Summary
International Symposium on Antennas and Propagation
2008
Session Number:1A
Session:
Number:1A-2
Simple Man-made Object Classification Approach Using The Correlation Coefficient in Circular Polarization Basis
Ryoichi Sato, Yoshio Yamaguchi, Hiroyoshi Yamada,
pp.-
Publication Date:2008/10/27
Online ISSN:2188-5079
Summary:
Wide area terrain monitoring or classification is one of the most important applications in Polarimetric Synthetic Aperture Radar (POLSAR) sensing. In particular, man-made object extraction is a subject of great interest. The power decomposition [1], [2] and the polarimetric correlation coefficient [3], [4] are the well-known useful target classification techniques. For complex urban area monitoring, however, it is sometimes difficult to distinguish the man-made objects as buildings or houses from other natural distributed ones, since the polarimetric scattering feature from the man-made objects is strongly dependent on the direction or orientation of their alignment. When the man-made objects are obliquely aligned to the radar illumination direction, the classification accuracy of the above mentioned techniques may become low, since the useful polarimetric indices for classifying the man-made objects (strong doublebounce scattering and large correlation between the circular polarized components) is no longer observed from the target areas.　Recently, an accuracy improvement of the classification for the obliquely aligned man-made objects has been reported by utilizing the modified correlation coefficient [5], [6]. The modified correlation coefficient is obtained by normalizing the conventional correlation coefficient in the circular polarization basis by that under the reflection symmetry condition. So it enhances the return from the non-reflection symmetry areas, in which there exist lots of the oblique man-made objects, and enables us to precisely detect or classify the oblique targets. On the other hand, by utilizing the modified coefficient only, one often encounters an alternative problem that the scattering contributions from the orthogonally aligned man-made objects are underestimated. In this paper, we propose a simple man-made object classification algorithm based on the correlation coefficient approach, to accurately classify the man-made objects not only orthogonally but also obliquely aligned to the radar illumination direction. We here make use of both the conventional and modified correlation coefficients. The total power is utilized in the simple classification algorithm as an additional parameter for choosing the appropriate polarimetric index, i.e. conventional correlation coefficient or modified one? It is verified from the image analysis for L-band Pi-SAR data that the present classification algorithm based on the combination between the correlation coefficients and the total power leads us precisely classify the man-made objects, regardless of the direction of their alignment.