Processing polarimetric SAR time series over urban areas with binary partition trees

Abstract

This paper addresses the processing of temporal series of Polarimetric Synthetic Aperture Radar (PolSAR) data for urban areas employing the Binary Partition Tree (BPT) representation. Instead of working directly with the original pixels of the data, a hierarchical region-based representation is employed. Thanks to the multi-scale properties of the BPT, the data may be decomposed into a set of homogeneous regions corresponding to the different observed temporal evolutions, while preserving the contours and small details of the scene. Then, this information is employed in order to detect and interpret the changes over the urban areas. The technique is evaluated with RADARSAT-2 PolSAR time series data over Barcelona, Spain.