Research on electromagnetic scattering parameter inversion technology of ground feature unit

In this paper, we propose a surface parameter inversion method based on convolutional neural network (CNN) for microwave images. Using the simulated backscattering data generated by the integral equation method(IEM),microwave images are generated by the fast Fourier algorithm and Stolt interpolation method to provide training sets for deep CNN networks. The results show that the error of root mean square height and correlation length is 1.799% and 2.26%.