Evaluation of the optimal number of clusters for unsupervised flood mapping using Interferometric Synthetic Aperture RADAR data

In this paper, we are dealing with image clustering in regard to the flooding extent delineation using Synthetic Aperture RADAR (SAR) and Interferometric SAR (InSAR) data. Even though we focus on flood mapping, it is not necessarily correct to consider the data division into two clusters (flooded and not flooded regions). In the context of unsupervised classification, the selection of optimal clusters number is a crucial task that affects the understanding of the clustering result. Therefore, the main objective of this work is to specify the required number of clusters needed in order to get an accurate flood map using the improved FCM approach that takes into account the InSAR coherence information. Indeed, we are solving this cluster analysis problem using fuzzy internal validity criteria namely Partition Coefficient and Partition Entropy indices. Lastly, we present experimental results concerning the Mallegue river flooding event that happened in 2005 in the North-West of Tunisia.