The key problem in any decision-making system is to gather as much information as possible about the object or the phenomenon under study. In the case of the radar targets the frequency et angular information is integrated to form an ISAR image, which has a high information content. A supperresolution technique (ESPRIT 2D) is used in the paper in order to reconstruct the target image. The bidimensional scattering center maps, extracted for their different angular aspects lead to classes of Fourier descriptors, which provide invariance of the considered signatures with respect to the spatial target motion. A new neural network was developed to classify the signature classes obtained in this way for several different targets.