Improving tracking of small fast objects in 360 degree wildlife videos
Re-engineering SORT-based algorithms for low-cost small object tracking from omnidirectional footage
Computer Vision and Pattern Recognition
Summary
Tracking small, fast-moving animals in videos taken with 360-degree cameras can be tricky because the special camera view creates unusual visual breaks and makes objects look very small. The authors changed an existing tracking method called SORT to work better with these cameras by making it aware of the camera’s unique image shape and how objects move across image edges. They also introduced a new collection of wildlife videos to test their method. Their changes helped track flying animals more accurately without needing complicated appearance-based calculations, and the system can run efficiently on ordinary computers.
multi-object trackingSORT algorithmequirectangular projectionomnidirectional cameraKalman filterHOTA metricMOTA metricIDF1 metricYOLOX detector
Authors
Xin Shu, Meegan Gower, Yvonne Buckley, Anil Kokaram
Abstract
Multi-object tracking (MOT) has advanced rapidly in urban surveillance and autonomous driving, yet many trackers rely on ReID- and transformer-based appearance encoders and are designed for standard FoV cameras. These assumptions break down for low-cost omnidirectional deployments, where equirectangular projection introduces seam discontinuities and targets appear to be small and fast-moving. We address multi-object tracking of flying animals captured in remote environments using omnidirectional cameras. We propose a lightweight framework that re-engineers SORT-based tracking for this geometry, including (i) a Seam-Aware Motion Model that keeps the Kalman state continuous across the seam, (ii) a composite seam-aware association cost that pairs a wrapped Euclidean term with GIoU, and (iii) OmniSmall, a new benchmark of omnidirectional wildlife footage. On our new dataset, with ground-truth detections, our modifications improved over OCSORT by +8.51 HOTA, +9.41 MOTA, and +10.17 IDF1; with YOLOX detections the gain narrows to +1.95 HOTA. Our proposed methods improved tracking performance on OmniSmall and remained competitive on JRDB without adding appearance encoders while keeping the tracking stage CPU-only. Our dataset and source code are available at: https://github.com/Xin-Shu/OmniSORT.git.