Polarimetric scattering analysis for detecting largely-oriented man-made objects based on eigenvalues/eigenvectors analysis to the rotated coherency matrix

This paper proposes a simple algorithm for detecting largely-oriented man-made object by using angular index β, in addition to the conventional angular index a. The angular indices are derived by the eigenvalues/eigenvectors analysis to the rotated coherency matrix of fully polarimetric SAR data. Here the orientation angle compensation (OAC) is carried out to the coherency matrix before doing the eigenvalues/eigenvectors analysis. In the proposed algorithm, thresholds Th₁, Th₂ for α, β are included and must be determined for accurate detection of the objects. In order to determine them, we carry out the Finite-Difference Time-Domain (FDTD) polarimetric scattering analysis for relatively large scale man-made objects models. The dependency of the polarimetric scattering feature on the variation of squint angle is in detail investigated. Resultantly, it is found from the FDTD analysis that the candidates of appropriate thresholds for α and β are obtained for largely-oriented man-made objects detection.