Improving coastal protection with AI drones

When dykes break, this usually leads to severe flooding and damage. A twelve-strong team from Magdeburg-Stendal University of Applied Sciences is using drones to detect cracks and rotten areas in dykes, dams and bridges at an early stage. As part of a four-year project, the Hydraulic Engineering and Hydraulic Engineering Testing research group developed an AI drone that was specially designed for this task. The combination of drone use and artificial intelligence is the special feature of this project, as the researchers emphasize.

Professor Dr.-Ing. Bernd Ettmer has been working as Professor of Hydraulic Engineering and Hydraulic Engineering Testing at the university since 2008. He and his team deal with dams, dams, hydropower plants, locks, bank reinforcements and flood protection. Since 2017, they have been using drones to take surface photos. They use these images and data from measuring boats to create terrain models to gain insights into the conditions of the areas. "For us, the drone is initially just a measuring instrument. You give it a route, it flies autonomously and we have a complete computer model of the area it flies over," Ettmer explains to Informationsdienst Wissenschaft (IDW). The team used the research project to identify cracks and damage to buildings, even underwater. The team used AI algorithms to ensure that the photos taken by the drone did not have to be laboriously examined for damage by employees.

Read also

Remote control failed: North Sea water threatened to flood Amsterdam

Learning processes improve damage detection

The AI has to be trained with images of a crack or damage to be able to recognize them. Initially, they use cut-out pieces of paper as symbolic cracks, which are adapted to natural cracks in shape and color as training progresses. Ettmer and his team have set up a test dyke in the university's laboratory halls, over which the drone flies to take pictures. The AI technology learns to identify these images as damage. As the project only started at the beginning of this year, the first few months were spent laying the groundwork. Dr. Daniel Hesse, senior engineer for hydraulic engineering, is primarily responsible for the drones in the project. "We are currently operating the drones with a standard RGB camera. The next task is to check which sensors we need to adequately implement the project. These could be spectral cameras, laser scanning or thermal cameras," explains Hesse.

After the tests in the laboratory halls, the drone will be trained on the university grounds before being used on dykes. Ettmer explains what they want to achieve with the project: "Ideally, the AI will evaluate the images taken by the drone during the flight. This means that potentially damaged areas can be detected and reported at the same time." This allows them to identify critical areas more quickly and deploy human resources in a targeted manner to repair damage. By the end of the project in 2027, the researchers hope that the combination of drone and AI will provide a valuable tool for the prevention of flooding and inundation.

(nen)