Toward Enhanced Tracking of Tiny Airborne Objects Using YOLOv8 and Particle Filter
Jeonghun Lee, Sungwook Cho, Heemin Shin, Seung-Bum Kim · 2025
The rapid proliferation of civilian drones worldwide has led to numerous issues, including challenges in air traffic management and public safety. In particular, the misuse of illegal drones for unethical activities such as terrorism poses a significant threat, potentially causing catastrophic damage. To mitigate these risks, continuous detection and tracking of drones are essential. Most civilian drones are classified as tiny airborne vehicles, appearing very small in the sky and exhibiting erratic movements, which complicates sustained detection and tracking. To address these challenges, this paper proposes a method that utilizes deep learning object detection technology to extract feature information of objects and continuously detect and track them through probabilistic state estimation. First, the YOLOv8 deep learning model is employed for real-time object detection. The YOLOv8 model effectively detects small, fast-moving airborne objects and acquires their feature information. Subsequently, a particle filter is utilized to track the detected airborne objects by estimating their probabilistic states based on the obtained feature information. Finally, the Hungarian algorithm ensures the optimal matching between detected objects and their tracked states. Experimental results demonstrate that the proposed method significantly improves tracking performance, especially in scenarios involving long distances and rapid movements, compared to existing methods. Consequently, this paper presents a robust and efficient approach for the detection and tracking of tiny airborne objects.