Dynamic Objects

The detection and tracking of moving objects has application in advanced driver assistance systems and autonomous robotics. Intelligent vehicles that monitor outside traffic participants have the potential to warn about unforeseen events and to trigger emergency stops in situations were humans are unable to react. The potential to reduce road accident related fatalities has attracted many researchers. However, the problem of moving object detection and tracking is challenging. Typical objects in traffic scenarios differ greatly in appearance and size. The diversity in different types of vehicles alone is large. Additionally, targets move with velocities varying from approximately 2 km/h for pedestrians to over 50 km/h for vehicles. In typical urban environments, often more than 5–10 objects appear in the same area. Partial and full occlusion occur frequently and ambiguities in data association are the consequence.

We have developed methods for the detection and tracking of moving vehicles and pedestrians using particle filters. We first separate foreground und background using geometric and temporal cues and then initiate candidate tracks by searching for object-specific motion patterns. Finally, objects are tracked using a geometric measurement model for laser data.