AN AUTOMATED ALGORITHM FOR FLOOD DETECTION USING MULTISPECTRAL IMAGES

摘要

Detecting floods is one of the most important applications of remote sensing. There have been lots of techniques for flood monitoring and mapping including visual interpretation, post-classification approach, and image differencing based change detection. Commonly used classification-based techniques can provide detailed and precise information on flooding events. The performance of change detection techniques can be also affected by the water turbidity and the presence of residential area. The primary objective of this study is to automatically detect the flooded areas under varying environment conditions. In this study, multi-temporal SPOT-4 and 5 images acquired over Cagayan de Oro, Philippine are investigated. Spectral features in the multispectral SPOT data are evaluated to increase the detectability of inundated riverine areas. And consequently, an automated flood detection algorithm that is robust to weather and varying environment conditions of residential area is proposed.