Research On Spectral Registration And Pansharpening Of Hyperspectral Image

Hyperspectral imaging technology perfectly combines the spatial structure information of the image with the spectral information of the features reflecting the material composition.Hyperspectral imager utilizes phase shift velocity compensation technology to achieve high-resolution short-wave infrared hyperspectral image,and the device has the characteristics of multiple bands,narrow band widths,high spectral resolution,large amount of data,and integrated spectrum.Because of these advantages,hyperspectral images have been applied in many fields,such as forest fire prevention,oil and gas detection,hydrological and ecological monitoring,and geological exploration.Therefore,it is of strong significance to improve the quality of spatial resolution and to fully utilize overlapping spectral information of hyperspectral images.It is of great significance to improve spectral fineness and spatial resolution quality of hyperspectral images for subsequent work such as qualitative analysis of materials and classification of land surveys.Due to the limited technology of equipment currently,the resulting hyperspectral image has many problems such as discontinuous spectrum,overlapping spectrum,large amount of data,and low spatial resolution.In previous studies,there has been little research on the spectral processing of overlapping regions.However there have been many algorithms for improving the spatial resolution of hyperspectral images,but there are still problems such as spectral distortion.This paper mainly discusses the fusion of spectral overlapping regions and the improvement of spatial quality of hyperspectral images.The content is as follows:(1)Spectral ranges may overlap in hyperspectral images leading to increased data storage,and a certain range of the spectrum may not be sufficient to reflect material composition.Aiming at these problems,this paper adopts a spectral registration method based on the one-dimensional SIFT algorithm to fuse the spectrum of overlapping regions.Experiments show that the fusion result reduces the amount of hyperspectral image,and the fusion spectrum can better reflect the type of material to which the features belong,which makes the classification results of hyperspectral images significantly improved.(2)Aiming at the problems of previous work that the quality of the spatial resolution of hyperspectral images has not been well improved and the spectrum misidentifies,this paper proposes a pansharpening algorithm that utilizes the details from both hyperspectral and panchromatic image.The method firstly uses a Gaussian filter to extract the detail information of the hyperspectral and panchromatic image respectively and decomposes the hyperspectral and panchromatic images to different scales and resolutions,and secondly fuses the high-frequency detailed information separately.The fusion result is adjusted by the coefficients while fusing the detail information;Finally,the fused detail information is added to the hyperspectral image.Compared with other algorithms,the method proposed in this paper has better visual effects and better results on evaluation indicators.