Change detection in urban areas of high-resolution polarization SAR images using heterogeneous clutter models

Based on the clutter statistical characteristics of fully polarimetric SAR image, this paper presents a novel method for SAR change detection based on the heterogeneous clutter model. We use spherically invariant random vectors (SIRV) distribution model to fit the urban areas of full polarimetric images. Then, the degree of evolution between the statistical characteristics of multi temporal full SAR images is measured by the polarimetric likelihood ratio test (PLRT) model. Afterwards, Minimum error method (KI) threshold segmentation is applied to extract the `real´ changed area for change detection. Two RadarSat-2 fully polarimetric images for Suzhou city in China, acquired on April 9, 2009 and June 15, 2010 separately, are used for our experiment. Compared with the detection results based the Wishart distribution model, it is shown that the heterogeneous clutter not only fits urban areas better, but also can avoid large number of false alarms.