FC-Track: Overlap-Aware Post-Association Correction for Online Multi-Object Tracking

Abstract

Reliable multi-object tracking (MOT) is essential for robotic systems operating in complex and dynamic environments. Despite recent advances in detection and association, online MOT methods remain vulnerable to identity switches caused by frequent occlusions and object overlap, where incorrect associations can propagate over time and degrade tracking reliability. We present a lightweight post-association correction framework (FC-Track) for online MOT that explicitly targets overlap-induced mismatches during inference. The proposed method suppresses unreliable appearance updates under high-overlap conditions using an Intersection over Area (IoA)-based filtering strategy, and locally corrects detection-to-tracklet mismatches through appearance similarity comparison within overlapped tracklet pairs. By preventing short-term mismatches from propagating, our framework effectively mitigates long-term identity switches without resorting to global optimization or re-identification. The framework operates online without global optimization or re-identification, making it suitable for real-time robotic applications. We achieve 81.73 MOTA, 82.81 IDF1, and 66.95 HOTA on the MOT17 test set with a running speed of 5.7 FPS, and 77.52 MOTA, 80.90 IDF1, and 65.67 HOTA on the MOT20 test set with a running speed of 0.6 FPS. Specifically, our framework FC-Track produces only 29.55% long-term identity switches, which is substantially lower than existing online trackers. Meanwhile, our framework maintains state-of-the-art performance on the MOT20 benchmark.

Figures (5)

• Figure 1: Samples from the results on MOT17 validation set. TrackTrack and our proposed FC-Track use the same detection results. For clarity, we show only a single representative example. On the third frame, TrackTrack exhibits an ID switch when the target overlaps with another target person, our FC-Track maintains consistent tracking throughout the occlusion.
• Figure 2: System overview of the proposed post-association correction framework FC-Track. Given matched tracklet–detection pairs at frame $f$, associations are first divided into overlapped and non-overlapped groups using overlap status inherited from frame $f-1$. Non-overlapped pairs are directly accepted, while overlapped pairs are re-evaluated through appearance similarity comparison using stored features and current detections appearance feature. Corrected matches are merged to produce final tracking results. Then the IoA of all tracklets are computed to save as overlap status, and appearance features are updated for the next frame, enabling temporal consistency and online error correction.
• Figure 3: Concept of Intersection over Area (IoA). The overlapped green region represents the intersection between two bounding boxes, while IoA is computed as the ratio of the intersection area to the area of the reference bounding box. Since either box can serve as the reference, two IoA values can be obtained for a box pair.
• Figure 4: Comparison of the performances of FC-Track over different minimum correction similarity distance threshold $\tau_{min}$ and similarity distance difference threshold $\tau_{dif}$. The results are evaluated on the validation set of MOT17.
• Figure 5: Comparison of the performances of FC-Track over different update IoA threshold $\tau_{update}$ and overlap IoA threshold $\tau_{overlap}$. The results are evaluated on the validation set of MOT17.