News Release

AI-trained CCTV in rivers can spot blockages and reduce floods

Machine learning-equipped camera systems can be an effective and low-cost flood defence tool, researchers show

Peer-Reviewed Publication

University of Bath

The ‘AI on The River’ software is trained to accurately detect debris blocking culvert trash screens

image: 

The ‘AI on The River’ software is trained to accurately detect debris blocking culvert trash screens

view more 

Credit: University of Bath

AI-trained CCTV in rivers can spot blockages and reduce floods 

Machine learning-equipped camera systems can be an effective and low-cost flood defence tool, researchers show 

Smart CCTV systems trained to spot blockages in urban waterways could become an important future tool in flood prevention, new research published today has found. 

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

Culverts, which number over 1 million in the UK and are in almost every city, town or built-up area, allow streams and rivers to flow under roads, railway embankments and housing developments, meaning they are an important but hidden part of waterways and infrastructure. Trash screens, usually a set of bars, are mounted at culvert entrances to stop debris from passing through. 

When a culvert trash screen is blocked, flooding can happen quickly. If the flow of water into a culvert is restricted, it can build up and pool, leading to risks to the structural integrity of the waterway and the local environment. 

Helping flood defences worldwide 

The machine learning process created by the team is already attracting attention from flood prevention organisations in countries including South Africa, where monitoring equipment is available but data that could be used to train an AI to do the same job is scarce or not collected. 

Dr Andrew Barnes, a lecturer in Bath’s Department of Computer Science and a member of the Centre for Climate Adaptation & Environment Research, and the team that developed the software underpinning the early warning system. He said: “We’ve been able to develop an efficient model that can capture and identify blockages before they become a problem – it’s proactive, so doesn’t wait for a flood to happen before raising the alarm. 

“We’ve developed the system to be flexible and scalable – it could be applied almost anywhere, giving it huge potential in countries where flooding is an issue but where the resources to develop similar tools locally may be scarce.” 

Identifying blockages with high accuracy 

Focusing on a culvert site in Cardiff, the team used machine learning to train a camera system to automatically spot potential obstructions, resulting in it being able to identify likely blockages with close to 90% accuracy. In most cases in the UK, culverts are monitored manually over CCTV by local authority staff watching a bank of screens. 

Using AI and machine learning to create early warning systems would allow local authorities in charge of keeping waterways flowing to focus resources where they are needed and respond to potential blockages quickly and in a focused way. 

The proactive nature of the system also offers major safety benefits to response teams, as they can attend sites immediately rather than having to work in dangerous flooded conditions. 

Dr Thomas Kjeldsen, a Reader in Bath’s Department of Architecture & Civil Engineering and a member of the Centre for Regenerative Design and Engineering for a Net Positive World (RENEW), added: “Climate change means the risk of flooding is growing all around the world. This work opens the potential for the development of new, light weight and cost-efficient flood management systems in urbanised areas, enabling authorities around the globe to adapt to the changing climate. This study is a first step toward a sustainable solution to flood forecasting, and it has opened a multitude of areas for exploration and exploitation.”  

The paper, CCTV Image-based classification of blocked trash screens, is published today in The Journal of Flood Risk Management: https://onlinelibrary.wiley.com/doi/10.1111/jfr3.13038 (DOI: 10.1111/jfr3.13038). 

The research was supported by the Engineering and Physical Sciences Research Council (EPSRC) grant number EP/W034034/1 Reclaiming Forgotten Cities - Turning cities from vulnerable spaces to healthy places for people (RECLAIM). 

ENDS 

For more information or to request interviews, contact Will McManus in the University of Bath press office: wem25@bath.ac.uk / +44(0) 1225 385 798 

Images are available at: https://tinyurl.com/bdxsnj5a  

 

The University of Bath

The University of Bath is one of the UK's leading universities, with a reputation for high-impact research, excellence in education, student experience and graduate prospects. 

We are ranked in the top 10 of all of the UK’s major university guides. We are also ranked among the world’s top 10% of universities, placing 150th in the QS World University Rankings 2025. Bath was rated in the world’s top 10 universities for sport in the QS World University Rankings by Subject 2024.

Research from Bath is helping to change the world for the better. Across the University’s three Faculties and School of Management, our research is making an impact in society, leading to low-carbon living, positive digital futures, and improved health and wellbeing. Find out all about our Research with Impact: https://www.bath.ac.uk/campaigns/research-with-impact/

Research from Bath is helping to change the world for the better. Across the University’s three Faculties and School of Management, our research is making an impact in society, leading to low-carbon living, positive digital futures, and improved health and wellbeing. Find out all about our Research with Impact: https://www.bath.ac.uk/campaigns/research-with-impact/


Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.