Frustrating as it may be, you can’t always depend on the weather forecast. Even when it is not supposed to snow, the ground can suddenly turn white. Even for private individuals, this scenario is aggravating. But for companies such as Deutsche Bahn, it can also be very expensive as, for example, snow clearing services may need to be activated at short notice.
Thanks to modern camera technology and intelligent image recognition, however, this problem vanishes like yesterday’s snow. Video analysis allows the current conditions to be evaluated in near real time and corresponding measures to be implemented promptly. As an initial step, DB Services, together with DB Station&Service, set up weather cameras in 22 test stations over the past year. These cameras record images of a specific area every 15 minutes and send these images automatically to the winter maintenance control centre via mobile communications. In this way, in addition to weather reports, the winter maintenance manager has a visual impression of the current situation and can initiate processes accordingly. Until now, however, the images have been evaluated by employees, who not only looked for snow, but – for reasons of data protection – also manually deleted any images that depicted people.
Thousands of images each day
Even with only 22 cameras in use, more than 2,000 images have already been transferred per day. By next winter, cameras are expected to be installed in around 200 stations. It has quickly become obvious that technical support is needed. This support is being provided by the big data startup ZERO.ONE.DATA as well as more experts from DB Systel. To enable weather-related changes to be detected automatically, however, a software solution developed by DB Systel had to be trained to evaluate all transmitted images. Test image data from the previous winter helped the system to identify snow unambiguously. At the same time, the system had to be able to detect people in the images and disguise their identities permanently. The success rate for the first test using historical material was already 95 percent – a value that is expected to be even higher during live operation. In a later stage of development, if snowfall is detected, an order to clear the snow will be sent automatically to the relevant service provider. It will also be possible to check automatically whether the snow has been cleared within the agreed time window.
Technology with great potential
Besides analysing video footage from weather cameras, there are many more potential use cases. At DB Station&Service, more than 80 million euros are currently being invested in developing and modernising video technology. Cameras will then be used to monitor restricted areas, track routes, count people and analyse passenger traffic. However, they can also be used to identify unattended luggage. For example, if a camera detects an unattended suitcase, security personnel could be notified automatically much more quickly in the future.
Together with on-board cameras, the cameras will also be able to improve passenger safety by detecting aggressive behaviour. Another interesting use of cameras is the identification of rust damage in outdoor facilities such as those at DB Energie or DB Netz. At DB Netz, camera analysis could be used, for example, to monitor clearance and to document infrastructure. At DB Cargo, video analysis could be used to help identify damage to wagons and containers or to determine wagon identification numbers.
Great potential and many ideas
The material for video analysis can come from a variety of different sources, facilitating the automatic support of business processes and evaluation of even large volumes of image data. DB Systel also anticipates additional applications and services for which cameras do not even need to be installed, such as services that can be provided using smartphone cameras. Consequently, a prototype for an Apple iOS app has been developed by the Mobile Team, that helps visually impaired passengers to identify whether the door of an ICE train is open or closed. The user walks along the platform while pointing the smartphone camera in the direction of the train and the app emits a signal when it detects an open door.
Even more functions are conceivable in the future, such as automatic recognition of the train number or a reserved seat. Clients can also license this technology for their own apps. For Ellen Engel-Kuhn, head of the disability liaison office at DB Vertrieb, this is an important step: “The ‘door finder’ app – or a future app that can recognise the train number or a reserved seat – would provide blind and visually impaired passengers with more autonomy and a much more comfortable journey.”
But regardless of where video analysis is used, the detection of defined scenarios and the shorter response times provide tailor-made support and optimisation for operating processes.