Research On Resolution Enhancement Algorithm Of Remote Sensing Image Based On Data Fusion

Recent years,with more and more remote-sensing are launchef into space,the means for acquiring remote sensing images are various.How to effectively use the multi-source of remote sensing images,avoid weaknesses,and realize the complementary advantages of multi-source data has become a hot issue for researchers today.This paper has carried out research on remote sensing image fusion technology.Based on the analysis of the respective characteristics of TM and MODIS data,remote sensing images with high spatial and temporal resolution are obtained by merging remote sensing images with high spatial resolution and high temporal resolution,then increase the dynamics observation frequency of crops and serve modern intelligent agriculture.This paper first studies a typical space-time adaptive fusion algorithm(STARFM algorithm)that combines high spatial resolution and high temporal resolution of remote sensing images,and finds that the STARFM algorithm has two disadvantages: One is the directional dependence problem of the fused data reflectivity,namely the BRDF effect problem,and the other is the existence of the mixed pixel problem.In view of these two deficiencies,this article uses the Ross-Li model to correct the BRDF effect of MODIS data,and using two cycles of data to calculate the reflectivity of the center pixel to improve the accuracy of the results,thereby reducing the number of mixed pixels.influences.Experimental results show that the improved STARFM algorithm can better reduce the influence of mixed pixel and BRDF effects.Compared with the STARFM algorithm,the source entropy,correlation coefficient,average absolute error and mean square error of the experimental results obtained by the fusion method of this paper are all improved.At the same time,image fusion algorithm based on neural network,especially super-resolution convolution neural network-based(SRCNN)algorithm has been widely used in recent years.However,SRCNN uses Gaussian distribution or encoder allocation methods to initialize the weights.The uncertainty of these methods can affect the accuracy of reconstruction.Because of its convergence and optimization ability,the particle swarm optimization algorithm can set initial parameters of the network.In this paper,the PSO algorithm is used to optimize the convolutional neural network,initialize the SRCNN weights,so as to achieve the purpose of improving the accuracy of the resolution reconstruction.Experimental results show that the PSO algorithm has a high resolution of remote sensing images,and the source entropy,correlation coefficient,average absolute error and mean square error of the experimental results are all improved.