Object identification from multi-frequency radar returns

Methods are presented for identifying object classes from a limited number of radar measurements. The radar measurements are the backscattered signal amplitudes at up to twelve harmonically related frequencies in the lower portion of the object´s response spectrum. These amplitudes constitute a set of features upon which classification is based. The maximum and average probabilities of misclassification are computed for various classes represented by test samples corrupted by Gaussian noise. Various methods are derived which reduce classification errors. Results are presented which demonstrate that reliable classification can be obtained in the presence of noise.