Relative amplitude and phase features for resonance-region radar target identification

A set of features is proposed for radar target identification. The features are derived from multifrequency radar returns in the resonance region. They are related to target geometry, target aspect angle, wave polarization, and frequency, but are independent of other radar-system parameters. A nearest-neighbor algorithm is proposed for target identification that utilizes the feature set. Computer simulation of the algorithm based on measured radar returns of scaled model ships found that with three frequencies the probability of misidentifying a target was 0.1 if the noise level was about one half the standard deviation of the feature magnitudes.<<ETX>>