Scanning Radar Super-Resolution Based on Fast Iterative Shrinkage Thresholding Network

Airborne scanning radar imaging is widely used both in military and civilian fields. However, the azimuth resolution of the imaging system is constrained by the size of the antenna. Iterative optimization super-resolution algorithms based on regularization can be used to overcome this limitation. But these methods demand manual tuning of parameters, which can be laborious. Deep unfolding network is a method of unfolding iterative optimization algorithms to deep learning networks, which combines the interpretability of iterative algorithms and the advantages of deep learning. Considering the excellent performance of the deep unfolding network in other signal processing tasks, this paper introduces the deep unfolding network based on fast iterative shrinkage-thresholding algorithm (FISTA) called FISTA-Net into the field of scanning radar imaging. For the task of scanning radar azimuth superresolution, we improve the model to fully learn the characteristics of radar azimuth data. Simulation results verify the effectiveness of the proposed method.