Improving urban impervious surface classification by combining Landsat and PolSAR images: A case study in Kitchener-Waterloo, Ontario, Canada

Urban impervious surface mapping using moderate-resolution optical images such as Landsat images could be challenging due to the complexity of urban land cover. The study aims to combine optical and PolSAR images to improve accuracy of impervious surface classification. A scene of Landsat-5 TM image and a scene of RADARSAT-2 full-polarized imagery of Kitchener-Waterloo were used. The classification accuracies of Landsat image with the combination of different polarizations were compared. The results demonstrated the improvement of impervious surface classification with the combination of RADARSAT-2 PolSAR imagery with Landsat imagery. The major improvement was distinguishing between dark and bright impervious surface. In addition, generally more polarizations generated better results, and HV had the most contributions compared to the rest three polarizations. The results of the study may serve as a reference for further application for combining PolSAR and optical images.