Can smart cameras detect faults on a railway?

Mon 05 Sep 2022 13:44

To use smart cameras to detect fault in placements of clips on a railway – that is what Irida Shallari’s research project is about. The project's goal is to create a solution that facilitates the constructions of a railway by using AI and deep learning.

Irida Shallari checks the information from the camera on a computer.

The aim of the research project EDGY, Embedded Deep Learning Systems, is to develop a smart camera that can be used when building railways, to make sure that it is done in a correct way to enhance the quality of the construction. One example where the cameras can be used is to detect faults in the placement of the clips that attach the rail to the sleepers. It can also be used to facilitate the maintenance of the rail.

The camera set-up will be attached onto the machines that constructs railways. It will continuously take images of the sleepers and the clips to make sure that they are placed correctly. If any damage has been done to the clip itself or other parts of the railway, for example wires for electricity or grounding, the smart camera will be able detect that.

The camera set-up.

Camera set-up based on AI and deep learning

The technology in the smart camera is based on AI. The images that the camera capturers will continuously be processed on a dual system relying on both deep learning and traditional image processing. This system will be able to define if the clip is correctly placed, if the angle is good or if there is a risk of, for example cutting a wire based on the placement of the clip. To avoid all these issues and build a high-quality railway, capturing the images and processing them will basically be done in real time.

– The computational part, to process the images, will mainly rely on the deep learning algorithm that we are using, Irida Shallari explains.

We aim to build the camera nodes at as low cost as possible. We want to enable industries to implement smart sensors in their industrial setups, she continues.

A digital twin

If we look further into the future, this set up could also be used to set up a digital version, a digital twin, of for example a railway system. A digital copy could assist in planning maintenance since you would have a real time view of the situation.

The project started in March 2022 and will run for two years. Pandrol, a manufacturer of the machines that construct railways, are part of the project and will work together with Irida to try and implement the set up on their machines.

Irida Shallari and Anders Olsson from Pandrol.

Irida Shallari and Vincenzo Gallo is working on the smart camera.
Irida Shallari and Vincenzo Gallo pull the cart with the camera set.
Three people are working on the camera setup.
Green clips along part of a railway rail.

The page was updated 9/5/2022