ARTEMIS: advanced AI-based monitoring of railway infrastructure.
ARTEMIS is an innovative system currently being developed as part of an Innosuisse project in collaboration with the Eastern Switzerland University of Applied Sciences (OST) and is already in use in a pre-production environment.
The name ARTEMIS stands for Automatic InfRasTructurE MonitoRIng System. The platform combines deep-learning algorithms with advanced data processing to detect railway infrastructure objects automatically from image data. At the same time, it determines their exact position through geolocation and captures important additional information such as text or numbers.
How ARTEMIS works
The detected objects and their condition are stored in a database. This can support initial asset inventorying or, through comparison with existing datasets, identify changes in the infrastructure over time. It enables more efficient management of railway infrastructure throughout its entire life cycle. ARTEMIS also supports the automated creation of precise “as-is” models of the recorded railway environment.
The database currently contains around 100 different types of railway infrastructure objects, including sleepers, signals, overhead line masts, balises and safety-relevant infrastructure elements. Ongoing development aims to improve detection accuracy and positioning quality while reducing the manual effort required to train new object categories.
First deployment with RhB
In line with an agile development approach, ARTEMIS has already been successfully tested in practice. Together with the Rhaetian Railway (RhB), a detection campaign was carried out on the section between Landquart and Klosters. Image data from the 2025 measurement campaign conducted by the FahrwegDiagnose joint venture was used to detect and locate nine object categories, including signals and balises.
A total of 14,700 images were analysed. At least one relevant object was detected in 1,375 images, allowing around 1,200 unique infrastructure elements along the route to be successfully identified and georeferenced.
Impact on operations
The results of this project support the updating of asset databases and create a reliable basis for monitoring and maintenance. ARTEMIS enables systematic digitalisation and continuous condition assessment of railway infrastructure, thereby helping to improve safety across the entire network.
Automated detection reduces the need for manual inspections and work in the track area, lowering operational effort while improving occupational safety. Thanks to the high temporal resolution, conditions can be reassessed regularly and changes can be identified at an early stage.
By comparing newly detected objects with existing data, the system ensures consistent validation and high data quality. At the same time, it creates the basis for predictive maintenance by enabling targeted measures and reducing unnecessary interventions – without compromising reliability.
Outlook
The findings from the project and user feedback feed directly into further development. In the long term, ARTEMIS has the potential to enable scalable and intelligent monitoring of railway infrastructure through the use of artificial intelligence.
