A Novel Physics‐Informed Dynamic Student’s t ‐Mixture Model for Aerial Targets RCS Characteristics

ABSTRACT

The statistical characterisation of a target’s radar cross section (RCS) is notoriously challenging due to its complex, angle‐dependent nature, often manifesting as multimodal and heavy‐tailed distributions contaminated by specular glints. Conventional statistical models, including Gamma, Weibull, Log‐normal distribution (LND) and mixed LND (MLND) models, are often inadequate as they lack robustness to outliers and are based on a physically inconsistent assumption of static mixture weights. To address these deficiencies, this paper introduces a physics‐informed dynamic Student's t ‐mixture model (PI‐DS‐tMM). Our framework makes a dual contribution: it employs the heavy‐tailed Student’s t ‐distribution for inherent robustness against outliers, and, critically, it replaces static weights with a dynamic, physics‐informed weighting function that explicitly models the angle‐dependent activation of scattering mechanisms. Extensive validation on both simulated and measured RCS data from diverse aerial targets demonstrates that the PI‐DS‐tMM achieves a superior goodness‐of‐fit over conventional models, as quantified by the Kolmogorov–Smirnov test. The proposed model provides a more accurate and physically consistent representation of RCS fluctuations, offering a principled framework poised to enhance the performance of advanced radar applications such as target detection, tracking and recognition.