Signal Processing in UWB Subsurface Radiolocation by Artificial Neural Networks

The signal processing problem for identification of underground objects is considered. The source of signals is the ultra short impulse electromagnetic waves reflected from air-ground interface and underground objects and received by subsurface radar antenna system. The source of the irradiated ultrawideband field is the plane electromagnetic wave with Gaussian time dependence that is incident normally to the air-ground interface. The numerical simulation of the electromagnetic problem to find the electrical field strength in receiving points above the ground at some fixed height is carried out by Finite Difference Time Domain method The received signals are processed by artificial neural networks for the object identification and classification of its position. The electrically small metal cylinder is used as an example of an object for the classification and recognition. The influence of the structure of an artificial neural network and its training peculiarities on the precision of the classification is investigated.