GPS Driven Camera Selection in Cyclocross Races for Automatic Rider Story Generation

Jelle De Bock, Alec Van Den Broeck, Steven Verstockt


Cyclocross races are a very popular winter sport in Belgium and the Netherlands. In this paper we present a methodology to calculate the proximity of riders to a number of cameras that are located on a cyclocross course in order to automatically select the correct camera for each rider. The methodology is based on two main input sources. The first input is the course with cameras positioned along it. As the course and camera information is usually available as pdf and isn’t directly processable by computer programs, we propose the conversion GeoJSON. The second requirement for our methodology is accurate location tracking of the athletes on the course with the help of wearable GPS trackers. We present an experimental camera proximity algorithm that uses both input sources and finds for every rider at any given moment in the race the closest camera or vice versa. The output of this methodology results in automatic identification of the filmed riders by a given camera at a given moment in the race and might benefit post-processing of the camera video streams for further computer vision-based analysis of the streams, for example, to pre-filter the camera streams or to generate rider and team stories.


Paper Citation