Inversion of Target Electromagnetic Parameters Based on Convolutional Neural Networks

This paper presents a study on the inversion of relative permittivity and geometric dimensions of complex targets based on deep learning convolutional neural networks. Taking the B2 bomber as the target object, the raw data is derived from the FEKO software, and bistatic RCS data is simulated using the Physical Optics (PO) algorithm. This simulated data serves as the input to the convolutional neural network, with the corresponding output being the relative permittivity and size of the target. Finally, the network's output results are compared with those of the test set. The results show that the fitting coefficients (R^2) of relative permittivity and size reach 0.975 and 0.941, respectively, indicating the accuracy of convolutional neural networks in handling nonlinear regression problems.