Real-Time Flood Monitoring with Computer Vision through Edge Computing-Based Internet of Things

Natural disasters such as severe flooding can cause catastrophic losses to properties and human lives. Constant real-time water level monitoring prior to a flooding event can minimise damages and casualties. Many of the currently deployed water level monitoring systems typically use a combination of float-type or ultrasonic sensing, image processing and computer vision techniques. However, these systems incur high computing and hardware requirements, which hinder the deployment of such systems in resource-constrained and low-cost environments. The recent development of technologies empowered by the Internet of things (IoT) and edge computing have enabled real-time systems to be deployed at a significantly lower cost and a far more distributed manner. In this paper, we propose an architecture for flood monitoring using RGB-D cameras with stereoscopic capabilities to measure the water level in an open environment. Our system uses image preprocessing techniques to account for chromatic aberration due to overexposure, followed by postprocessing before the depth readings are extracted. Data processing and water level information extraction are entirely performed on an edge computing device, therefore greatly reducing the amount of data transmitted to the cloud server. We practically implemented and experimentally validated this system in the real world, under a wide range of weather and lighting conditions. Our results showed promising outcomes and demonstrated the applicability of our proposed system in a wider context.