Research On Reconstruction Algorithms Of Compressive Sensing Sar Imagimg

To achieve high-resolution images, synthetic aperture radar (SAR) faces the considerable technical challenges such as high-rate A/D convert and huge amount of data samples usually. Compressive Sensing (CS) theory shows that the high-resolution images can be reconstructed from extremely smaller set of measurements than what is generally considered necessary by Nyquist Sampling theorem, and offers many advantages in SAR imaging filed. Therefore, in recent years, a new approach named CS which applied to SAR has attracted more and more attention.CS reconstruction as one of the main content of CS theory research, the main process is accurate to recover the original signal from a small number of measurements. The effective reconstruction algorithm design is an essential part to make the CS theory become practical. In this paper, we mainly discussed the CS reconstructions based on adaptive filtering framework and smoothed ι0norm, and extended the smoothed ι0norm reconstruction algorithms to solve complex-value problem in SAR. The author’s main research works and contributions are as follows:ι0-LMS signal reconstruction algorithm base on adaptive filtering framework is improved. We reviewed the adaptive filtering reconstruction algorithms, focusing on an approximate representation of the ι0norm as well as how to reduce the steady-state mean square error (MSD) in this paper. To improve the performances of ι0-LMS, the Laplace function of representation of ι0norm which control parameter of approximate degree continuously changes, and the zero-attractor intensity gradually attenuates in the iterative process. Those adopted improvements can speed up the convergence and reduce the MSD.Smoothed ι0norm (SLO) reconstruction algorithm is improved. The approximate ι0norm function is represented by the’steep’higher approximation hyperbolic tangent function represented. Contrary to least mean square method, the improved Newton SLO (INSLO) algorithm adopts a modified Newton algorithm to optimize the approximate ι0norm. It is experimentally shown that INSLO obtain a better reconstruction performance.The SLO algorithms are extended to the complex-value case. In order to reconstruct the complex-value signal in CS radar imaging directly, a complex ι0norm approximation is presented, and the complex domain and the real domain with a unified form are pointed out. Then, in the same reconstruction framework and steps, the direct reconstruction of complex-value case is completed. The experimental results of complex domain demonstrate that the extended reconstruction is also effective in the complex-value case.