Reference DB Cargo

Digital Diagnostics on Freight Trains with AI

AI Identifies and Classifies Freight Car Markings

With a fleet of around 83,000 freight cars, DB Cargo is the largest rail freight service provider in Germany. It offers its services throughout Europe and beyond.

Each car has a label with the car number, tare weight, load capacity, load limit number, route class, brake equipment, and many other details. To ensure that freight cars reach their destination as quickly as possible, it is essential that car markings are recognized correctly and quickly in marshalling yards. Until now, employees have not had any digital support for this task.

As part of the “AI@Digital Diagnostics Freight Wagons” project, DB Cargo has teamed up with experts from Telekom MMS and DB's internal AI experts AI Factory and vsion. ai to develop image processing with AI to recognize the markings on freight cars as they pass through camera bridges and process them digitally – a challenging task given the variety of car types, different designs and marking formats, and the varying external condition of the cars.

The introduction is imminent and should make work easier for employees.

-> Increase efficiency and innovation in your business processes thanks to advanced technology: Our experts will be happy to advise you on AI video analysis with AI Vision!

Reference at a Glace

Task

The reading of the lettering of freight wagons by employees is to be digitally supported.

Solution

A large number of specially trained AI models enable object recognition and content extraction of the addresses on freight wagons.

Result

Automated comparison of recognized addresses on freight cars with relevant IT systems.

In the challenging task of reading wagon labels using AI, we worked very trustingly and on an equal footing with the DB teams AI Factory and vsion.ai, as well as Telekom MMS. This AI solution makes work much easier for our employees. Achim Leister, Project Manager, DB Cargo

Benefits for our Customers

Automation of manual processes ensures greater efficiency
High acceptance among employees, as monotonous tasks are eliminated
Better predictability for customers thanks to fewer disruptions and faster train delivery

Requirements

Solution

Benefits

Requirements

Recognizing all Labeling Formats on a Wide Variety of Wagon Types With AI

Wagon labels are relevant for the handling of wagons during train formation. In some cases, addresses must be manually checked and compared with specifications in the IT system by employees on the track or in the maintenance workshop. This activity takes up valuable process time during train dispatch and is now to be digitized with the help of AI.

DB Cargo can use AI to locate and read the 15 most important addresses in a wide variety of forms on a freight car. It is primarily supported in this by the internal service providers AI Factory and vsion.ai. Telekom MMS has solved a challenging part of the tabular address recognition in this project.

Back in 2019, Telekom MMS developed an application for digital damage assessment of freight cars, which can be used to prepare cars for maintenance more efficiently. This application uses image data obtained when wagons pass through camera bridges. Every day, 9,750 wagons pass through the 13 camera bridges at marshalling yards in Germany – that is 300,000 freight wagons per month and around 3 million per year. This means that 95 percent of DB Cargo's freight car fleet is already recorded using optical sensor technology. This technology should now also be used to read the addresses on freight cars using AI technology. The goal was to digitally compare the data between the labeling on the freight cars and the data stored in the IT system.

For the current proof of concept, the team of experts from Telekom MMS, together with DB Cargo, first piloted the use case “automated recognition of test fields to check the work performed when re-marking the tare weight.”

Solution

AI Models for Object Recognition and Content Extraction

In order for the optical sensors already installed on camera bridges to be able to capture the lettering on each freight car, it was necessary to train two new AI models. Like their predecessors, these models use computer vision and are integrated into the DB Cargo cloud infrastructure as executable Docker containers. For the first detection model, the AI must reliably locate the lettering on the freight car. This is a major challenge, as there are dozens of freight car types of various designs. In addition, some of the cars are heavily soiled, weathered, or sprayed with graffiti. The solution to this problem is known as “object detection.” This model was jointly developed by DB's internal teams AI Factory and vsion.ai.

In the next step, the second AI model created by Telekom MMS, AI Factory, and vsion.ai extracts the contents of the label. To do this, letters and numbers must be recognized and assigned to table fields so that the data can be logged automatically. To this end, the AI expert team has developed a mapping of table templates: Once the tables have been broken down into individual cells, OCR (Object Character Recognition) is used for character recognition. The read content is then transformed into a standardized transfer format and made available for further processing.

Benefits

AI Assistance System Increases Efficiency of Technical Handling of Freight Cars

As a result of the proof of concept, a complex address on a freight car is recognized and read sufficiently reliably, continuously, and automatically with the help of two AI models. This proves the technical feasibility of (partial) automation of data reconciliation in the future. This reduces time-consuming manual processes, eliminates monotonous work steps and the associated susceptibility to errors, and allows the necessary work to be carried out much more quickly and efficiently. For DB Cargo employees, this means less work in adverse weather conditions. The AI solution has therefore been very well received by the workforce.

However, the final quality control remains in human hands – because AI is also to be understood as an assistance system in this application.

In the next stage of development, camera-based digital diagnostics will be put into productive use. Many more applications are planned for this technology. For example, it can be used to diagnose the condition of the brake shoes on rail cars. The camera bridges could also be used to record the contents of freight cars, for example to identify types of scrap metal or damage to tarpaulins or protective films.

About DB Cargo

DB Cargo, a division of Deutsche Bahn, bundles national and international rail freight transport activities. The DB Cargo network comprises 16 national companies and uses one of the largest rail networks in the world. Thanks to its own companies and numerous cooperations and joint ventures with partner railways, DB Cargo is now the only provider to offer freight transport services across almost the whole of Europe. Around 30,000 employees ensure that cross-border transport is managed efficiently, in a customer-oriented and environmentally friendly manner. Almost 60 percent of DB Cargo's transport services are now provided across Europe. With around 83,000 freight cars and around 2,600 locomotives, the company has the largest fleet on the continent.

Our AI Solution

Benefit from advanced technology that offers solutions tailored specifically to your company's requirements! More about AI Image and Video Analysis with AI Vision

Our Contact for Your Questions:

Dr. Cornelia Mossal

Corporate Communication

+49 35 1 - 28 20 - 0
Send e-mail

For tomorrow.
And beyond.

Subjects

Media

Channels