Modified Four-Component Scattering Power Decomposition of PolSAR Data by Using Rotated Dihedral Component

Polarimetric Synthetic Aperture Radar (PolSAR) is a radar that can identify and classify targets using the polarization scattering characteristics. The model-based scattering power decomposition, which is one of the PolSAR data analysis methods, estimates rough land cover conditions based on the ratio of fundamental polarimetric scattering components contained in the observation data. However, there is a problem that urban areas where building walls are not aligned parallel to the observation trajectory are misclassified as natural vegetation or forest. Although modified methods that applies polarimetric basis rotation to compensate the effect of the oblique angle of buildings in such urban areas has been proposed, the problem still remains. The cause of the misclassification is thought to be a mismatch between the scattering model and the actual scattering phenomenon, or bias in the polarimetric basis rotation for angle compensation of the urban area. In this paper, we propose a modified scattering model that takes into account the oblique angles of urban areas directly and evaluate the validity of the model theoretically. Furthermore, error in the orientation angle estimation is discussed by numerical examples and experimental results. The ALOS-2 dataset that is L-band fully polarimetric dataset is used in the experiment.