Research And Application Of Multispectral And Panchromatic Remote Sensing Image Fusion Algorithm

With the maturity of sensor technology,a large number of new high-resolution remote sensing satellites have emerged,which provides a solid data foundation for image acquisition and application.However,due to the limitation of signal-to-noise ratio,only low resolution multispectral(MS)images and single band panchromatic(PAN)images can be obtained.Image fusion can complement the favorable information into the same image to obtain a high-resolution MS image.High quality fusion images are of great significance in the subsequent practical applications such as disaster monitoring and ground feature classification.Therefore,the fusion of MS and PAN images has always been a research hotspot and difficulty in remote sensing.At present,a large number of scholars have conducted extensive research on image fusion and proposed many algorithms,but the current fusion algorithm is still difficult to balance the spectral and spatial information of the image,and its applicability to new highresolution spaceborne images is poor.In addition,most fusion algorithms only use subjective and objective evaluation of fusion effect,and lack of practical application to indirectly test the effectiveness of the algorithm.Based on the above problems,this paper takes a variety of high-resolution remote sensing images as the research object,and proposes two algorithms with different characteristics for fusion and application research.The main work of this paper is as follows:(1)Aiming at the spectral distortion and insufficient spatial extraction caused by ignoring the low correlation of structural information of MS and PAN images by component replacement method,a new fusion algorithm based on HCS and multi-scale fast guided filtering is proposed.The algorithm optimizes the intensity component I extracted by HCS in MS image through fast guided filtering,extracts the high frequency of I and PAN image in multiple scales,obtains the spatial details highly related to MS image,enhances it by residual homogeneity learning,and accurately injects it into MS image to obtain fused image.Experiments show that the algorithm is suitable for different highresolution remote sensing images,and its spectral and spatial performance is better and stable than the traditional fusion methods in this paper.(2)Aiming at the problem that the multi-resolution analysis method is complex and can not process the R,G and B components of the image well,a fusion algorithm based on PCA and NSST transform is proposed.The algorithm makes full use of the multi-scale and multi directionality of NSST and the dimensionality reduction of PCA and eliminates the correlation of R,G and B.WLS is introduced for multi-scale detail extraction,and three bottom features based on phase consistency are used to fuse the high and low frequencies with different rules.Experiments show that the combination of component replacement method and multi-resolution analysis method can effectively complement each other and obtain high-quality fused images.(3)In view of the fact that only subjective and objective evaluation of fusion effect can not reflect the application of fusion results in classification,this paper further indirectly evaluates the advantages and disadvantages of fusion algorithm through classification accuracy,and analyzes the impact of fusion on classification.Based on the classification results,the best fusion algorithm and the best classification method are evaluated.According to the classification results,the double fidelity of spectrum and space is obtained,which makes the classification accuracy higher.At the same time,it proves the effectiveness and superiority of the fusion algorithm proposed in this paper.