Scatterer identification using neural networks

The authors illustrate how Kohonen self-organizing neural networks have been used to classify different types of radar scatters which make up complex targets. The goal was to measure the target´s radar cross section (RCS) under conditions to which distinct scatterer types respond differently. Traditional clustering algorithms were used to find dense regions in the Kohonen network which represent the examplars of different scatterers. The investigation focused on a small, metal coated and faceted target. This network was trained and tested using computer-generated RCS data for the target at three different aspect angles, each color coded by scatter type. It has been shown that differences between leading and trailing edge and tip scattering can be distinguished using this approach