Subsurface object recognition in a soil using UWB irradiation by Bow-Tie antenna and artificial neural network

Background: Subsurface radiolocation problems have an important place in the modern world, such as in geology, building, and humanitarian demining. A complex problem that impedes the widespread use of subsurface radars is the processing and interpretation of the parameters of the reflected electromagnetic field. Objectives: The main purpose of this work is to solve the problem of recognition of objects buried in a soil by bow-tie antenna and artificial neural network (ANN). Materials and methods: The problem of recognition an ideally conducting cylindrical object that is situated below the earth's surface is solved by an ANN. The air-ground interface is irradiated by a bow-tie antenna, which is excited by means of a nanosecond impulse current. The irradiation by nearly point-like source in contrast to plane transient electromagnetic wave incidence considered in our previous works is characterized by the significant decrease of field energy reached a hidden object, reflected, and received by antenna. Moreover, the descent of the field energy becomes more sensible proportionally to the distance from the object to the radar. The complications can call into question the possibility the application of the approach on the base of ANN. The electromagnetic problem is solved numerically by using the FDTD method. The time dependences of amplitudes of differently polarized electric field components, which were obtained in four points above the earth's surface were used as the initial data. The points form the shape of a square. The ANN was trained by the obtained data to determine the position of the object beneath the ground. Results: ANN recognition quality was tested by test data with the addition of Gaussian noise and data obtained when the receiving system is moved relative to the object by shift of the value that was absent in training set. Conclusion: Such type of antenna system in combination with the ANN shows good results for determining the distance to the object even in the presence of noise.