Study Of SAR Imaging Methods Based On Extended Sparse Structure

Synthetic aperture radar?SAR?is an active coherent imaging radar.Its advantages of all-weather,long-range and high-resolution make it useful in both military and civilian applications.In recent years,the rapid development of compressive sensing?CS?technology has provided a new direction for SAR imaging.SAR signals can sparsely represented based on CS theory for the sparsity of SAR echoes,which overcome the problems of big data and high sampling rate.In many practical applications,SAR signals often have block-sparsity.In addition,in order to obtain more detailed target information,we consider electromagnetic scattering characteristics to be more complex.Therefore,in this paper,we studies SAR imaging method based on extended sparse structure,and mainly studies two types of extended sparse structure:block-sparse structure and object-level scattering structure.The main research contents are as follows:Firstly,based on the study of CS theory and the electromagnetic scattering characteristics of the target,an SAR imaging model based on the attributed scattering center?ASC?model is given.Secondly,an iteratively reweighted least square?IRLS?SAR imaging method based on block-sparsity is proposed,which uses the log-sum function instead of the l₀-norm as sparse constraints.A strategy to minimize the upper bound of the cost function?called surrogate function?is used to simplify the original problem.The implementation of this method is the same as IRLS,and the update of the weight coefficient use a coupling mechanism to encourage block-sparse solutions.Simulation experiments verify that the method promotes the performance of encouraging block-sparse solution compared with Polar Format Algorithm?PFA?and Least Squares?LS?methods.In addition,in a hierarchical Bayesian framework,a pattern-coupled block-sparse Bayesian learning SAR imaging method is proposed for block sparse structures,assuming that the estimation of each element is related to the hyperparameters of itself and its neighboring elements.And the problem finally is solved by combining with Expectation-Maximization?EM?method.Numerical results show that this method has improved SAR imaging performance compared with PFA and LS.Thirdly,a variable scale iterative sparse SAR imaging method is presented.The quasi-Newton iterative method that does not need to calculate the Hessian matrix is used to reduce the amount of calculation required for imaging.Simulation experiments verify the imaging effect of the method in sparse scenes and block-sparse scenes.Finally,considering the background of impulsive noise,a robust SAR imaging method based on ASC model is discussed.Since the l₂-norm is very sensitive to impulsive noise,in this paper,we uses l_p-norm with 0?p?2 as the metric for the data fidelity and l₁-norm as a sparse constraint to establish an l₁-l_p optimization problem.This problem is solved by the alternating direction method of multipliers?ADMM?combined with the proximity operator.Finally,the high-resolution SAR image is achieved via extrapolation of the phase history based on the estimated ASC attributes.In addition,this paper further validates the effectiveness of the method after amplitude clipping processing,which limits the signal to a certain amplitude range,thereby partially suppressing impulsive noise.Numerical results show that the proposed method can effectively suppress impulse noise,and achieve object-level imaging,while preserving the integrity of the target structure,whether or not the amplitude of signal is limited.