Advanced Physical Optics-inspired Support Vector Regression for Efficient Modeling of Target RCS

This paper proposes an advanced physical optics-inspired support vector regression (APOI-SVR) for efficiently modeling the radar cross section (RCS) of conducting targets. Specifically, an improved physical optics-inspired kernel function is newly proposed by introducing two angular frequency parameters, thereby enhancing the capability of characterizing the various fluctuation patterns in RCS with respect to observation angles. Furthermore, considering the critical role of data preprocessing in facilitating the model’s ability to learn the underlying RCS patterns accurately, a physics-based data preprocessing method is introduced. Numerical validations based on two exemplary targets demonstrate that APOI-SVR effectively reduces the predictive root mean square error (RMSE) by over 24.7% compared with the benchmark model. Afterward, APOI-SVR is adopted to quickly establish the RCS feature map of an aircraft model, the results show that it is comparable to numerical simulations in accuracy but less than one-tenth in time cost, indicating the practicality of APOI-SVR for efficiently analyzing the RCS characteristics of targets.