Through the Image Analysis Railway Infrastructure (IARI) project, the Computing Centre will supply and develop the machine learning and AI technology, which will be fed images and video annotated by Bane Nor staff.
The project will develop a system to inspect and monitor key components including track, switches and overhead contact lines, using camera-based data capture and automatic image analysis. This will enable more frequent and faster inspections without delaying services.
“The solution will make it possible to inspect and analyse the railway infrastructure in a fast, secure and objective manner without stopping traffic,” says Mr Jørgen Torgersen, Bane Nor's administrator responsible for the project. “Pictures and video from measuring trains, passenger trains, work machines and drones in addition to fixed-mounted cameras, which Bane Nor has already installed, are examples of areas where automated image analysis can be useful.”
The research effort will focus on adapting existing general deep learning-based algorithms for rail uses and training these algorithms to recognise relevant errors and nonconformities.
The system is expected to increase the number of faults detected, and enable earlier detection, with this increased workload equalling the reduction in inspection work carried out.
“The use of modern technology such as AI and machine learning in traditional businesses such as railways is important to meet the future transport needs,” says Mr Sverre Kjenne, executive vice-president of digitalisation and technology at Bane Nor. “The railway is being developed and digitised to provide a better service for most people. We are grateful for the support from the Norwegian Research Council and proud that they are investing in this project.”
