Contribution Of Supervised And Unsupervised Classification In Inundation Mapping: A Case Study

Abstract

Radar images play a major role in mapping flooded areas because of their ability to bypass cloud cover. However, their use presents major challenges linked to the way they are acquired, in particular the phenomena of overlap, foreshortening and radar shadow. The latter, in particular, is a delicate issue because of its potential to be confused with bodies of water. This is also the case for airport runways, as part of this research, a comparative analysis was undertaken using two methods, unsupervised classification coupled with change detection and textural analysis and supervised classification using Random Forest classifier. we used the Kappa coefficient and a number of other metrics to assess the accuracy of the supervised classification.