A Bayes classifier for change detection in synthetic aperture radar imagery

Multitemporal synthetic aperture radar (SAR) observations of natural areas permit monitoring of the characteristics and evolution of structural and electrical properties of natural surfaces as a result of meteorological and phenologic cycles. Change detection is one possible way of analyzing multitemporal SAR data which aims at quantifying the relative changes in radar backscatter of the same surface recorded at two different times under the same imaging conditions. A method is proposed for detecting the quantifying eventual changes in radar backscatter and mapping out ensembles of pixels of spatially and radiometrically homogeneous and similar changes using a Bayes classifier. Examples using real multitemporal SAR data are given