We have developed a fully automated approach to critical infrastructure monitoring by using deep learning to analyse bridge images acquired by drones. Saving time and resources, our methodology offers an efficient and precise way to detect infrastructural maintenance needs.
Drones equipped with cameras and sensors have become indispensable tools for data collection in several domains, like monitoring critical infrastructure and 3D modelling. Uncrewed aerial vehicles (UAV) can collect images from large areas and access parts of infrastructure that can be harder to get to manually, saving time and money in the process.
Detecting vulnerabilities and damage with deep learning
We have developed a methodology to significantly improve the processing of drone-generated data for monitoring critical infrastructure, based entirely on automated image analysis.
Our primary focus was optimising inspections of bridges, where rust, cracks in the concrete and road damage can be critical issues in need of repair. A fully automated method was achieved with AI software rooted in deep learning to scan each image and detect vulnerable or damaged sections in need of maintenance work. The algorithm was trained on image data collected in collaboration with The Norwegian Public Roads Administration and Orbiton’s own images and annotations.
Project: InfraUAS – Monitoring critical infrastructure using uncrewed aerial vehicles
Partner: Orbiton AS (now Norse Asset Solutions) and The Norwegian Public Roads Administration (Statens vegvesen)
Funding: The Research Council of Norway.