A flash flood detected area using classification-based image processing for sentinel-2 satellites data: A case study of Zafaraana Road at Red Sea

Abstract

Natural crises manifested in floods and droughts are counted as the most severe impacts of climate change on the world. In this regard, flash floods are the most common cause of economic and human losses worldwide. However, the present study focuses on the flash flood-affected area between Zafaarana and Ras Ghareb coastal roads. Sentinel-2 satellite images of recent years before and after the flash flood have been utilized to detect flooded areas and investigate their environmental conditions.

Initially, the captured images were pre-processed to compare the environmental conditions before and after flooding. Consequently, the Normalized Difference Water Index (NDWI) was utilized to classify water bodies in different bands. Finally, an image difference feature (IDF) model with computation of per-pixel features, merging image disparities, and calculation of the characteristic value phases was constructed to extract various image differences after photo processing, that's to identify flooded pixels in the images and assess their performance in the proposed model. The proposed IDF model was compared by rating each model on the same test set, while changing the training set. In conclusion, the proposed algorithm shows an accuracy of 98.9%, which is a better flood image processing technique than other methods. The insights from this research will help decision makers in structuring their rescue strategies and evacuation maps during and before the environmental crisis.