Rapid situational awareness during floods using remote sensing and machine learning

Kurzfassung

Rapid disaster response is critical for saving lives and mitigating the impacts of natural hazards. Traditional approaches for analyzing remote sensing data – including satellite, airborne and UAV imagery – to support situational awareness are often time-consuming and resource-intensive, limiting the speed of emergency response. Recent advances in machine learning offer new opportunities to automate image analysis and enable efficient processing of the rapidly growing volume of remote sensing data while addressing the spatial and temporal dynamics of disaster scenarios.

This talk presents machine learning methods for rapid crisis mapping that have been developed and investigated at the German Aerospace Center (DLR) and its Center for Satellite Based Crisis Information (ZKI). These approaches have been deployed during several national and international flood events to support disaster response operations. We discuss key aspects of data acquisition and present operational results that were used by emergency management agencies during flood disasters. In addition, we showcase experimental results from ongoing research projects, including satellite-based flood monitoring, the detection of exposed assets and the identification of damaged roads and buildings from very high-resolution satellite, airborne and UAV imagery.