关键词: 合成孔径雷达, 交替方向多乘子法, 结构特征增强, 相变图

Abstract:

In synthetic aperture radar (SAR) imagery, the conventional convex optimization algorithms based on least absolute shrinkage and selection operator (Lasso shorted) may lead to the loss of structural features in weak scattering and affect the accuracy of the sparse signal recovery when sparsity enhancement is carried out. To solve this problem, a novel algorithm based on two-layer sparse group Lasso of alternating direction method of multipliers (SGL-ADMM) is proposed in this paper. Based on the block structure (group) feature of the scatterer as a prior, the proposed method firstly introduces the proximal operator corresponding to ?₁ norm according to the classified feature of the SAR data. Then, it realizes the sparse feature enhancement in the first-layer and second-layer SAR group by using the Gauss-Seidel strategy to perform dual iterative operation for the proximal operators in the ADMM framework. In addition, the ?_F norm in the mixed norm is the Gaussian penalty item, which can smooth the SAR echo data for realizing the enhanced structural features of SAR, accordingly. Therefore, the proposed algorithm can realize the joint enhancement of sparse feature and structure feature during echoed complex SAR data processing. In the experiments, the simulation data and the measured data of SAR, SAR ground moving target imaging (SAR-GMTIm) and inverse SAR are selected to compare the proposed algorithm with the traditional algorithm from the qualitative and quantitative perspectives. The phase transition diagram (PTD) method is used to verify the reconstruction ability of the proposed algorithm in the qualitative analysis, so as to verify the effectiveness and superiority of the proposed algorithm in SAR sparese and structural feature enhancement.

Key words: synthetic aperture radar (SAR), alternating direction method of multipliers (ADMM), structural feature enhancement, phase transition diagram (PTD)