Research On Remote Sensing Image Fusion Technology Considering Image Spatial Heterogeneity

With the development of imaging technology and satellite launch technology,the spatial resolution,time resolution,and spectral resolution of remote sensing images continue to increase,and there are more and more image sources is coming.How to make good use of these data has become an important research direction.Remote sensing image fusion,as the pre-stage of using remote sensing images,has many advantages such as reducing information redundancy,suppressing noise,and improving image quality.There is significant variance in the spatial distribution of the ground features on the earth.As the information carrier of the real ground features,the pixel of remote sensing image should also be spatially heterogeneous,and the existing fusion algorithms seldom consider it.Therefore,this paper carried out technical research on image fusion from the following aspects:i.First of all,this paper introduces mature image fusion algorithms,including based on replacement fusion algorithms,such as IHS algorithm,PCA algorithm,GS algorithm,and Brovery-based fusion algorithm,as well as multi-scale-based image fusion algorithms,such as Laplacian Pyramid,Gradient pyramid and wavelet transform,etc.And analyze the internal principles of the fusion of these algorithms,and point out the advantages and disadvantages of these fusion algorithms.Introduced commonly used image fusion evaluation indicators,including subjective evaluation and objective evaluation.ii.Based on the assumption of spatial heterogeneity in remote sensing images,this paper proposes a remote sensing image fusion algorithm based on GWPCA,and verifies it on Landsat8 images and Gaofen-1 satellite images,using entropy of information,standard deviation,average gradient,and spectrum angle map.As the quality evaluation index of the fusion image,it is found that the GWPCA algorithm is significantly better than the traditional image fusion algorithm in terms of spectral retention.iii.Researchers may be more concerned about the fusion effect of one or several types of ground features rather than the entire image.Therefore,this paper focuses on four different objects: water,cultivated land,bare land,and woodland,explored the fusion results of these four objects,found that the image fusion effects of the four objects under different bandwidths have obvious differences,and found the optimal bandwidth of these four objects.