Landmines detection using finite-difference time-domain and artificial neural networks

In this paper, the performances of the neural-network approach for discrimination of the landmine depth from the ground surface and the landmine radius are carried out. A sensor for landmines detection consists of two microstrip antennas is used. One of the antennas is used as transmitting antenna and the second is considered as receiving antenna. Circular cylindrical shape metallic landmine is used. The neural-network process data are obtained from the FDTD formulation of the electromagnetic scattering. The inputs in the input layer of the neural network are the magnitude of the mutual coupling between the microstrip antennas, |S21| while the outputs are landmine depth from the ground surface and landmine radius. Good agreement with exact profile has been observed. The computation time and computer memory of the inverse problem are considerably reduced.