A Fast 3-D MIMO Imaging Method Based on Khatri–Rao Decomposition and Sparse Bayesian Learning

In multiple-input–multiple-output (MIMO) radar imaging, the resolution of images is strictly constrained by the array aperture. In this article, the problem of 3-D imaging in an MIMO radar system is investigated. To exploit the sparsity of the target in the imaging scene, the sparse Bayesian learning (SBL)-based methods have been established by discretizing the imaging area and designing the dictionary matrix. In this article, different from the present MIMO imaging methods and compressed sensing (CS) methods, the original dictionary matrix is decomposed into three partial dictionary matrices, and a novel sparse reconstruction model for MIMO imaging is established, which can greatly reduce the running time. Then, a novel SBL-based method named Khatri–Rao decomposition and SBL (KRDSBL) is proposed, where the expectation–maximum (EM) method is utilized to estimate the unknown variables, including range, azimuth, and pitch vectors of scattering points and the noise variance. Additionally, the posterior distribution for all unknown variables is derived. Simulation and real data experiments have verified that the proposed algorithm can obtain high-resolution images in a relatively short period of time.