Future driver assistance systems will not only be limited to one vehicle, but also interact with their environment. Future driver assistance systems will therefore require a detailed knowledge of their surrounding. Video sensors offer a large potential for environmental perception. However, the range of these sensors is very limited.

The combination of video data with a digital map, similar to those used by current navigation systems, allows predicting the course of the road far beyond the sensor range. At the same time, computer vision can be used for localization in a digital map, e.g. for navigation.

On the other hand, information in current digital maps is often outdated, so the data from video sensors can be used for updating the digital map. Thus the digital map turns from a static database into a powerful knowledge representation of what the video sensors have “seen”.

First results have shown that the fusion of GPS, Video and a standard digital map already delivers results that can be used to increase the sensor range. Current research focuses on developing algorithms to localize the vehicle only based on video data, and on developing strategies to identify and update outdated map information at the same time.

In robotics, a similar problem for in-door environments is known as the SLAM (simultaneous localization and mapping) problem. Since these techniques mostly use laser-scanners for sensory perception, the specific geometrical properties of video sensors must be considered additionally.