| With the development of Aeronautics,space and remote sensing science technology,remote sensing satellites carrying various sensors continuously launched,a large number of data of remote sensing images with different spatial and spectral resolutions have a blowout of increase.Especially in recent years,all kinds of high resolution civilian remote sensing satellites have been developed in many countries,so that a large amount of images has been applied to life,work and research.But how to get more accurate and comprehensive information by the available remote sensing image,which is in accord with the application in different fields,is an urgent problem to be solved.In traditional,people mostly use pixel-level image fusion algorithms to obtain remote sensing images that satisfy their own research and applications.However,with the increase of the resolution of remote sensing images,the amount of data needs to be processed increased.Traditional pixel-level image fusion requires processing of all pixel information of an image,and it has gradually failed to meet the needs of big data processing.Then,with the Compressive Sensing(CS)algorithm,the scholars put forward a remote sensing image fusion algorithm based on compressed sensing by fusing a small amount of compressed data after sampling.When the fusion image data,which contains high spatial resolution and spectral resolution information,is obtained,the amount of data calculating is reduced and the demand for huge data processing is alleviated to a certain extent.In this paper,the partitioned method is applied to the fusion process of compressed sensing image,and the optimal fusion coefficient is obtained by Particle Swarm Optimization(PSO),and a new type of block compressed sensing image fusion algorithm based on particle swarm optimization(PSO-BCS)is proposed.The algorithm reduces the memory allocation in the operation process and reduces the demand for the computer hardware when the algorithm runs.In addition,obtaining the fusion coefficient by particle swarm optimization algorithm can solve the problem that the fusion coefficient can't be adaptive with the change of the fusion data when traditional experience values are fused.In this paper,in order to verify the versatility of the algorithm,different spatial and spectral resolution images are used in the experiment,to analyze the characteristics of different target information in the fused image.High-resolution images choose GF-2 satellite images(true color combination band321),and low spatial resolution image choose Landsat 8 satellite image data(true color combination band432).In the part of experiment,the pixel-level remote sensing image fusion algorithms(color space transform HIS fusion,principal component analysis transform PCA fusion,color standardization transform Brovey fusion)are firstly analyzed and implemented.Then analyzing and implementing the CS fusion algorithm and the algorithm of this paper.The experimental results were evaluated from the subjective evaluation content and the objective evaluation index.The subjective evaluation process is mainly to compare the details of different features such as artificial features and vegetation,and the spectral characteristics of the fusion results such as hue,intensity,saturation,and visual spatial resolution of the fusion image,which ware enhanced and integrated.The enhancement of the edge and texture details of the image in the image was evaluated in three aspects.The objective evaluation process is to evaluate the fusion image effect through five different objective evaluation index parameters.The content of the evaluation involves the evaluation of the information volume of the image,the retention of the spectral information,and the sharpness of feature edge information and texture in the image.As a result,it was found that the detailed texture of the fused image obtained by the algorithm of this paper is clearer,the edge information of the object is improved,and the spectral information is kept relatively well. |
PDF Full Text Request