IARI – Image Analysis Railway Inspection

In the IARI project, NR and the Norwegian railway infrastructure owner, Bane NOR, are developing AI-based image recognition for inspection of railway infrastructure.  

Inspection is essential to ensure a safe and effective railway. However, manual inspection is timely and costly demanding, and will often require a cease in traffic to ensure the safety of the personnel. Thus, to expand or replace some inspection tasks with automated methods have a great potential for cost reduction and improving inspection. 

Identify critical inspection tasks

NR develops algorithms that recognize faults on the infrastructure by using images from train or drone mounted cameras. We also use data from IR and depth sensors that give complimentary information. The deep learning based algorithms are trained on large image datasets with known faults. In cooperation with Bane NOR, we work to identify critical inspection tasks that are suitable for image analysis and establish training datasets.