Research On Sparse Reconstruction Algorithm Of Interferometric Hyperspectral Image Based On Compressed Sensing

Since th e second half of the 20th century,remote sensing technology has made great progress in theory,technology and application.Interference in hyperspectral image data is made based on the push type sweep type Fourier transform imaging principle of Large Aperture Static Imaging Spectrametry(LASIS)pushed sweeping resulting 3d image data via satellite,high resolution,its vast amounts of data to the data storage and transmission on limited bandwidth channel caused a certain degree of difficulty,so in view of the data itself characteristic design is suitable for high interference in hyperspectral data transmission method is imperative.As a new theoretical framework,compressed sensing provides new research ideas for signal description and processing.Unlike the existing sampling theorem,the theory uses a rate far less than Nyquist's sampling law to sample the signal,and then reconstructs the original signal with a high probability from these small observations.This efficient sampling method greatly reduces the sampling rate,so it has great application prospects in many research fields.This efficient sampling method greatly reduces the sampling rate,so it has great application prospects in many research fields,and has great significance for the transmission of interferometric hyperspectral images.Interferometric fringes contain abundant spectral information.However,when reconstructing the image using the traditional regular orthogonal matching pursuit algorithm,it is necessary to calculate the absolute value of the measurement matrix and the residual inner product.Because of the interference fringes with fixed position and large fluctuation of amplitude in the interferometric hyperspectral image,the variance of inner product calculation results is large,which will lead to the excessive number of atoms selected in the second selection according to regularization criteria in each iteration,thus resulting in the absence of atoms with higher matching degree in subsequent stages,resulting in support concentration.The proportion of atoms with high matching degree is lower.In order to solve the above problems,an orthogonal matching tracking algorithm based on correlation threshold is proposed in this thesis,which firstly adopts block processing and then selects interference fringe blocks.Because vertical interference fringes have strong unidirectional characteristics,this thesis extracts the interference fringes in the image according to the horizontal total variation value and carries out adaptive sampling.Then this thesis use a related coefficient threshold value to determine the threshold instead of a ROMP selection algorithm in the second,using relevant threshold can not only guarantee the correlation of eachselected atoms is high enough,and can choose the appropriate multiple atomic guarantee enough cycles,avoid subsequent matching degree higher the omission of the atom.Compared with the traditional ROMP algorithm,a large number of experimental data show that the precision of sparse reconstruction proposed in this thesis can be significantly improved.The above algorithm adopts correlation threshold for secondary selection,but fixed correlation threshold coefficient also limits the selection of correlation threshold that is more suitable for each image.Therefore,this thesis further proposes an orthogonal matching tracking algorithm based on adaptive threshold.This algorithm ensures that the relevant thresholds which are all optimal for the image are adopted in image reconstruction,and further improves the sparse reconstruction accuracy of the algorithm.