Event-based cameras as the key to precise sleeper detection

Event-based cameras as the key to precise sleeper detection

Sleepers hidden under the ballast are a particular challenge when tamping. Event-based cameras could make machine operators’ work easier, protect sleepers from damage, and pave the way for fully automating the tamping process.

The problem that arises when tamping ballasted tracks is that the sleepers are often completely hidden under the ballast. Vegetation or foreign objects may interfere with existing systems for sleeper detection, such as eddy current sensors. Plasser & Theurer is currently developing an innovative solution to precisely detect the position of the sleepers, even under these difficult conditions, and to prevent damage to superstructure components.

A key component is the event-based camera (see box “Efficient image processing modelled on nature”). It records the movements of the ballast grains as soon as the lifting and lining unit lifts the track directly before tamping it. The image data is analysed in real time using specially developed algorithms. By precisely recording where ballast is moving and where it is not, the position of the sleepers can be determined.

The technology is still in the development stage, but a successful test run in collaboration with our Technology Centre in Purkersdorf has already proved its efficiency. An event-based camera was used to precisely determine the position of the sleepers. As the amount of ballast increased, it became more difficult to detect the movements exactly; however, the system still produced promising results, even under these conditions.

These findings form the basis for developing it further and integrating it into the Plasser TampingAssistant, which automates positioning of the work units and facilitates efficient operation, especially near turnouts and crossings. In the long term, the system should work autonomously and thus make a fully automated tamping process possible.

The next steps involve refining the algorithms. This is crucial, as the system must be able to reliably detect the sleepers in a very short time window during the lifting and lowering process – an important capability on the way to a fully automated tamping machine that also works efficiently and precisely when ballasting.

Efficient image processing modelled on nature

Event-based or neuromorphic cameras are based on a principle that is modelled on the human eye. Unlike conventional cameras, which record images at a fixed rate, these cameras measure asynchronous changes per pixel. Each pixel works independently and only reports movements or changes in brightness, which significantly reduces the amount of data recorded and increases the processing speed.

This technology has several advantages. As only relevant image information is recorded, memory requirements decrease significantly and energy consumption is also lower. At the same time, these types of cameras enable an extremely fast response time.

Event-based cameras are used in areas where high speed, efficiency, and precision are important when recording movements or changes. They open up new possibilities for numerous applications in industrial image processing.

Further information

Show all articles