Detection of Multitransition Abrupt Changes in Multitemporal SAR Images

This paper addresses the detection of abrupt and step-wise changes in multitemporal sequences of synthetic aperture radar (SAR) images. Specifically, the major motivation is to deal with the change detection of urban features with multitemporal step patterns. In contrast to most techniques presented in the past, that only use couples of adjacent images, in this work multi-SAR images are used for the change detection purpose. The proposed method is appropriate for the generation of new SAR satellite systems that has a high revisit time, generating many successive image of the same area. We attacked on two aspects of the problem: detection of multi-transition instants (TIs) of the SAR amplitude time series and detection of the probable changes. For this purpose, the analysis of variance (ANOVA), technique is used. ANOVA achieves a representation of the relation of all group changes. The features extracted by ANOVA are utilized for the supervised classification procedure. The proposed technique provides promising results for the classification of abrupt changes like in urban areas.