Research On NDVI Time Series Reconstruction Method Based On FY-3D Data

As an important part of the terrestrial ecosystem,vegetation is a natural link between the atmosphere and the surface.It is a common method to monitor vegetation growth by using remote sensing technology.After decades of development,satellite remote sensing technology has collected massive remote sensing data,which is also an important means of long time series and large space scale ground monitoring.However,due to the influence of cloud,aerosol and sensor characteristics,the acquired surface vegetation information data has a lot of noise and discontinuity,which reduces the use efficiency of remote sensing data and limits the ability to analyze land surface changes.Therefore,how to fill the missing remote sensing image,remove the noise contained in the image,reduce the impact of the sensor’s own performance on the image quality,and reconstruct a complete and high-quality remote sensing image with time series has gradually become a research hotspot in the remote sensing application field.This study takes the North China Plain as the study area,based on FY-3D MERSE-Ⅱ data,MODIS＿NDVI product data set and FY＿NVI On the basis of cloud detection,different NDVI time series reconstruction methods are used to reconstruct the above data for the main data sources such as FY＿NVI product data set and auxiliary data such as land cover,and multiple evaluation methods are used to systematically evaluate the reconnection results,In this way,we can get the applicability of different data sources to reconstruction methods and explore the spatial heterogeneity of reconstruction methods to different land cover types.The results show that each of the five reconstruction methods has its own reconstruction characteristics and data applicability FY＿NVI product data has the best reconstruction effect and fidelity evaluation.The reconstruction curve is relatively smooth,but the reconstruction result is low;The application of A-G fitting method to different data sources is different,FY＿NVI The reconstruction effect and fidelity evaluation of FY＿NVI product data are the best,MODIS＿The reconstruction effect and comprehensive evaluation of NDVI product data take second place.It is poor in the application of FY-3D MERSE-Ⅱ data.The reconstruction curve is smooth,and some missing values are filled;D-L hyperbolic method versus FY＿NVI The reconstruction effect of NVI product data is the best,and the reconstruction effect of FY-3D MERSE-Ⅱ data is better than that of MODIS＿NDVI product data,MODIS in fidelity analysis.The fidelity of NDVI product data is the best,and the fidelity of Fengyun data and NDVI product is relatively consistent.The reconstructed curve is a flat curve,with over fitting phenomenon,and some growth details are lost,which is quite different from the original curve.The HANTS method is inferior to other reconstruction methods in three data source reconstruction.The reconstructed curve is a smooth curve,which can effectively remove noise,but there is an over fitting phenomenon.This method is applicable to the reconstruction of NDVI time series data with large time interval or continuous missing;MVC method is the best of several methods in image reconstruction,which can remove the noise in the image very well.The reconstructed curve has poor smoothness and virtual height.