Segmentation and Visualization of Flooded Areas Through Sentinel-1 Images and U-Net

Abstract

Floods are the most common phenomenon and cause the most significant economic and social damage to the population. They are becoming more frequent and dangerous. Consequently, it is necessary to create strategies to intervene effectively in the mitigation and resilience of the affected areas. Different methods and techniques have been developed to mitigate the damage caused by this phenomenon. Satellite programs provide a large amount of data on the earth's surface, and geospatial information processing tools help manage different natural disasters. Likewise, deep learning is an approach capable of forecasting time series that can be applied to satellite images for flood prediction and mapping. This article presents an approach for flood segmentation and visualization using the U-Net architecture and Sentinel-1 synthetic aperture radar (SAR) satellite imagery. The U-Net architecture can capture relevant features in SAR images. The approach comprises various phases, from data loading and preprocessing to flood inference and visualization. For the study, the georeferenced dataset Sen1Floods11 is used to train and validate the model through different epochs and training. A study area in southeastern Mexico that presents frequent floods was chosen. The results demonstrate that the segmentation model achieves high accuracy in detecting flooded areas, with promising metrics regarding loss, precision, and F1-score.

Keywords

• Deep learning (DL) and Sentinel-1
• flood mapping
• flood segmentation
• flood with deep learning
• Sentinel-1
• U-Net and natural disasters