RivMon - Development of a camera- and AI-based risk assessment for Hazard detection and mitigation during heavy rainfall for local aereas watercourse sections

The challenge

Heavy rainfall and flash floods are becoming more frequent across Europe, yet smaller rivers, streams, and drainage channels are still poorly monitored. Unlike large rivers, these local watercourses often lack reliable forecasting systems, real-time hazard assessment, and continuous monitoring infrastructure.

Current systems face several major problems:

• Limited sensor coverage and unreliable measurements

• No automatic detection of blockages, vegetation, or flow obstructions

• Missing real-time flood forecasting for small catchment areas

• High false alarm rates caused by faulty or incomplete sensor data

• Lack of dynamic flood visualization and risk assessment

• Insufficient hydrological data for accurate simulations and emergency planning

These limitations make it difficult for municipalities, emergency services, and water authorities to react quickly and effectively during heavy rainfall events.

The RivMon project

RivMon is an innovative research and development project that aims to solve these challenges through the combination of artificial intelligence, camera technology, hydrological modeling, and real-time data processing.

The project develops:

• AI-based camera systems for automatic detection of hazards such as blockages, debris, and excessive vegetation

• Camera-based hydrological monitoring to measure water levels, flow velocity, and discharge in real time

• Dynamic flood forecasting models for predicting flood development in small watercourses

• Automated risk analysis systems to identify endangered infrastructure and critical areas

• Real-time visualization platforms for municipalities and emergency services

• AI-generated situation reports to support operational decision-making during flood events

Project goal

The goal of RivMon is to create a scalable and intelligent early warning system that improves flood preparedness, reduces false alarms, and enables faster and more targeted responses during extreme weather events.

TU Dresden - Main Research Goals

The contribution of TU Dresden focuses on the development of advanced camera-based hydrological monitoring methods for small watercourses.

The main goal is to transform camera systems into intelligent measurement tools capable of capturing critical hydrological parameters in real time. This includes:

• Development of robust camera gauges for water stage measurement, combining AI and photogrammetry  

• Flow velocity measurement through AI-supported image velocimetry methods

• Discharge calculation based on water surface geometry and flow dynamics

• Precise camera calibration and georeferencing for accurate long-term monitoring, with the development of self-calibrated cameras 

• Research on stereo camera systems to improve measurement reliability in unstable flow situations

A key challenge addressed by TU Dresden is the reliable extraction of hydrological data under changing weather, lighting, and flow conditions. The research aims to create scalable and automated methods that can be deployed across many different watercourse sections without requiring complex manual calibration.

The developed technologies provide the hydrological foundation for RivMon’s real-time flood forecasting, hazard assessment, and early warning capabilities.

Partner

• SPEKTER GmbH
• Fraunhofer Gesellschaft e. V. Institute of Optronics, System Technologies and Image Exploitation(IOSB); Industrial Automation Branch(INA) of the Fraunhofer IOSB
• TUD Dresden University of Technology; Institute of Photogrammetry and Remote Sensing; Junior Professorship in Geosensor Systems

Funding

This project is funded by the Federal Ministry for Economic Affairs and Energy (BMWE) on the basis of a decision by the German Bundestag.