used to jointly complete an earth observation task,there are some problems such as inconsistency in radiance measurements and spectral band setting of different sensor.In order to solve these problems,according to the characteristics of the GF-1 satellite image,this paper couples the medium-resolution satellites(Landsat-8 and Sentinel-2)surface reflectance information,and through the processing of radiometric normalization,multi-source remote sensing data with spatial consistency,temporal consistency and radiation consistency are formed.The main research work of this paper is as follows:(1)Taking two groups of remote sensing images(GF-1 WFV1 and Landsat-8 OLI,GF-1PMS1 and Sentinel-2B MSI)as examples,the correlation between surface reflectance and Normalized Difference Vegetation Index(NDVI)of different sensors is compared and analyzed.The comparison results show that due to atmospheric effects,radiance calibration accuracy,inconsistent spectral response,etc.,data differences between sensor images can be caused.But there is a high linear correlation between the two groups of sensor data(GF-1WFV1 and Landsat-8 OLI,GF-1 PMS1 and Sentinel-2B MSI),cross-calibration and spectral information conversion can be used to reduce radiometric differences and achieve radiometric normalization.(2)Based on the Landsat-8 OLI surface reflectance image,according to the characteristics of wide coverage and wide-angle observation of GF-1 WFV sensor,a cross-calibration method based on atmospheric radiative transfer model is proposed to realize radiometric normalization.The verification experiment of calibration coefficient shows that the surface reflectance of GF-1 WFV4 image is close to the measured data and Landsat-8 OLI image's surface reflectance.Uncertainty of the measured data,errors generated during the cross-calibration experiment,and errors of the atmospheric radiative transfer model will allaffect the cross-calibration results.(3)Coupled with the Sentinel-2B MSI surface reflectance information,the radiometric normalization with the GF-1 PMS1 image is completed by means of relative radiometric correction.In view of the shortcomings of the existing radiometric normalization methods,this paper uses regularized Iteratively Reweighted Multivariate Alteration Detection(IR-MAD)rules to select high-quality no-changed points to solve the radiometric normalization equation.This method is universal and can be applied to images under different conditions.After relative radiometric correction,the target image(GF-1 PMS1)is closer to the reference image(Sentinel-2B MSI)in the surface reflectance value,image color and ground objects spectral reflective curves,which shows that this method is effective to improve the surface reflectance accuracy of GF-1 images and improve the radiation consistency of multi-source images.The radiometric normalization method of GF-1,landsat-8 and sentinel-2 satellites is of great significance for the collaborative application of multi-source remote sensing data and the production of quantitative remote sensing products,which provides a reference for the establishment of the radiometric normalization processing system of multi-source remote sensing data. |
PDF Full Text Request