Automatic recognition of aircraft by means of radar signals is the subject of this dissertation. In particular, the problem of utilizing polarimetric information from the monostatic, wideband, scattering matrix to target classification and analysis is considered. The approach taken is to effect a decomposition of the target in terms of its predominant scattering centers and their polarimetric properties. This polarimetric target model is based on a new formalism for the description of the vector behavior of wideband electromagnetic planewaves, called transient polarization.; The transient polarization response of a radar object is a 3-dimensional, time-dependent electric field locus. Conceptually, this signature can be envisioned as the result of transmitting a short "circularly polarized" pulse toward the object. The incident pulse illuminates each of the object scattering structures with all possible orientations of the electric field vector. The object substructures scatter electromagnetic energy with a transient polarization state which is characteristic of the substructure geometry.; Transient polarization states are expressed as a 3-tuple of time-dependent polarimetric parameters {dollar}{lcub}A(t),epsilon(t),tau(t){rcub}{dollar}. These are respectively, amplitude, ellipticity, and tilt. In the limit of zero bandwidth, that is, in the continuous wave case, these parameters describe an ellipse in the plane of the wavefront. Scattering centers are identified by the peaks in the amplitude. The polarimetric parameters of canonical targets, simplified aircraft target combinations, and scaled model commercial airliners are analyzed.; The performance of radar target identification systems employing polarimetric features is evaluated by means of Monte-carlo simulation studies. Both decision-theoretic and language-theoretic techniques are used for this purpose. In the former case, distance measures in polarization space are employed in the non-parametric nearest neighbor algorithm. In the latter case, the polarization space is partitioned and symbols are assigned to the various polarimetric regions. A target is represented as a string of polarimetrically-related symbols, and is classified using syntactic methods.; Target classification simulations show that polarimetric pattern representations extracted from transient polarization responses may be used for reliable and flexible target classification. In addition, the polarimetric techniques developed in this study may be used as a tool for concise characterization and analysis of time-domain electromagnetic scattering.
展开▼