Radar Forward-Looking Imaging for Complex Targets Based on Sparse Representation With Regularized AK-SVD and GGAMP-VSBL Algorithm

Radar forward-looking imaging is a leading topical issue in the radar research domain, which plays a significant role in many fields, such as target estimation, forward-looking detection, and precision guidance, by generating a stochastic radiation field through the wavefront modulation of the transmitted signals and associating the target echo with the radiation field to achieve the forward-looking imaging of the target in the beam. This method has superior performance on sparse targets, while the performance for complex targets will degrade severely. This article, therefore, proposes a radar forward-looking imaging method for complex targets based on sparse representation with regularized approximated K-singular value decomposition (RAK-SVD) and damped Gaussian generalized approximate message passing-based variational sparse Bayesian learning (GGAMP-VSBL) algorithm. First, the imaging principle and model are introduced to realize forward-looking imaging. Second, a regularization parameter is added to the classical dictionary learning (DL) K-singular value decomposition (K-SVD) algorithm to obtain the RAK-SVD algorithm, and it is used for complex targets to improve the accuracy of sparse representation. Third, the variational Bayesian inference (VBI) methodology is applied to sparse Bayesian learning (SBL) to form the variational sparse Bayesian learning (VSBL) algorithm, and Gaussian generalized approximate message passing (GGAMP) is embedded into VSBL to obtain the GGAMP-VSBL algorithm to speed up image reconstruction. Meanwhile, a new variance updating formula is adopted to further reduce the computational complexity and improve the robustness of the algorithm. The experimental results show that the proposed method can effectively image complex targets and has favorable performance in different complex scenes.