Collecting pedestrian trajectories

For the proper understanding and modelling of pedestrian dynamics, reliable empirical data are necessary for analysis and verification. To this end, we have performed a series of experiments with a large number of persons. Such experiments give us the opportunity to selectively analyse parameters independent of undesired influences and adjust them to values seldom seen in field studies.We are developing software for the time-efficient automatic extraction of accurate pedestrian trajectories. Depending on the camera system the software is able to detect and track people on planar or uneven terrain with or without markers.In this paper, we summarise the experiments we have accomplished and the possibilities of our extraction techniques, in particular the newly introduced algorithm of markerless detection in stereo recordings. The markerless detection based on groups of ellipses approximating isolines of the same distance to an overhead stereo camera.