PIERS ONLINE, VOL. 6, NO. 5, 2010 425 An Electromagnetic Target Classification Method for the Target Sets with Alien Target: Application to Small-scale Aircraft Targets M. Secmen and G. Turhan-Sayan Department of Electrical and Electronics Engineering, Middle East Technical University, Ankara, Turkey Abstract— The electromagnetic target classification is a challenging problem since the scat- tered field from a target is highly dependent on operating frequency, polarization and aspect angle. In order to minimize adverse effects of these dependencies an intelligent classifier con- taining some distinguishable target features is needed. In addition, in order to be suitable for real target applications, the properties of operating with moderate frequency bandwidth and discriminating an alien target from a target set containing friend targets are important. In this study, an electromagnetic target classification method for isolated targets using noisy data in the classifier design to obtain high accuracy performance in a wider SNR range and having the ability of discrimination of an alien target without any priori information is introduced. The proposed method is mainly based on a late-time resonance region target classification technique, which was reported recently to use the multiple signal classification (MUSIC) algorithm and natural-resonance mechanism modeled by singularity expansion method (SEM) for target fea- ture extraction, and modified for target sets containing alien target(s). The proposed classifier design method is demonstrated and tested for a target set of five friend and one alien small-scale aircraft targets. According to the test result, the proposed method gives high accuracy rates for this target set. 1. INTRODUCTION The scattering from an electromagnetic target is a complex mechanism due to the fact that the scattered fields are strongly dependent on operation frequency, polarization (both transmitter and receiver) and aspect angle (azimuth and elevation). Besides, the additional environmental noise makes the electromagnetic target classification problem more complicated. Therefore, in order to diminish the dependencies of all these mentioned facts an intelligent classifier distinguishing test targets with some specially designed features is needed. Besides, the criteria of the accuracy, satisfactory noise performance, high decision speed in real time, small memory requirements and simplicity should be satisfied. In addition, in order to be suitable for real target applications, the properties of operating with moderate frequency bandwidth and discriminating alien targets from friend targets are important. A finite size electromagnetic target has three different scattering regions with respect to op- erating frequency: Rayleigh region, resonance region and optical region. The resonance region of a target (target dimensions are comparable to λ) in which the proposed method in this paper also works has stronger operation frequency but slighter aspect angle dependency with respect to radar cross section change that about 15–20 degrees changes in aspect angle is needed to observe moderate changes in scattered fields for most of the targets in this region. The resonance region methods mainly base on the Singularity Expansion Method (SEM) for- mulated by C. E. Baum [1]. According to this method, the scattered frequency response of an electromagnetic target is represented as a sum of an entire function and a meromorphic function resulting in a time limited function plus superposition of damped sinusoidals in the time response such that H (s, Ω) = G(s, Ω) + ∞  n=1 R n (s, Ω) (s - s n )(s - s ∗ n ) , h(t, Ω) = g(t, Ω) + ∞  n=1 b n (Ω)e αnt cos(w n t + θ n ) (1) Here, time limited function g(t, Ω) is defined as forced response due to the interaction of the incident wave with the target and it continues until the wave fully passes the target. Afterwards, only the infinite summation part exists (the late-time response). This time limited function has no standard form as the summation part and its magnitude is stronger as compared to the summation