The AI4Flood project—AI for Near-Real-Time Satellite-Based Flood Response—aims to implement, train, and comprehensively validate and test various machine learning algorithms for the extraction of flood-affected areas.
Disaster managers frequently request satellite-based crisis information for flood monitoring in order to make targeted use of often limited resources and to prioritize individual activities in the event of a disaster. The AI4Flood project aims to improve existing satellite-based emergency mapping methods using SAR data by testing and validating new machine learning algorithms for the extraction of water bodies during flood events. Particular focus is placed on automating the visual image analysis process through the use of Convolutional Neural Networks (CNNs) for the semantic segmentation of systematically acquired Sentinel-1 SAR data with high spatial and temporal resolution.
