Multi-object tracking (MOT), a core technology in computer vision, aims to continuously identify and associate multiple targets within video sequences. In drone videos, motion blur caused by rapid camera movement severely degrades detection accuracy, often resulting in fragmented tracking trajectories. Traditional algorithms assume static cameras and linear target motion, making them unsuitable for dynamic scenes where motion coupling occurs between the camera and targets. This paper proposes a selective video deblurring-based tracking algorithm that dynamically filters blurry frames for restoration using a PSNR-based threshold while preserving the original data of clear frames. The strategy leverages optical flow analysis between adjacent frames to localize blurred regions, thereby avoiding pixel distortion caused by full-frame processing. This enhances both per-frame detection accuracy and cross-frame tracking stability. Experiments on the VisDrone dataset demonstrate that the proposed method outperforms traditional approaches in terms of core metrics such as Multiple Object Tracking Accuracy (MOTA), Multiple Object Tracking Precision (MOTP), and Identity F1 Score (IDF1), effectively mitigating identity switch issues. This approach provides an efficient solution for multi-target tracking in dynamic UAV scenarios and holds promising potential for real-time applications such as traffic monitoring and disaster rescue.