Deep Learning-Based Reconstruction for Near-Field MIMO Radar Imaging

Near-field multiple-input multiple-output (MIMO) radar imaging systems are of interest in diverse fields such as medicine, through-wall imaging, and surveillance. The imaging performance of these systems highly depends on the underlying image reconstruction method. While sparsity-based methods offer better image quality than the traditional direct inversion methods, their high computational cost is undesirable in real-time applications. In this paper, we develop a novel deep learning-based reconstruction method for near-field MIMO radar imaging. The main goal is to achieve high image quality with low compu-tational cost. The developed approach has a two-staged structure. The physics-based first stage performs adjoint operation to back project the measurements to the reconstruction space, and DNN-based second stage converts these backprojected measurements to a scene reflectivity image. As DNN, a 3D U-Net is used to jointly exploit range and cross-range correlations. We illustrate the performance of the reconstruction method using a synthetically generated dataset. The results demonstrate the effectiveness of the developed method compared to commonly used analytical approaches in terms of image quality and computation time.