Flood Area Analysis Using Satellite Image

Abstract

Urban flooding presents a serious threat to hu- man safety, transportation systems, and public infrastructure. This study proposesan image processing and deep learning- based framework for analyzing flood-prone areas using satellite imagery. The approach integrates optical and radar satellite data to identify flood-affected zones using spectral indices and temporal change detection techniques. Using a Convolutional Neural Network (CNN) model based on VGG16, the system classifies satellite images into flooded and non-flooded categories. The model achieved an accuracy of approximately 96 percent, demonstrating its potential to support disaster management teams in timely decision-making. The results highlight the utility of combining deep learning with remote sensing data for flood impact assessment and disaster mitigation.