Back on Track: Next-Gen Railway Asset Management with Computer Vision

Managing the Dutch railway network in order to get trains to their destination safely and on time requires accurate knowledge of the location and state of the various assets on the rail. But how to cope with the bookkeeping when you have over 7000 km of railway? While at the same time maintenance is outsourced to multiple contractors who also have their own systems? ProRail has developed a computer vision solution to verify, update and correct the existing database, rather than sending people out in the field and relying on manual labour. Special trains, equipped with cameras, regularly capture the entire Dutch railway on film. Using convolutional neural networks, assets are identified and localized. Not only is it more efficient than manual inspection, it also has a higher accuracy. In this talk, we will focus on the implementation of computer vision algorithms in a large organization, such as ProRail. We will elaborate on the challenges that we faced, for example handling millions of images and connecting to the existing infrastructure. By putting the database back on track with computer vision, we create a sustainable solution for ProRail.

This lecture is made possible by Vantage AI.