Abstract:

A new distance measure derived from the polarimetric coherency matrix was proposed on the basis of classical Wishart distance originally developed for polarimetric SAR image classification. A change detection methodology based on the proposed polarimetric distance measure was then developed. An experiment of change detection using two RADARSAT-2 quad-polarized images acquired in 2009 and 2010 respectively over a suburb area of Beijing was carried out. The change detection capability of the polarimetric distance measure was compared to the log-ratio images formed by the radar intensities of different polarization channels and the total scattered power (the Span). The results showed that the polarimetric distance measure could highlight all types of land cover changes, while only part of change types were visible on intensity-related log-ration images. Moreover, the difference image generated from the polarimetric distance presented highest separability between changed and unchanged pixels. Therefore, the proposed change detection method could be widely applicable for change detection from polarized SAR images.

Key words: Polarimetric Synthetic Aperture Radar, change detection, polarimetric coherency matrix, distance measure