Near-Field Low-Storage SAR Imaging With Spatio-Temporal Sub-Nyquist Samples

Range migration algorithm (RMA) is widely used in millimeter-wave (MMW) near-field synthetic aperture radar (SAR) imaging. In comparison to imaging algorithms based on compressed sensing (CS) or neural network, RMA emerges as a more practical solution due to its lower computational complexity and higher imaging efficiency. However, achieving high-resolution images through RMA often requires equally spaced and dense sampling which may inevitably lead to huge storage. Furthermore, RMA dictates that adjacent sampling locations in both the horizontal and vertical directions adhere to strict equidistance, resulting in practical inflexibility and inconvenience. To overcome these problems, an enhanced RMA based on undersampling in the time and space domains is proposed. To compensate for precision loss caused by temporal undersampling, a parameter estimation method based on the maximum likelihood criterion is employed. Additionally, a matrix completion algorithm (MCA) based on sparse Bayesian learning (SBL) is also introduced to recover the missing cells due to spatial sparse sampling. Simulation results and application cases demonstrate the effectiveness of proposed algorithm. By employing the enhanced RMA, the required imaging data can be diminished to approximately 10% compared to traditional RMA.