Target recognition by self-organizing map (SOM) type unsupervised clustering using electromagnetic scattered signals in resonance region

This paper investigates the use of unsupervised learning for electromagnetic target recognition in resonance region. Wigner distribution based target features extracted from late-time target responses at arbitrarily observed aspect angles are used to design a target classifier by using the self-organized map (SOM) algorithm. Effects of having unequal amounts of training data for different library targets are investigated in particular. Small scale aircraft modeled by conducting wires are used as test targets in demonstrations.