Detection of buried dielectric cavities using the finite-difference time-domain method in conjunction with signal processing techniques

In this paper, we address the problem of detecting low-dielectric contrast cavities buried deep in lossy ground by using the finite-difference time-do main (FDTD) method in conjunction with signal processing techniques for ext rapolation and object identification. It is well known that very low freque ncy probing is needed for deep penetration into the lossy ground, owing to a rapid decay of electromagnetic (EM) waves at higher frequencies. It is al so recognized that numerical modeling using the FDTD method becomes very di fficult, if not impossible, when the operating frequency becomes as low as 1 Hz. To circumvent this difficulty, we propose a hybrid approach in this p aper that combines the FDTD method with signal processing techniques, e.g., rational function approximation and neural networks (NNs). Apart from the forward problem of modeling buried cavities, we also study the inverse scat tering problem-that of estimating the depth of a buried object from the mea sured field values at the surface of the earth or above. Numerical results for a buried prism are given to illustrate the application of the proposed technique.