Sea You Later: Metadata-Guided Long-Term Re-Identification for UAV-Based Multi-Object Tracking
It addresses long-term tracking challenges for UAV-based maritime computer vision, representing an incremental improvement in a domain-specific area.
The paper tackled the problem of re-identification in multi-object tracking for UAVs in maritime scenarios, achieving state-of-the-art performance with a HOTA of 69.5% and IDF1 of 85.9% on the SeaDroneSee dataset.
Re-identification (ReID) in multi-object tracking (MOT) for UAVs in maritime computer vision has been challenging for several reasons. More specifically, short-term re-identification (ReID) is difficult due to the nature of the characteristics of small targets and the sudden movement of the drone's gimbal. Long-term ReID suffers from the lack of useful appearance diversity. In response to these challenges, we present an adaptable motion-based MOT algorithm, called Metadata Guided MOT (MG-MOT). This algorithm effectively merges short-term tracking data into coherent long-term tracks, harnessing crucial metadata from UAVs, including GPS position, drone altitude, and camera orientations. Extensive experiments are conducted to validate the efficacy of our MOT algorithm. Utilizing the challenging SeaDroneSee tracking dataset, which encompasses the aforementioned scenarios, we achieve a much-improved performance in the latest edition of the UAV-based Maritime Object Tracking Challenge with a state-of-the-art HOTA of 69.5% and an IDF1 of 85.9% on the testing split.