Unsupervised Land Cover Classification Using H//TP Space Applied to POLSAR Image Analysis

Koji KIMURA  Yoshio YAMAGUCHI  Hiroyoshi YAMADA  

IEICE TRANSACTIONS on Communications   Vol.E87-B   No.6   pp.1639-1647
Publication Date: 2004/06/01
Online ISSN: 
Print ISSN: 0916-8516
Type of Manuscript: PAPER
Category: Sensing
total power,  anisotropy,  polarimetric entropy,  alphabar,  Wishart distribution,  iterative ML method,  

Full Text: PDF(4.2MB)>>
Buy this Article

This paper takes full advantage of polarimetric scattering parameters and total power to classify polarimetric SAR image data. The parameters employed here are total power, polarimetric entropy, and averaged alpha angle (alphabar). Since these parameters are independent each other and represent all the scattering characteristics, they seem to be one of the best combinations to classify Polarimetric Synthetic Aperture Radar (POLSAR) images. Using unsupervised classification scheme with iterative Maximum Likelihood classifier, it is possible to decompose multi-look averaged coherency matrix with complex Wishart distribution effectively. The classification results are shown using Pi-SAR image data set comparing with other representative methods.