Evaluating traffic trajectories from stationary multiple object tracking systems
Markus Steinmaßl, Karl Rehrl (2025): Evaluating traffic trajectories from stationary multiple object tracking systems In: Journal of Location Based Services
Stationary multiple object tracking (MOT) systems beside the road infrastructure are increasingly used to acquire detailed trajectory data of traffic participants. While comprehensive methods for validating MOT algorithms have been previously proposed, there is a need for evaluation methods focusing on tracking results and their purpose of use for different applications. Especially if such systems are installed in new locations, the quality of the resulting tracking data is in question. In previous works, methods to evaluate the quality of object positions have been proposed. However, methods addressing quality criteria such as accuracy of detection, classification, and tracking have not been considered previously. To address this gap, this work identifies various trajectory quality criteria and discusses methods to evaluate them. Novel approaches to evaluate tracking stability and detection accuracy, treating the generating system as a black box, are proposed. Their applicability is demonstrated on an openly available real-world dataset.
