2D Off-Grid Laplace Adaptive Pattern-Coupled Sparse Bayesian Learning for ISAR Imaging of Incomplete Data

In inverse synthetic aperture radar (ISAR) imaging, traditional methods become ineffective in scenarios with incomplete data, such as sparse aperture (SA) and limited bandwidth (LB). Although sparse signal processing (SSP) techniques have been introduced to enhance performance in ISAR imaging under these conditions, the two-dimensional (2D) off-grid problem and the structural features of ISAR images still require further investigation to improve imaging quality. In this article, a novel 2D Laplace adaptive-pattern-coupled sparse Bayesian learning (LaAPC-SBL) with 2D off-grid compensation is proposed to effectively capture the 2D structural features of ISAR image and correct the off-grid error simultaneously. ISAR imaging signal model with 2D off-grid error compensation is first established. Subsequently, within the Bayesian framework, the 2D LaAPC prior is proposed to capture structural features. The model variables are efficiently inferred using joint generalized approximate message passing and variational Bayesian inference. The computational efficiency of inference process is further enhanced by leveraging the Kronecker structure of the observation matrix. Experimental results of simulated and measured data indicate that the proposed algorithm effectively preserves the structural features of ISAR images and mitigates the 2D grid mismatch error, even under the condition of incomplete data, while maintaining high computational efficiency.