Research On The Generation Method Of GF-2 Multispectral Reflectance Products With High Spatial And Temporal Resolution

Remote sensing technology has the characteristics of macroscopic,periodicity,real-time and high efficiency,and has been widely used in meteorology,oceanography,agriculture and forestry,hydropower,transportation,military and other fields.High spatial and temporal resolution remote sensing images are of great significance for monitoring land surface change,estimating crop yield and inverting phenological parameters.Due to the limitation of satellite sensor hardware conditions and launch costs,existing satellite data have mutual constraints between high spatial resolution and high temporal resolution,and remote sensing images with high timeliness and precision cannot be obtained at the same time,which limits the application and research of remote sensing technology.Therefore,how to use the existing hardware platform and data resources,through software technology and algorithm model,improving the timeliness and accuracy of remote sensing image and meeting the practical application,has important practical significance.Aiming at the problem that remote sensing satellites can not acquire high time and space resolution images at the same time,this paper takes Pingding County in Shanxi Province as the research area,GF-2 and time series GF-1 WFV the surface reflectivity data in 2020 as the research object to set up based on temporal filtering fitting,panchromatic multispectral model of multispectral remote sensing image fusion and spatiotemporal fusion comprehensive reconstruction framework.In this framework,the discrete time series GF-1 WFV data are firstly filtered and denoised,fitted and interpolated to reconstruct daily data.Secondly,GF-2panchromatic and multispectral data were fused to improve the spatial resolution of multispectral data.Then,the reconstructed daily GF-1 WFV data and the fused GF-2multispectral data were used as the input data of the spatiotemporal fusion model,and finally the daily surface reflectance data with spatial resolution of GF-2 panchromatic image was predicted and generated.In the process of reconstruction,qualitative and quantitative evaluation are carried out on the results of each stage to provide reliable guarantee for the application of reconstruction results.The research work and achievements of the thesis are as follows:(1)Daily reconstruction of low resolution temporal remote sensing images.GF-1 WFV surface reflectance data of discrete time series can be effectively interpolated through SG filtering(Savitzky-Golay)and least square polynomial fitting method to achieve daily data reconstruction.The reconstructed data series conform to the law of actual data change,and the generated temporal NDVI data has higher accuracy.Compared with the existing public data set Mu Sy Q,the generated temporal NDVI data has higher consistency,strong correlation of quantitative indicators,and more prominent data integrity.(2)Gf-2 panchromatic multispectral whole scene fusion.MTF-GLP-HPM(The Modulation Transfer Function-generalized Laplacian pyramid-High Pass Modulation)fusion method was used to fuse the whole scene GF-2 data.The results show that the fusion algorithm has a stable effect and can effectively improve the spatial resolution while maintaining the overall spectral color.For different types of land cover areas,compared with the original multispectral data,the SSIM index of fusion results can reach above 0.9,and RMSE is less than 0.03.(3)Continuous time-spectrum high-resolution surface reflectance data reconstruction based on SG filter fitting,MTF-GLP-HPM and STARFM(Spatial and Temporal Adaptive Reflection Fusion Model).Using low resolution images reconstructed from time series reconstruction and high resolution fusion image as input data,the continuous time series multispectral data with spatial resolution of GF-2 panchromatic images were obtained by STARFM model based on single data pair.The results show that STARFM algorithm has high efficiency and can meet the requirements of large area data prediction.A 16-fold difference in the spatial resolution of the input data can also get better fusion results.The final reconstruction results show that the proposed method can completely reconstruct multi-spectral images with high spatial and temporal resolution,and the reconstruction results can correctly display the rule of surface reflectance changes with time,and have a high quantitative evaluation index.It is proved that this method can make full use of the spatial-temporal spectral information of the existing data resources and comprehensively consider the advantages and characteristics of various algorithms.It can solve the problem of the restriction between the spatial-temporal performance of remote sensing data from the software technology level,and provide a solution for the acquisition of high spatial-temporal resolution remote sensing data.