Scholarlink

<p>Unmanned aerial vehicles (UAVs) have become important platforms for multi-object tracking. In this paper, we proposed Quick-UAV OCSORT, a fast and robust tracking-by-detection (TBD) algorithm that extends the Observation-centric SORT (OC-SORT) framework. Our method introduced a lightweight LightMBN network for enhanced re-identification, a pre-matching strategy to reduce redundant ReID calls, an improved B-IoU + algorithm for robust small object matching, and a Camera motion compensation (CMC) module to mitigate camera jitter. Experiments on the VisDrone2019 dataset demonstrate improvements of 0.88% in MOTA, 5.51% in IDF1, and 3.01% in HOTA, while maintaining an average processing speed of 19.58 FPS.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quick-UAV OCSORT: a fast and robust UAV-based tracking-by detection algorithm

  • 期刊论文

  • 发表日期：

    2025年5月12日

查看全文
  • Peng He,
  • Zhihong Zhao,
  • Xinyue Liu

摘要

Unmanned aerial vehicles (UAVs) have become important platforms for multi-object tracking. In this paper, we proposed Quick-UAV OCSORT, a fast and robust tracking-by-detection (TBD) algorithm that extends the Observation-centric SORT (OC-SORT) framework. Our method introduced a lightweight LightMBN network for enhanced re-identification, a pre-matching strategy to reduce redundant ReID calls, an improved B-IoU + algorithm for robust small object matching, and a Camera motion compensation (CMC) module to mitigate camera jitter. Experiments on the VisDrone2019 dataset demonstrate improvements of 0.88% in MOTA, 5.51% in IDF1, and 3.01% in HOTA, while maintaining an average processing speed of 19.58 FPS.