Target classification with multiple frequency illumination

Target classification may be viewed as a crude form of inverse scattering that determines which of
M
prespecified body shapes is most consistent with the scattered field data. A method is described for discriminating between different target shapes by appropriately processing the backscattered waveforms or target signature observed in radar receivers. Although this general subject has received considerable attention in the recent literature, the present paper is among the first to theoretically examine the automatic target classification problem for a specific application with currently operational radars. The method for shape classification uses scattered returns from two or more collocated sources operating at different frequencies and polarizations. The basic technique consists of appropriately comparing the received data with the contents of a stored signature library that has been obtained a priori by either calculations or measurements. The technique is demonstrated for the case of radar space object identification which often involves very noisy observations of the target signature. The target classification algorithm is synthesized from decision theoretic principles, which, coupled with a rigorous technique for target response calculation, provides for reliable classification of simple space objects during their first pass in view of the sensor. The classification technique requires no a priori information concerning target motion and estimates the aspect angle automatically. It operates sequentially on incoming data, gives an improved estimate of target class after each additional data point, and provides a confidence measure associated with the current decision. Finally, the classifier determines if the target is not in the signature library by means of its target rejection capability.