Off-grid radar coincidence imaging based on block sparse Bayesian learning

Radar coincidence imaging (RCI) is a high-resolution and instantaneous imaging technique without the limitation of relative motion between target and radar. In sparse-based RCI, the assumption that the scatterers are located at the pre-discretized grid-cell centers is commonly used. However, the generally existent off-grid degrades the imaging performance considerably. In this paper, the algorithm based on block sparse Bayesian learning (BSBL) framework is developed to solve the off-grid RCI in the range-azimuth space. Applying the Taylor expansion, the unknown true dictionary is approximated to a linear model. Then target reconstruction is reformulated as a block sparse recovery problem. BSBL is then applied to solve the problem by assigning appropriate priors to the coefficients and exploiting the block structure and intra-block correlation. Results of numerical experiments demonstrate that the algorithm can yield superior imaging performance, compared with other block sparse recovery algorithms.