Study On Sparse Sampling And Imaging Method Based On Compressed Sensing

As a new kind of information acquisition and processing theory, compressed sensing theory has become a hot research field of signal processing. This theory states that under the conditions of ‘signal compressibility’ and ‘non-relevant of observation system and representation system’, signal can be restored from a small amount of sampled data with a high probability. This makes the signal super-resolution on the space, time, and spectrum become possible. As most of the radio signals are compressible, which means the coded coefficients under a quadrature or over-complete dictionary are sparse. Therefore compressed sensing has broad application prospects in wireless communications and imaging, and many other applications For example, in synthetic aperture radar imaging, the received radar echoes received can be regarded as the echo of some strong scattering centers superimposed, this sparsity meets the need of the sparsity requirement of compressed sensing that signal should be sparse in some domain. This makes the radar imaging with compressed sensing possible.At present, although compressed sensing in radar imaging showed initial success, but there are still some problems. Firstly, the current compressed sensing radar imaging method using only the sparsity priori and imaging with less azimuth pulses. However, with the reduction of the number of azimuth pulses, the imaging quality fall rapidly. Secondly, because of the range down-sampling tends to decrease target energy, compressive sensing radar imaging were mainly sampled in azimuth dimension. However, with urgent needs of broadband / ultra-wideband microwave imaging in safety testing, non-destructive control and other areas, the range-azimuth joint super-resolution technology has become an urgent research problem. To sovle these problems, this paper studies the sparse sampling and imaging method based on compressed sensing.The main work of this thesis are as followers:(1) A range-azimuth joint radar imaging method based on compressed sensing is proposed. Firstly we analyzed the spasity in echo signal of synthetic aperture radar, studied the structure of sparse basis, and achieved joint undersampling on both fast time and slow time dimension. The method is used for super-resolution on SAR and ISAR imaging. The results showed that, compared to conventional microwave imaging method, compressive sensing imaging method can achieve low sidelobes and higher image quality with small number of pulses.(2) A sparse imaging method base on significant priori and weighted L1 optimization is proposed, In addition to the sparsity priori, significance and geometric structure of the target can be used as a priori information to improve the image quality under low sampling rate. Firstly, we use the low-resolution imaging results to get the saliency map, which distinguish a significant target area and background area. Secodnly, different weights to the target and background are calculated in the reconstruction process to suppress background clutter, while enhancing the strong scattering points in the target area. This method is experimented on a Yak-42 ISAR data with 256 azimuth pulses, and the result shows that the weighted L1 optimization base on saliency priori can differtntly treat target and background, and enhanced target scattering points, while suppressing background clutter.(3) A cooperative sparse imaging method based on graph Laplacian regulization is proposed. In addition to the target sparsity priori and significance, the relevance between the nearby range cells can further improve the imaging quality. On the basis of saliency based L1 optimization imaging, we studied the similarity of the nearby range cells, and construct the graph Laplacian regularization, which introduces a structural constraint to the original sparse optimization problem. An approach based on augmented Lagrangian multiplier method is proposed to alternately solve it. This method is experimented on a Yak-42 ISAR data with 256 azimuth pulses, and the result shows that the graph Laplacian regularization can effectively suppress the isolated scatterers in background clutter, while the impact for strong scatterers on the target is very small.(4) We studied an analog signal AIC(Analog-to-Information Converter) sampling method based on compressed sensing, a hardware-based MWC simulation platform is designed, mainly focused on multiband signal in communication systems. We analysis the principle and structure of MWC systems. Experimental results validate its reconstruction and stability, which laid the foundation for the hardware implementation of range sampling in sparse radar imaging.