Integrating Fragmented RCS Data of Complex Targets Using Gaussian Processes
Donghai Xiao, Jingcong Yang, Lixin Guo, Wen Jiang, Tao Hong · 2025
This paper introduces an efficient method for the integration of fragmented RCS data using Gaussian process regression (GPR). By leveraging a physics-inspired covariance function derived from the physical optics (PO) approximation, we establish a surrogate model to characterize the frequency-dependent RCS of complex targets. This method enables precise RCS interpolation between the observed frequency bands and extrapolation beyond them. Validation experiments based on the simulated data of the SLICY model and an aircraft model demonstrate small root mean square errors (RMSEs) for both interpolation and extrapolation, less than 0.8 dBsm. Additionally, the proposed method can easily offer the confidence interval (CI) of the predicted result, thus convenient to quantitatively evaluate its reliability.