Air Target Classification Using High-Resolution Range Profiles and Deep/Machine Learning Techniques

High-resolution range profiles (HRRPs) offer detailed structural information about air targets, making them crucial for classification and identification in both military and civilian contexts. This paper presents a comprehensive study on air target classification utilizing HRRP data through various deep learning (DL) and machine learning (ML) techniques. The research focuses on classifying five distinct missiles (AGM-65 Maverick, AGM-84 Slam ER, AGM-114 Hellfire, AGM-119 Penguin, AIM-7 Sparrow) and three fighter jet configurations (including Eurofighter, F-22 Raptor, Su-57 Felon, Mig-29 Fulcrum, and T-38 Talon), as well as common aircraft platforms (An-26, Yak-42, and Cessna Citation S/II). The study employs simulated HRRP data generated from 3D CAD models using both C-band (5520 MHz) and X-band (10 GHz) radar parameters with bandwidths ranging from 400 MHz to 600 MHz. The validity of the simulation approach is confirmed by comparing generated HRRP data with published measured data for common aircraft. Experimental results indicate that deep learning classifiers generally outperform hybrid DL-ML methods in terms of average accuracy, with pre-processing steps further improving performance. Detailed analysis highlights the significant impact of target features, radar parameters, noise conditions, and target physical characteristics on classification accuracy, noting that larger targets typically yield better classification results. Classification performance was evaluated using standard metrics such as precision, recall, and the F1-score. The results support the integration of HRRP-based deep learning classifiers for rapid, reliable identification of diverse airborne threats, illustrating their potential as force multipliers in both civil and military radar systems.