Real Time Flood Response Sysytem: Water Level Mapping and Stranded People Detection

This paper is concerned with using YOLOv5 deep learning models for the automatic detection of human presence as well as the magnitude of flooding from flood videos. The system analyzes flood images and videos acquired by a drone, identifying lost people and evaluating flood intensity depending on water height. Another strength of the model is that it uses ARIMA regression for severity prediction so that disaster management will be of great use. Most of the applications are developed to be user friendly and it includes a feature for uploading the dataset, preprocessing the image, generating the model and real time detection. From this study with the help of AI driven flood monitoring systems, quick rescue operations can be facilitated and help in disaster response management systems. Real-time video processing is another direction of future work as well as expansion of the dataset.