Multimodality Swarm Targets Identification Using Polarimetric Rotation Domain Radar Signatures

Technical developments of uncrewed aerial vehicle (UAV) derive swarm formulations of UAVs in various application scenarios. The evolution of swarm modalities elevates radar target signature complexities and proposes higher requirements for radar systems to identify single and swarm modalities in low-altitude airspaces. This article uses radar rotational domain polarization and scattering complexity signatures to compose a feature space to distinguish single and swarm modalities for quadcopter drones and birds. A signature selection strategy is designed based on a subspace projection algorithm to maximize signature differences in the polarimetric rotation domain. The feature space is constructed from radar tracks of dynamic targets. The random forest model studies training datasets and constructs an effective target classifier with feature importance ranking functionality. Quadcopter drone and bird models are developed to simulate their polarization signatures in single and swarm modalities. Validation experiment results indicate that the method could identify targets under different modalities. Relevance studies between swarm complexity and classification accuracy prove the necessity of scattering the entropy signature. A preliminary experiment in identifying UAV composition model with corner reflector loads indicates the method has potential in more refined UAV target identification problems.