SFTrack: A Robust Scale and Motion Adaptive Algorithm for Tracking Small and Fast Moving Objects

Inpyo Song, Jangwon Lee

Abstract: This paper addresses the problem of multi-object tracking in Unmanned Aerial Vehicle (UAV) footage, which is crucial for various UAV applications such as traffic monitoring systems and real-time suspect tracking by police. However, this task is highly challenging due to the fast motion of UAVs and the small size of target objects in the videos, resulting from high-altitude and wide-angle views of drones. To tackle these challenges, we introduce a refined method that involves a new tracking strategy, which initiates the tracking of target objects even from low-confidence detections commonly encountered in UAV scenarios. Additionally, we propose revisiting traditional appearance-based matching algorithms to enhance the association of low-confidence detections. To evaluate the effectiveness of our method, we conducted benchmark evaluations on two UAV-specific datasets (VisDrone2019, UAVDT) and a general dataset (MOT17). The results demonstrate that our approach surpasses current state-of-the-art methodologies, showcasing its robustness and adaptability in diverse tracking environments. Moreover, we have improved the annotation of the UAVDT dataset by correcting several errors and addressing omissions found in the original annotations. This refined version of the dataset will be made available to facilitate better benchmarking in the field.