Stricken man-made object detection using scattering power decomposition with NNED and rotation of the covariance matrix

This paper proposes a simple man-made object detection method by utilizing the nonnegative eigenvalue decomposition (NNED) [1] and a unitary rotation [2] to the covariance matrix. It is observed from the result of the PolSAR image analysis using the proposed method that the double-bounce scattering contribution P_d becomes stronger than that by the conventional scattering power decomposition even when the stricken residential houses are built on rough and slope ground surface. Furthermore, in order to analytically verify the improved effect for the stricken man-made object detection, the Finite-Difference Time-Domain (FDTD) polarimetric scattering analysis is also carried out for a simplified man-made objects model, which consists of quad dielectric rectangular parallelepipeds on finite circular plate with rough surface. The rectangular parallelepipeds are considered as a simplified man-made buildings model. It is cleared from the FDTD result that the enhancement of P_d can be realized by doing the C₂₂ rotation, in addition to the NNED procedure.