Domain: Smart Cities
Vehicle tracking in city of Berlin.
The internationally renowned Technische Universität Berlin is located in Germany’s capital city at the heart of Europe.
Their academic activities are focused on building a distinctive profile for the university,
ensuring exceptional performance in research and teaching, providing their graduates with excellent qualifications and a modern approach to university administration.
In cities, it often is not possible to deploy traffic cameras in locations with optimal field of views. Trees or infrastructure may be in the way, sometimes larger vehicles obstruct the visibility of other vehicles, bikes, or pedestrians. Furthermore, large intersections make it hard to track vehicles’ movement with a single camera. In the setup of this challenge, two approaches have been implemented to address these issues. i) Multiple cameras to observe larger areas and ii) cameras on top of buildings to get a larger field of view. However, traffic analysis software solutions lack in different aspects, depending on the light conditions, the viewing angle, and tracking within a single and across multiple fields of view / video streams.
DataThe challenge has the following sample datasets available for download
- Tracking must be reliable accounting for occlusion, weather conditions, road conditions, lighting conditions.
- Solution should also identify different types of vehicles e.g. bicycles, motor bikes, busses, trucks, emergency vehicles.
- Relative position (coordinates) should be tracked in time and space.
- Identify a vehicle in one video stream (re-identify a vehicle if it is occluded by, for example, a bus).
- Speed and size estimate of the vehicle should be also provided.
- Within a short period of time, the vehicle should be re-identified if it shows up in an adjacent camera’s video stream.