航空学报 > 2026, Vol. 47 Issue (4): 331967-331967   doi: 10.7527/S1000-6893.2025.31967

无人机视频多目标特征关联技术研究进展

伍瀚1,2, 孙浩1,2, 刘奎1,2, 计科峰1,2(), 匡纲要1,2   

  1. 1.国防科技大学 电子科学学院,长沙 410073
    2.国防科技大学 电子信息系统复杂电磁环境效应国家重点实验室,长沙 410073
  • 收稿日期:2025-03-12 修回日期:2025-03-29 接受日期:2025-05-28 出版日期:2025-06-10 发布日期:2025-06-06
  • 通讯作者: 计科峰 E-mail:jikefeng@nudt.edu.cn
  • 基金资助:
    国家自然科学基金(61971426)

Multi-object feature association in UAV videos: Recent progress and perspectives

Han WU1,2, Hao SUN1,2, Kui LIU1,2, Kefeng JI1,2(), Gangyao KUANG1,2   

  1. 1.College of Electronic Science and Technology,National University of Defense Technology,Changsha 410073,China
    2.State Key Laboratory of Complex Electromagnetic Environment Effects on Electronics and Information System,National University of Defense Technology,Changsha 410073,China
  • Received:2025-03-12 Revised:2025-03-29 Accepted:2025-05-28 Online:2025-06-10 Published:2025-06-06
  • Contact: Kefeng JI E-mail:jikefeng@nudt.edu.cn
  • Supported by:
    National Natural Science Foundation of China(61971426)

摘要:

无人机视频已成为智能监控、智慧城市、态势感知、低空经济以及军事侦察等军民用领域不可或缺的信息来源。无人机视频多目标特征关联旨在持续预测各目标位置并维持其身份标识,是多目标跟踪等任务的核心。目前,相关综述多聚焦于目标检测与跟踪,本研究对无人机视频多目标特征关联技术研究进展进行系统综述。首先,归纳梳理无人机视频多目标特征关联典型研究成果,并根据应用场景和数据源特性对其进行分类,涵盖了多视角和多光谱特征关联相关研究成果。其次,深入分析各类方法的典型算法、优缺点及适用场景。然后,总结整理无人机视频多目标特征关联的主流公开数据集,包括单视频数据集、多视角视频数据集以及多光谱视频数据集,并基于VisDrone、MDMT和VT-Tiny-MOT 3个典型数据集,对现有代表性方法的性能等进行系统对比,分析不同方法之间性能差异的根本原因,为后续研究奠定基础。最后,探讨无人机视频特征关联面临的挑战与未来的研究方向,特别是基础模型构建与多模态深度融合等,以期为无人机视频多目标特征关联技术的深入研究提供参考。

关键词: 无人机视频, 特征关联, 多目标跟踪, 多视角视频, 多光谱视频

Abstract:

Unmanned Aerial Vehicle (UAV) videos have become essential sources of information in both civilian and military domains, including intelligent surveillance, smart cities, situational awareness, low-altitude economy and military reconnaissance. Multi-object feature association in UAV videos aims to continuously predict target positions and maintain the identity of each target, serving as the foundation for tasks such as multi-object tracking. However, existing reviews predominantly focus on UAV object detection and tracking, lacking a systematic review for multi-object feature association in UAV videos. This paper provides the first systematic review of the research progress on multi-object feature association in UAV videos. First, existing methods are summarized and categorized based on application scenarios and data source characteristics, which covers multi-view and multi-spectral feature association approaches for the first time. Then, the representative algorithms are analyzed in depth, including their strengths, limitations, and applicable scenarios. In addition, mainstream public datasets used in this research field are summarized, including single-view, multi-view, and multi-spectral UAV video datasets. Representative datasets such as VisDrone, MDMT, and VT-Tiny-MOT are selected to evaluate and compare existing methods, with the purpose of analyzing the root causes of the performance differences among existing methods and laying the foundation for subsequent studies. Finally, the paper highlights the key challenges that remain in UAV multi-object feature association and discusses future research directions, particularly in the areas of foundation model development and multi-modal deep fusion. This review aims to provide valuable insights for advancing research in this field.

Key words: UAV videos, feature association, multi-object tracking, multi-view videos, multi-spectral videos

中图分类号: