Landmine detection and classification using UWB antenna system and ANN analysis

The use of an artificial neural network to detect landmines by ultrawideband radar with special antenna system is presented in the work. The system emits a short impulse electromagnetic wave that irradiates a ground that contains hidden objects. The field reflected from the objects and ground surface is received by four antennas of different polarization orientations. The signals from the antenna outputs are transformed to discrete form with constant time step. The data is normalized on the root of the signal energy and used to form six stitched differential signals that are input ones for the first layer of an artificial neural network. It is trained to recognize the type of a landmine or other hidden object and its distance to antenna system. Influence of an additive noise in the input signals on the detection and recognition is investigated.