Signature-based target identification and pattern recognition

In the identification of a particular target (or target class), out of a library of possible targets, one can use the scattering signatures from a transient (broadband) type of radar. These signatures are parameter sets (using aspect-independent and/or aspect-dependent parameters) in some scattering model. There are various such models of interest, based on the symmetries (or partial symmetries) of the various target features (geometric shapes, including constitutive parameters), of both local and global varieties. Such symmetries have significance for the signal processing as well, so that the signatures can be better extracted from the noise and clutter background. Based on such symmetries, one can organize the target features, models, signatures, and signal processing into the habitats of a zoo. Part 1 of this article does this. In Part 2, target signatures and pattern recognition are considered. Pattern recognition is a well-established set of techniques for categorizing objects by data features. In the context of radar-target identification, one can use parameters in electromagnetic-scattering models and sets of the same (signatures) for such features. This leads to a target-identification scheme, in which a target in all its aspects to a radar is treated as a pattern class, where a pattern is a set of data features.<>