Discrete Tomography Approach for Subsurface Object Detection by Artificial Neural Network

The problem of the underground object detection by short impulse electromagnetic wave is presented in this work. The plane electromagnetic wave is incident on the boundary between air and model of the ground normally. The electromagnetic problem of the wave propagation and its reflection on subsurface objects is solved numerically by FDTD method. The time dependences of the reflected wave received under the boundary are analyzed to detect subsurface objects. For this purpose the artificial neural network (ANN) uses the signals received in points under the boundary at fixed height. Time dependencies of received electromagnetic field is discretized with a constant time step. Additional information for the ANN is obtained by time-spatial processing that based on discrete tomography approach. The set of points presented the received time dependences is multiplied on pre-calculated time-spatial attenuation matrix. The matrix is formed on ray tracing method, antenna pattern, wave attenuation and time delays of wave in media. Underground spatial points serve as a secondary source of electromagnetic field. The work of the ANN is verified on testing data that correspond to intermediate positions of a hidden object.