AI-powered video surveillance of rivers can detect blockages and reduce flooding

Intelligent video surveillance systems, trained to detect blockages in urban waterways, could become an important future tool in flood prevention, according to a new study published today.

Researchers from the University of Bath have shown that their AI-based detection software, ‘AI on The River’, trained to accurately detect natural debris, litter or waste blocking trash screens mounted in culverts, can be integrated with existing video surveillance systems to provide early detection. warning of probable flooding.

Culverts, of which there are over a million in number in the UK and found in almost every town or built-up area, allow streams and rivers to flow under roads, railway embankments and housing estates, meaning they constitute an important but hidden part of waterways and infrastructure. Trash screens, usually a set of bars, are mounted at culvert entrances to prevent debris from passing through.

When a culvert waste screen becomes blocked, flooding can occur quickly. If the flow of water in a culvert is restricted, it can pool and accumulate, causing risks to the structural integrity of the watercourse and the local environment.

Helping flood defenses around the world

The machine learning process created by the team is already attracting the attention of flood prevention organizations in countries like South Africa, where monitoring equipment is available but data that could be used to drive a AI to do the same job are rare or not collected.

Dr Andrew Barnes, Senior Lecturer in the Department of Computer Science at Bath and member of the Center for Climate Adaptation and Environmental Research, and the team who developed the software underpinning the early warning system. He said: “We have been able to develop an effective model that can capture and identify blockages before they become a problem. He is proactive, so he does not wait until a flood occurs to sound the alarm.

“We developed the system to be flexible and scalable: it could be applied almost anywhere, giving it enormous potential in countries where flooding is a problem but where the resources to develop similar tools locally may be rare.”

Identify blockages with great precision

Focusing on a culvert site in Cardiff, the team used machine learning to train a camera system to automatically detect potential obstructions, allowing it to identify likely blockages with an accuracy of close to 90%. %. In most cases in the UK, culverts are manually monitored via CCTV by local authority staff who monitor a series of screens.

Using AI and machine learning to create early warning systems would allow local authorities responsible for maintaining waterway flow to focus resources where they are needed and respond quickly and targetedly to potential blockages.

The proactive nature of the system also provides major safety benefits for response teams, as they can get to sites immediately rather than having to work in dangerous flood conditions.

Dr Thomas Kjeldsen, reader at the Department of Architecture and Civil Engineering at Bath and member of the Center for Regenerative Design and Engineering for a Net Positive World (RENEW), added: “Climate change means the risk of flooding is increasing all over the world. This work opens the possibility of developing new lightweight and cost-effective flood management systems in urbanized areas, enabling authorities around the world to adapt to climate change. This study is a first step towards a sustainable solution for flood forecasting. , and this opened up a multitude of areas for exploration and exploitation.