Recognition of conductive objects based on the characteristics of reflected electromagnetic wave

The problem of electromagnetic wave diffraction by the metal objects has been solved using integral equation technique. The diagrams of backward scattering have been plotted for four different objects. Based on the feature vector, which has been constructed by applying wavelet packet signal decomposition, a neural network has been trained. We have performed the testing of ability of the neural network to recognize the object depending on the noise level. Various methods of the feature vector forming have been considered.