Research On Super-resolution Reconstruction Of Hyperspectral Remote Sensing Images

Hyperspectral remote sensing images not only contain two-dimensional space information,but also contain a lot of spectral information.It has a very broad application prospect in military and civil fields.In the imaging process of hyperspectral remote sensing images,the disturbance of atmosphere and the slight shaking of sensors will lead to the decline of image quality,which will seriously affect the practical application of hyperspectral remote sensing images in target detection,object recognition and other fields.Therefore,scholars have been exploring methods of obtaining high resolution images,and the super-resolution reconstruction of images has always been a hot spot of research.It is expensive to improve the resolution of hyperspectral images from hardware.So this thesis focuses on improving the spatial resolution of hyperspectral images through super-resolution reconstruction algorithm in software.The main work includes the following parts:(1)Aiming at the disadvantage of traditional projection on convex sets(POCS)algorithm which can not keep image edges well,a super-resolution reconstruction algorithm based on improved point spread function(PSF)is proposed.This thesis introduces the basic mathematical principle of super-resolution reconstruction algorithm,summarizes the basic principles of commonly used super-resolution reconstruction algorithms.Edge detection is carried out on the image edge,and improved PSF is applied to the detected edge pixels.The weight parameters of the improved PSF template in horizontal and vertical directions can be adaptively adjusted with the change of the detected edge slope.The effectiveness of the algorithm is verified by four different experiments.(2)Aiming at the deficiency that traditional POCS algorithm can only be applied to two-dimensional sequence images,a super-resolution reconstruction algorithm of three-dimensional hyperspectral remote sensing images based on POCS is proposed,considering the imaging characteristics of three-dimensional hyperspectral remote sensing images.The characteristics of hyperspectral imaging model are deeply studied,and a POCS constraint set is defined based on the spectral continuity of hyperspectral remote sensing images,which is added to the iterative optimization of POCS algorithm.The effectiveness of the algorithm is verified by four different experiments.