Modeling Sources and Sinks in Crowded Scenes by Clustering Trajectory Points Obtained by Videobased Particle Advection

This thesis provides an approach to analyze dense crowded scenes in real-time. Therefore an efficient algorithm models dense crowded scenes with the aid of particles, moved by the optical flow calculated between two consecutive frames. Thus trajectories can be obtained, without using a tracking algorithm. Areas of interest are modeled by a further processing step, yielding in reasonable results. Analysis of real surveillance video footage is very challenging, hence this thesis provides solutions to enhance the quality of trajectories. As dense crowded scenes are analyzed,many trajectories are interrupted thus making the choice of an appropriate clustering algorithm challenging. This thesis evaluates different clustering algorithms and their practicability in combination with the real-time particle advection algorithm.