Exploiting different neural-network-based algorithms for the detection and the characterization of cylindrical buried objects