RMDU-Net: 3-D MMW Imaging Via Range Migration Based Deep Unfolding Network

Limited by the Nyquist sampling rate, conventional matched filtering-based algorithms always confronts large-scale data dimension, causing difficulties in real-time processing. To address the issue, in this article, a 3-D millimeter-wave (MMW) radar imaging method via range migration (RM) based deep unfolding network is proposed, named as RMDU-Net, which utilizes the compressed sensing (CS) theory and unfolding networks to achieve high-resolution accurate reconstruction under a low sampling rate. The 512×512×6 large-scale imaging results can be obtained by RMDU-Net and the theoretical resolutions are up to 2.8mmx2.8mmx4cm. Besides, the proposed method is compared with traditional range migration algorithm (RMA), and the numerical evaluations are given. The practical experiments demonstrate this imaging approach is excellent in both imaging quality and computing efficiency compared with classical algorithms.