Application Analysis Of Spatial-temporal Fusion Method Based On High Score Special Satellite Data

At present,due to the limitation of physical conditions of remote sensing satellites,remote sensing images with high spatial resolution and high temporal resolution are difficult to obtain simultaneously.Generally,remote sensing images with high spatial resolution features are relatively small,and remote sensing images with high temporal resolution features are often revisited longer,which is a "spatial and temporal contradiction" in current remote sensing images.In order to solve this problem,some experts and scholars have proposed a technique that combines high temporal resolution features of low spatial resolution remote sensing image data and high spatial resolution features of high spatial resolution remote sensing image data,which is spatial and temporal fusion technology of multi-source remote sensing image.The spatial-temporal fusion technology combines the high temporal resolution and high spatial resolution features of multi-source remote sensing image,and then achieve the continuation of high spatial resolution data in time in order to alleviate the mutual restriction of time resolution and spatial resolution.The existing methods of spatiotemporal fusion can be divided into three categories: the method based on spectral transformation,the method based on data reconstruction and the method based on data learning.In this,the fusion method based on data reconstruction is most widely used.This paper mainly analyzes three kinds of spatial and temporal fusion methods based on data reconstruction,including the spatial and temporal adaptive reflectance fusion model(STARFM)method,the spatial and temporal fusion method based on semi-physical model and the spatial and temporal fusion method based on learning.The main work is as follows:(1)The basic theory,development background and research status of multi-source remote sensing image fusion technology are systematically summarized.(2)Two research areas with relatively flat terrain are selected.The high resolution remote sensing image data covering two research areas(hereinafter referred to as high score data)used as data sources,a pretreatment process is designed according to the characteristics of data sources on the basis of summarizing the data preprocessing needed for high score data in this paper.(3)The STARFM method,the spatial and temporal fusion method based on semi-physical model and the spatial-temporal fusion method based on learning,and the comprehensive evaluation index of the fusion results is summarized.(4)The GF-1 data with high temporal resolution features and GF-2 data with high temporal resolution features are fused in the study area to generate remote sensing data with high spatial resolution and high temporal resolution features.After that,a quality evaluation system of fusion results is established,which is composed of visual analysis,quantitative evaluation and scatter diagram evaluation.