Conventional compressive sensing (CS) imaging methods rely on the assumption that all scatterers in the imaging scene are located exactly on the pre-defined grids.However

since the scatterers are distributed in a continuous scene

the off-grid problem inevitably exists

which makes basis mismatch between echo measurement and the assumed sensing matrix

and leads to considerable performance degradation by CS-based methods.Therefore

this paper investigates the sparse imaging for off-grid target in frequency diverse multiple-input multiple-output (FD-MIMO) radar.A sparse autofocus imaging method based on Bayesian compressive sensing (SAF-BCS) is proposed.It employs the technique of variational Bayesian inference to achieve the imaging of off-grid scatterres in light of the criterion of maximum a posteriori (MAP).Compared with the conventional sparse recovery algorithms

the proposed method adequately utilizing the prior information of the target

is able to automatically tune parameters

and thus can provide a better capability to correct the off-grid errors

and to estimate the noise power

etc.Simulation results confirm that SAF-BCS is not sensitive to grid discretization

and has a robust imaging performance.