Flood Extent Mapping with Unmanned Aerial Vehicles Data using Deep Convolutional Neural Network
Vaishnavi Barkhade, Shruti Mahakarkar, Rahul Agrawal, Chetan Dhule, Nekita Chavan Morris
Abstract
Flooding is a common occurrence that results in human fatalities, severe environmental harm, and major infrastructural damage. A method for mapping areas with apparent and subterranean vegetation flooding that integrates CNN and region growing (RG). To determine the number of floods beneath plants that are hidden from photojournalism using the digital elevation model(dem), the Region Growing technique is applied, whereas to extract areas which are flooded a Convolutional classifier is used. The CNN-based classifier is trained using a data augmentation strategy to enhance the classification outcomes. This paper develops an automatic flood detection system for UAV aerial photographs using deep learning algorithms. Unmanned aerial vehicles (UAVs) have the potential to offer high-resolution data with the ability to quickly and accurately detect inundated areas under intricate urban environments. This research makes use of unmanned aerial vehicles to develop an automated imaging system that can identify waterlogged areas from aerial pictures. The suggested method combines CNN and region growth methodologies for mapping regions with visible and subsurface vegetation flooding, resulting in a more complete flood detection system.UAVs offer high-resolution data collecting as well as the rapid and precise detection of flooded regions in complicated urban contexts. The use of data augmentation improves the classification results of the CNN-based classifier.