Post-Processing of Multi-Target Trajectories for Traffic Safety Analysis
Thorben Janz, Andreas Leich, Marek Junghans, Kay Gimm, Shishan Yang, Marcus Baum · 2018
This work presents a method for qualitatively improving the output of an existing video and radar-based multitarget tracking system by means of post-processing. The proposed method is a novel two-pass track stitching method, which involves a breaking phase for tracklet creation and a linking phase to create putative assignments. Minimum-cost network flow techniques are utilized to find optimal tracks. The method has been tested with data from a research intersection in Braunschweig, Germany, which is operated by the German Aerospace Center for the purpose of traffic safety analysis. Furthermore, an evaluation with the publicly available MOTChallenge benchmark is provided. Improvements achieved are: 14% to 33% less ID switches and 40% less track fragmentations. Both criteria are especially relevant for traffic safety analysis.