Flood mapping in vegetated areas using an unsupervised clustering approach on Sentinel-1 and-2 imagery

摘要: The European Space Agency's Sentinel-1 constellation provides timely and freely available dual-polarized C-band Synthetic Aperture Radar (SAR) imagery. launch of these other SAR sensors has boosted the field SAR-based flood mapping. However, mapping in vegetated areas remains a topic under investigation, as backscatter is result complex mixture backscattering mechanisms strongly depends on wave vegetation characteristics. In this paper, we present an unsupervised object-based clustering framework capable flooding presence absence flooded based globally data only. Based image pair, region interest segmented into objects, which are converted to SAR-optical feature space clustered using K-means. These clusters then classified automatically determined thresholds, resulting classification refined by means several growing post-processing steps. final outcome discriminates between dry land, permanent water, open flooding, vegetation. Forested areas, might hide indicated well. presented four case studies, two contain For optimal parameter combination, three-class F1 scores 0.76 0.91 obtained depending case, pixel- thresholding benchmarks outperformed. Furthermore, allows easy integration additional sources when become available.