Towards Unsupervised Flood Mapping Generation Using Automatic Thresholding and Classification Aproaches

This paper presents an unsupervised flood monitoring approach developed by DARES TECHNOLOGY using Synthetic Aperture Radar (SAR) space-borne images in order to improve disaster management and coordinate response activities in front of flood crisis events. The methodology developed combines SAR pre-processing, histogram thresholding in blocks with bi-modal behavior and unsupervised segmentation approaches. The use of interferometric parameters, such as the coherence is also employed to refine results. The challenge of this detection approach is working towards an unsupervised approach, i.e., training data is no required and, therefore, information or ground-truth about the class statistics within the area of interest are not necessary. The methodology proposed will be validated over the floods occurred in Mumbai city on August 29, 2017 produced by heavy rain events. 4 Sentinel-1 SAR images will be employed for this purpose.