Radar target classification using RCS characteristics

This study investigates the feasibility of classifying anti-tank threats, specifically Rocket-Propelled Grenades (RPGs) and Anti-Tank Guided Missiles (ATGMs), based only on their Radar Cross Section (RCS) characteristics. A comprehensive target RCS database was constructed using electromagnetic simulations, and radar tracking scenarios were simulated to generate dynamic RCS measurements under varying Signal-to-Noise Ratio (SNR) levels and carrier frequencies of 10, 20, and 40 GHz. Two primary classification approaches are evaluated: a correlation-based Fitting Method and an Artificial Intelligence (AI) Method based on a Multi-Layer Perceptron (MLP) neural network. The results show that the Fitting Method achieves high classification accuracy, exceeding 90%, for non-rolling ATGM targets at 10 and 20 GHz. In contrast, the AI Method demonstrates superior performance in classifying rolling RPG targets by effectively capturing their spectral roll characteristics, particularly at SNR levels above 30 dB. At the higher carrier frequency of 40 GHz, the classification performance of both methods deteriorates, probably due to increased sensitivity to phase variations among multiple scattering centers, resulting in more rapid interference patterns in the RCS signature. The study concludes that an effective radar-based classification architecture better  employs a hybrid framework, combining the AI Method for rolling targets, such as RPGs and with the Fitting Method for non-rolling targets, such as ATGMs.

This work is towards an M.Sc. degree under the supervision of Prof. Gil Yudilevitch, Aerospace Engineering, Technion. Co-Advisor: Dr. Dotan Algranati.