Implement And Analysis Of A Push-broom Compressive Sampling Hyperspectral Imager

Compressive sensing theory is formulated as a novel sampling scheme to overcome traditional limitations imposed by the Shannon-Nyquist theorem.In the meanwhile,the spectral imaging techniques are impeded by the data explosion as a result of the higher imaging resolution,in which the data is multiplex of the data in traditional imaging techniques.Since there are massive redundant information in the hyperspectral imaging data cube and strong correlation between the spectral bands,compressive sampling is a unique option for the hyperspectral imaging acquisition.When introducing the theory of compressed sampling into hyperspectral imaging for the design of push-broom aerospace hyperspectral imager,in order to further reduce the demand for system hardware,the super-resolution design is mainly used on the coding template,which makes it possible to use lower data acquisition.Recovering high-resolution spectral images greatly improves the quality of the map.However,the use of super-resolution coding templates brings difficulties to the calibration measurement of the coding sampling matrix,and may face the problem of either calibrating the measurement data or the reconstructed spectral dimension.This paper studies how to reconstruct the map of high-resolution spectral dimensions under the condition of low-resolution small-scale coded template calibration measurement data.The main work of this paper includes the following aspects: 1.Realizing a spectral imaging experimental system by push-broom compression sampling mode with a coded mask(aperture).A compressed sampling spectral imaging experimental system based on coded mask and dispersive prism was built.The coded mask the image was pushed to change the sampling matrix.The calibration measurement of the coded matrices of each spectral segment and the acquisition of the target aliased data are combined for image’s three-dimensional data cube reconstruction.2.Proposing a method of taking the sub-pixel shift images of the calibrating frames to enrich the compressive sampling matrices and improving the quality of the reconstruction,and making a verification as it’s performance.This method uses the harsh calibrating data with wide spectral interval to produce new compressive matrices.From analysis of the modulation process of the dispersive prism,a conclusion that when the spectral sampling interval is small,the correlation between the coded sampling matrices of two adjacent spectral segments is very high and there is only one sub-pixel between the two patterns.Through a super-resolution and downsampling,images with sub-pixel displacement of the calibrated images are generated and have been used to produce compressive matrices of the spectral segments with tiny spectral intervals from the calibration band,contributing a higher sampling rate in spectral domain and a better quality of the reconstruction.3.Realizing the optimization method by the data from push-broom.In the calibration,there are a series frames of each spectral band acquired by pushing the code mask.All of these images were taken to produce compressive matrices in different spectral bands,and then utilizing the matrices to the reconstruction,which also provide an effectively improved outcome and higher compressive rate.The proposed method can reduce the requirement in the calibration of the coding mask and the realizing difficulty of the compressive sampling spectral imaging techniques.Besides,it has certain reference value for the application of the push-scanning compression sampling hyperspectral imaging technology device.