Distorted Born Iterative Method Combined with U-net Networks for Close-range Microwave Imaging

In recent years, microwave imaging technology has been widely used in nondestructive testing, medical imaging and target identification. Aiming at the problem of low resolution of traditional close-range microwave imaging methods, this paper proposes a Distorted Born Iterative method (DBIM) combined with U-net neural networks. The method first utilizes DBIM to initially construct unknown scatterers. Then, the preliminary permittivity approximation solution of DBIM is as input to train the U-net networks. The feature sampling capability of U-net is utilized to construct the unknown scatterer again. Finally, the proposed algorithm is applied to "Austria", MNIST database and Cylindrical tests. Simulation results show that the Root-mean-square error (RMSE) of the method is decreased by about 15%-30% and the structural similarity (SSIM) is improved by about 15%-25% compared with the traditional algorithm. DBIM U-net algorithm offers advantages in imaging accuracy and has some potential applications in close-range microwave imaging.