Dielectric Cylinder Discrimination with ANN Using Wideband Electromagnetic Response

The application of three-component stacked artificial neural networks (ANN) for discrimination of dielectric cylinders of different diameters by their radio images is considered. Neural networks include two sparse autoencoders and the softmax unit. Neural networks are not tied to the frequency range, unlike many well-known methods based on the resonant properties of objects, and they are a powerful tool for object recognition. Radio images were obtained using the method of auxiliary sources (MAS) for cylinders with the radius of 15 to 35 mm. The possibility of successful recognition was confirmed for the case of the diameter deviation of 1 mm and the presence of additive Gaussian noise with SNR of up to 5 dB.