Electromagnetic imaging for buried conductors in the slab medium by direct sampling method and U-Net

This paper presents convolutional neural network (CNN) with deep learning methods to reconstruct electromagnetic imaging of buried conductors in slab medium. First, we emit transverse magnetic (TM) waves to illuminate the buried conductors in the slab medium. Direct sampling method (DSM) is used to estimate the image by the measured scattered field. The estimated image is then inputted to CNN for accurate electromagnetic reconstruction. We analyze the reconstruction performance of different conductor shapes in the noise environment. Numerical results show that our proposed method is capable to reconstruct good images for conductors buried in the slab medium. In conclusion, in addition to simple shapes such as spherical and elliptical buried conductors, edge details of other irregular shapes can also be well reconstructed.