Real-Time Flood Warning System Development Using Computer Vision Method

Abstract

The purpose of this research is to develop a system for monitoring river water levels and sending real-time flood warnings via a Telegram bot with images and text. This information is sent to residents and business owners in flood-risk areas, as well as to flood management officials, in advance. The system allows them to prepare and prevent potential flooding. The system was implemented for the experiment at the Mekong riverbank in Luang- Prabang city by utilizing a computer vision algorithm to process images from a Hikvision ColorVu CCTV IP camera to monitor water level changes at the staff gauge with three colored warning zones, such as the green zone (safe zone), the yellow zone (watch zone), and the red zone (danger zone). Such monitoring is achieved through image processing by color detection and color segmentation to determine the water level position. The system then sends alerts to subscribed users via a Telegram Channel using the internet connection from a 4G smartphone hotspot. Experimental results indicated that the system can successfully monitor water level changes along the Mekong riverbank and automatically send flood warnings in real-time in both daytime and nighttime via the Telegram bot with an accuracy of 85.39% and a warning delivery time of less than 1 minute.