High Quality Imaging of Automotive Imaging Radar with Co-prime Sampling

In millimeter-wave (mmWave) automotive radar, physical measurements acquired by conventional uniform sampling are often limited by available resources. Sparse sampling can achieve high-quality imaging through sparse reconstruction with limited measurements. However, it often requires cumbersome sampling strategy designs, and the actual manufactured hardware may not meet these requirements well. Therefore, a simpler and easier-to-implement sampling strategy, with less sampling but still structured, along with a corresponding high-quality imaging method, should be explored. In this work, we introduce and explore the technique of co-prime sampling for high-quality imaging of automotive radar. We explore co-prime sensing to obtain more effective samples while reducing the number of physical sampling channels in a structured design. Specifically, we use the second-order statistics of the measured data, namely the covariance matrix, as an alternative for measurements to increase effective measurements. Through this process, we also establish the corresponding imaging forward measurement model of the alternative measurements, and then construct the scene’s sparsity-regularized imaging model. We adopt the first-order proximal gradient algorithm to design the solving method of the imaging model, where background information is specifically reserved for better scene image interpretation while target information is enhanced. Finally, through experimental verification with several sets of measured data, it is proven that high-quality imaging results are obtained with co-prime sampling and the corresponding imaging methods. This paves the way for automotive imaging radar sensing with a less but structured sampling strategy.