Abstract
Receiving incomplete signals is common in real inverse synthetic aperture radar (ISAR) imaging. If incomplete signals are used for imaging, severe grating lobes will be present in the obtained image. The gridless sparse recovery (SR) method, called atomic norm minimization (ANM), can reconstruct missing signals accurately and is particularly well-suited for sparse ISAR imaging. The Frequency-selective (FS) ANM (FSANM) method uses prior knowledge to improve the estimation performance of the ANM method. However, the estimated performance of the FSANM method is still limited by convex relaxation when frequency separation is unreasonable, which leads to unsatisfactory imaging results. To break this limitation and improve ISAR imaging performance, a novel gridless sparse ISAR imaging method was proposed, which can be called FS reweighted ANM (FSRAM). The proposed method introduces a non-convex metric to establish a connection between atomic \({\ell _0}\) norm and \({\ell _1}\) norm. According to the non-convex iterative solution model, the implementation of a semi-definite program (SDP) for FSRAM was derived. The simulated experimental results indicate that the proposed method maintains a high level of estimate performance even in cases when the frequency separation is not acceptable. The real experimental results show that the proposed method can obtain better quality ISAR images.