查看文章

Floods are natural disasters and pose a threat to the lives, property, and infrastructure of an urban area. Though their risk cannot be fully eliminated, several methods can be used to manage floods, once they occur. This includes identification of flood-prone areas, timely detection of the affected areas, mapping rescue routes and arranging logistics to carry out the rescue as soon as possible. The use of advanced innovative technologies for flood management such as image detection and machine learning can assist in effective flood management. This paper presents a novel approach through the integration of image processing and machine learning to detect flood-affected areas using a set of images. The three-step approach proposed in this study is based on landmark detection from images, training of a machine learning algorithm and classifying images from an area as flooded or non-flooded. The results based show an accuracy level of 90% depicting the significance of the proposed model for image-based flood detection.