Method model for obtaining EVI index based on Bayesian theory

Abstract

The invention belongs to the technical field of vegetation remote sensing, and discloses a method model for obtaining an EVI index based on the Bayesian theory, and the method model comprises the following generation steps: (1) building MODIS EVI priori information: carrying out the spatial superposition of MODIS surface classification data and land utilization data, and carrying out the judgmentand extraction of all types of pixels of an MODIS of each land type; obtaining an EVI mean time sequence by combining the MODIS reflectivity band data and the QC quality control band of the MODIS reflectivity band data; performing filtering reconstruction; (2) generating an EVI initial value: introducing the MODIS EVI prior information by utilizing a Bayesian theory; carrying out mixed pixel decomposition on the MODIS EVI priori information and the MODIS EVI observation value in combination with a land utilization map; and (3) generating a Landsat EVI predicted value, using an EVI data prediction model, and using the Landsat EVI observed value at the pairing moment to reconstruct the EVI initial value. Based on the Bayesian theory, Landsat high-spatial-resolution data and MODIS high-time-resolution data are fused, and finally the EVI data with high spatial-temporal resolution are obtained.

Description

technical field [0001] The invention belongs to the technical field of vegetation remote sensing, and in particular relates to a method model for obtaining high time-space time-series EVI index based on Bayesian theory. Background technique [0002] MODIS vegetation vegetation products currently generate two global terrestrial vegetation index products, one is the normalized difference vegetation index NDVI, and the other is the enhanced vegetation index EVI, which is used in vegetation remote sensing research. At present, there are many method models to obtain EVI index with high spatial resolution and high temporal resolution, which can be mainly divided into the following categories: data fusion, mixed pixel decomposition, data assimilation and several methods combination model. [0003] STARFM (Spatial and Temporal Adaptive Reflectance Fusion Model) is a typical data fusion method, which obtains the regression relationship between Landsat values ​​and MODIS values ​​thr...

AI Technical Summary

Problems solved by technology

Although this method is simple and easy to implement, it only uses high-resolution data to calculate the area components, and the spatial detail information of high-resolution images is not used. The values ​​of the sub-pixels are the same. However, due to the differences in water and soil environment and biological characteristics, even at a small scale, the vegetation index of the pixels of the same vegetation category will generally be different. Therefore, the image obtained by the decomposition of mixed pixels is not A real high-spatial-resolution image, and the decomposed image will have patches at the scale of the decomposition window, which will greatly affect the appearance

There is also a high chance of getting unrealistic decomposition values

[0005] Many existing models that fuse high spatial and temporal resolution data have many errors that can be optimized and weakened due to their incomplete principles and incomplete theories.

And in the research of vegetation remote sensing, the vegetation index NDVI is commonly used to reflect the dynamic changes of vegetation growth, but in the period of vigorous vegetation (vegetation coverage exceeds 80%), NDVI will appear saturated, and NDVI is easily affected by noise, especially in In water body areas and near urban construction areas, the ability to deal with atmospheric background and soil background is limited, and EVI has been improved in these aspects, but there are still errors in the EVI obtained by the existing method model

Images