Polarimetric Radar Target Decomposition and Classification

The polarization transformation properties of radar target characterize the radar scattering phenomena according to the theory introduced by E. Kennaugh, J. R. Huynen and S. R. Cloude. For this reason airborne and spaceborne polarimetric SAR are currently emergent technologies useful for providing maximum available information for application on remote sensing. The measurable quantities in the linear polarization basis have strong variability as function of the target aspect angle given the presence of strong disturbance effects like multiplicative noise. For this reason and given the difficult interpretation of the scattering coefficients in linear polarization basis, many authors have proposed several target decomposition theorems in order to give a likelihood interpretation of the observed phenomena. Target decomposition theorems have shown the basis for developing target classification and remote sensing inversion studies. In this Philosophy Doctoral dissertation the Einstein's photon circular polarization special unitary basis is used extensively in order to develop several lossless and sufficient target decomposition theorems providing orientation invariant parameters stressing the proper number of degrees of freedom of each case. Respectively the five novel target decomposition theorems proposed in this Dissertation are proven useful for: a) extracting the characteristic parameters of a coherent field, b) characterizing the statistics of a random field, c) decomposing the main features of a reciprocal deterministic target vector, d) modeling the degrees of freedom of a random reciprocal target, e) assessing the features of circular polarization dual coherent radar. The quantum theory of radar target scattering has been introduced by analyzing qualitatively the photon spin transformation properties of some elemental targets. Two Unsupervised Classification schemes based on the inner Hermitian product have been proposed generalizing the Cameron's approach to not symmetric and random targets. A Supervised Classification scheme based on the Cloude-Pottier eigen-features has been proposed for the identification of man-made target. A relationship of equivalence for the estimation of the coherency matrices of random target is also proved. Results have been validated via an extensive use of airborne and spaceborne fully polarimetric, simulated dual coherent radar and some anechoic chamber data sets. The new parameters proposed in this Ph. D. dissertation and the wide number of classes proposed are useful for assessing the advantages of fully polarimetric system versus dual coherent radar radiating circular polarization.