Transductive Radar Clutter Segmentation Network for Ionospheric Inhomogeneity Clutter Suppression

The ionosphere turbulent motions, detected by radar systems, result in a complex form of ionospheric inhomogeneity clutter. In this paper, we introduce transductive radar clutter segmentation network (TRCS-Net), a deep unfolding network architecture for joint suppression of ionospheric clutter. The proposed model utilizes transductive transfer learning to adapt knowledge from a source domain to specific target domain conditions of ionospheric clutter. A deep unfolding approach maps the iterative optimization procedure onto a neural network, jointly optimizing across multiple traditional clutter suppression methods. The training dataset uniquely fuses measured radar data with simulations characterizing the turbulent motion mechanisms that generate ionospheric irregularities. Experiments demonstrate the effectiveness of TRCS-Net in improving clutter suppression performance compared to individual algorithms, leveraging transductive transfer learning and simulation data.