This paper presents a new approach for classifying radar targets through the fractal analsis of their radar echo. Radar target echo and clutter models have been generated from Rayleigh and Weibof distributions. oth target and clutter signatures have shown fractional Brownian motion behavior. The radar echo of a target varies basically according to radar target cross section (RCS) and more specifically according to the size. and geometric shape of the target. Those variations have been efficiently captured and abstracted in terms of average holder constant. Radar target echo and clutter have also been transformed into invariant symmetrized dot pattern. (SDP) plot, where a correlation coefficient factor . R. has been computed. The two features, average holder constant and R have been presented to a multi-resolution neural network to classify seven types of aircraft in the presence of clutter. The multi-resolution neural net is composed of three sub-nets. each sub-net is a three-layered neural net with a back propagatin learning algorithm. Conclusive classification results have been obtained and analyzed in terms of confusion matrix format.