Composite Scattering Evaluation Using Isolate Scattering Data of Target and Rough Surface

Abstract

In many scenarios, it is easy to obtain the background scattering and the target scattering, respectively. However, it is not always simple to acquire the target's composite scattering, which is crucial for target detection. To surmount this obstacle, we use a convolutional neural network (CNN) inspired by a physical mechanism to first reverse the root mean square height (RMSH) and correlation length (CL). The target's composite model above the rough surface is then created using computer graphics and the Monte Carlo method. Finally, the composite scattering of the integrated model is estimated using by shooting and bouncing ray (SBR) method. The proposed method is tested through simulations and measurements, and the results indicate that the prediction error in the HH- and VV-polarization mode is less than 5%.