The Research On Fusion Of Remote Sensing Images With Coupled HCS And NSCT Methods

Remote sensing technology as a discipline of rapid development,has a considerable advantage in access to information via the sensor surface of the object.Remote sensing imagery,as a carrier for recording sensor information,has a rich expression form.When sensors acquire ground surface information,different sensors have large differences in spatial resolution,radiation resolution,and spectral resolution.Therefore,The single image data source has defects in the identification and extraction of geographical information.Using multi-source image data fusion processing method can effectively improve the complementarity of data,solve data redundancy and other issues,thereby improving the interpretation accuracy and application efficiency of image data.Therefore,the research of multi-source image data fusion algorithm has always been academia.Research hotspots.In recent years,China launched a series of domestic high-resolution satellites,effectively promoting the application of high-resolution satellite data in various industries in China.However,there are still many problems in data fusion research of newly launched domestic high-resolution images.Therefore,this paper proposes a hyperspherical color space(HCS)and nonsubsampled contourlet transform(NSCT).A combination of remote sensing image fusion algorithms and different fusion rules for different data sources are used to improve the quality of fusion images.The paper focuses on the fusion of multi-spectral bands and panchromatic bands of high-resolution satellites,and the synthesis of aperture radar images of high-resolution satellites and multi-spectral images of high-resolution satellites.The main tasks of this article are:(1)A fusion algorithm for multi-spectral and panchromatic bands of optical images is proposed.Firstly,the paper automatically corrects the panchromatic and multi-spectral bands of the high-resolution image,and performs HCS transform on the multi-spectral bands to obtain the ? component.Next,the ? components of the panchromatic and multispectral bands are separately decomposed using HSCT,and a plurality of high frequency subbands of the panchromatic band and a plurality of high frequency subbands and one low frequency band of a low frequency subband and a multispectral band are acquired.Bands,and for a plurality of high frequency subbands of the panchromatic band and a plurality of high frequency subbands of the ? component,an improved Laplace energy and method are used to calculate the energy sum,for each of the mutually corresponding high frequency subbands Comparing the numerical values,keeping the energy and high part as the fusion part,using the weighted local direction entropy method to calculate the entropy value for the only low frequency subband of the panchromatic band and the only low frequency subband of the ? component,retaining the part with higher direction entropy As a fusion part.Finally,the preserved results are inverse transformed by NSCT and HCS to obtain the final fused multispectral remote sensing image.(2)An image fusion algorithm for optical image multispectral band and synthetic aperture radar is proposed.The paper first preprocesses and geometrically corrects the highresolution number 1 multispectral image and the high-resolution SAR image,and performs the HCS transform on the multispectral image to obtain the ? component.Then,the SAR image is denoised to obtain more effective edge and texture information.Secondly,SAR image and ? component are decomposed by HSCT respectively,and multiple high frequency subbands of SAR image,multiple high frequency subbands and one low frequency subband of a low frequency subband and multispectral image are obtained.In addition,the Laplace energy and method are used to calculate the energy sum for multiple high-frequency subbands of the SAR image and multiple high-frequency subbands of the ? component,and the values of the highfrequency subbands in the same direction are calculated.In contrast,the part with high energy and value is reserved as the fusion part.The only low-frequency sub-band and I-component unique low-frequency sub-bands of the SAR image directly retain portions of the multi-spectral image.Finally,the preserved results are inverse transformed by NSCT and HCS to obtain the final fused multispectral remote sensing image.(3)Through the fusion experiment of different remote sensing image data,fusion quality evaluation of fusion results.Including subjective evaluation and objective evaluation.The subjective evaluation is through visual interpretation and comparison of the images before and after the fusion to determine whether it is easier to identify features of the features.The objective evaluation mainly focuses on the spatial resolution and spectrum retention.The spatial resolution was evaluated by means of mean,standard deviation,information entropy,average gray,correlation,and clarity.Spectral retention is determined by comparing the spectral functions of the five features of water,vegetation,roads,bare soil,and buildings in the original image and the fused image.For the high-resolution panchromatic and multispectral bands,the method is compared with the CN method,the SFIM method,the PCA method,the GIHS method,and the Gram-Schmidt method for subjective evaluation and objective evaluation.Simulation experiments show that this method is superior to other algorithms both in terms of visual interpretation and in spatial resolution and spectrum retention.For the high-scoop No.3 synthetic aperture radar and the high-resolution No.1 multi-spectral image,simulation experiments show that this method is stronger than the CN algorithm,the GIHS algorithm,and the SFIM algorithm in terms of spectrum retention.On the lower buildings,it is stronger than the CN algorithm,the GIHS algorithm,and the SFIM algorithm.On the higher buildings,it is slightly weaker than the SFIM algorithm,and stronger than the CN algorithm and the GIHS algorithm.It is stronger than other three algorithms in terms of water,roads,and vegetation.