Research On Spectral Reconstruction Algorithm Of Spatial Heterodyne Imaging Based On Compressed Sensing

Spatial heterodyne imaging spectroscopy is a kind of interferometric imaging spectroscopy technique which is modulated in space and time.By scanning and imaging the target object once,the spatial heterodyne imaging spectrometer can generate the atlas data cube with the target spatial information and spectral information.Space heterodyne imaging spectroscopy has wide application value in military reconnaissance,atmospheric monitoring and environmental detection.However,the graph data cube has a huge amount of data,and the data has the characteristics of redundancy.The storage,transmission and processing of the data require a lot of time and cost.Therefore,it is of great significance to study the high quality compression and reconstruction method of the data cube of space heterodyne imaging spectrometer.Compressed sensing theory realizes compression in the process of signal sampling,breaking through the limitation of traditional data sampling theorem,greatly improving the efficiency of data collection and reducing the waste of resources,which provides a new research idea for the transmission and processing of massive high-dimensional data.At present,many scholars have applied compressed sensing theory to the research of remote sensing.Firstly,according to the spectral correlation of spectral cube of spatial heterodyne imaging spectrometer,a compressed sensing reconstruction algorithm of spectral cube based on space-spectrum association is proposed.Firstly,the Kmeans algorithm is used to group the atlas data cube,then the reference image is reconstructed by smoothing L0 norm algorithm at a high sampling rate,and then the non-reference image is sampled at a low sampling rate,and the prediction and reconstruction of the non-reference image is completed by reference image.Through simulation experiments,the image quality(PSNR)reconstructed by the proposed algorithm is 2.6～3.3d B higher than that of the traditional reconstruction algorithm,and the reconstruction time is reduced by about 22～29 seconds,and the spectral curve of the target point is closer to the original spectral curve of the target point than that of the traditional algorithm.Then combining with the two-dimensional spatial imaging characteristics of atlas data cube,a compressed sensing reconstruction algorithm of Atlas data cube based on image segmentation is proposed.The algorithm firstly grouped the atlas data cubes,then reconstructed the reference image with high sampling rate by using block compressed sensing method,and divided the reconstructed reference image into foreground and background by using graph cutting algorithm.Then,according to the front background information of the image,different sampling rates of the front background were allocated,and all the other non-reference images were reconstructed by block using the allocated sampling rates.In this way,a complete atlas data cube is reconstructed.The simulation results show that the image quality(PSNR)of the proposed algorithm is improved by5～8d B and the reconstruction time is reduced by 22～48 seconds.The reconstructed spectral curve of target points is closer to the original spectral curve than the traditional algorithm,and its relative error is reduced by about half.The results show that the proposed algorithm is superior to traditional reconstruction algorithms in both spatial dimension and spectral dimension,which proves the feasibility of the proposed algorithm in cube compression of atlas data.