Online and Real-Time Tracking in a Surveillance Scenario

This paper presents an approach for tracking in a surveillance scenario. Typical aspects for this scenario are a 24/7 operation with a static camera mounted above the height of a human with many objects or people. The Multiple Object Tracking Benchmark 20 (MOT20) reflects this scenario best. We can show that our approach is real-time capable on this benchmark and outperforms all other real-time capable approaches in HOTA, MOTA, and IDF1. We achieve this with two contributions. First, we apply a fast Siamese network reformulated for linear runtime (instead of quadratic) to generate fingerprints from detections. Second, we extend the walking path as predicted by the Kalman filter with an additional motion model that also takes into account unforeseen changes in the intention of the tracked person. Thus, it is possible to associate the detections to Kalman filters based on multiple tracking specific ratings: Cosine similarity of fingerprints and Intersection over Union (IoU).