航空学报 > 2026, Vol. 47 Issue (4): 332264-332264   doi: 10.7527/S1000-6893.2025.32264

复杂背景下反无人机红外目标鲁棒跟踪算法

冯子成1,2, 张文龙1,2(), 刘冬辉1,2, 于起峰1,2   

  1. 1.国防科技大学 空天科学学院,长沙 410073
    2.图像测量与视觉导航湖南省重点实验室,长沙 410073
  • 收稿日期:2025-05-20 修回日期:2025-07-14 接受日期:2025-08-19 出版日期:2025-09-08 发布日期:2025-09-05
  • 通讯作者: 张文龙 E-mail:wenlong@nudt.edu.cn
  • 基金资助:
    国家自然科学基金(12202485)

Robust infrared target tracking algorithm for anti-UAV in complexbackgrounds

Zicheng FENG1,2, Wenlong ZHANG1,2(), Donghui LIU1,2, Qifeng YU1,2   

  1. 1.College of Aerospace Science and Engineering,National University of Defense Technology,Changsha 410073,China
    2.Hunan Provincial Key Laboratory of Image Measurement and Vision Navigation,Changsha 410073,China
  • Received:2025-05-20 Revised:2025-07-14 Accepted:2025-08-19 Online:2025-09-08 Published:2025-09-05
  • Contact: Wenlong ZHANG E-mail:wenlong@nudt.edu.cn
  • Supported by:
    National Natural Science Foundation of China(12202485)

摘要:

面向复杂背景下反无人机任务,红外目标跟踪面临着目标多尺度、遮挡、移出视野、背景干扰等诸多挑战。针对这些问题,提出了一种红外无人机鲁棒跟踪算法。首先,提出了一个全局跟踪模型,采用高效的一阶段无锚框框架,通过全局搜索以应对因遮挡和移出视野导致的无人机目标消失问题,并使用多层级结构跟踪不同尺寸的目标。其次,提出一个基于记忆网络的时空特征融合模块,利用视频中的时空域特征来提高目标的可判别性。然后,提出了一个目标增强与干扰抑制模块,通过记录和匹配帧间的目标和干扰特征以生成权重,并与得分图加权以提升算法的抗背景干扰能力。最后,提出了一个动态层级加速方法,通过删除冗余层级以提高运行效率。实验结果表明:该算法在2nd和3rd Anti-UAV数据集上分别取得92.4%和78.7%的精度以及69.4%和56.5%的成功率,并能够以26.9帧/s的速度运行,其性能均明显超过了现有算法,实现了在复杂场景中对无人机的实时和鲁棒跟踪。

关键词: 红外目标跟踪, 反无人机, 复杂背景, 抗干扰, 实时跟踪, 深度学习

Abstract:

In the anti-UAV task under complex backgrounds, infrared target tracking faces numerous challenges,such as multi-scale targets, occlusion, moving out of view, and background interference. To address these problems,a robust infrared UAV tracking algorithm is proposed. First, a global tracking model is designed, which adopts an effi⁃cient one-stage anchor-free framework to perform global search for handling target disappearance caused by occlusionor moving out of view, while employing a multi-level structure to track targets of varying sizes. Second, a spatiotemporalfeature fusion module based on the memory network is proposed, which leverages spatio-temporal featuresfrom video sequences to enhance target discriminability. Third, a target enhancement and interference suppressionmodule is introduced, which records and matches target and interference features between frames to generate weight⁃ing maps. These maps are applied to the score maps to improve the algorithm’s anti-interference capability. Finally, adynamic hierarchical acceleration method is proposed to improve the running efficiency by removing redundant hierar⁃chical computations. Experimental results demonstrate that the proposed algorithm achieves 92. 4% and 78. 7% preci⁃sion and 69. 4% and 56. 5% success rates on the 2nd and 3rd Anti-UAV datasets, respectively, with a running speedof 26. 9 framen per second, which significantly outperforms existing methods and achieves real-time and robust UAVtracking in complex scenarios

Key words: infrared target tracking, nti-UAV, complex background, anti-interference, real-time tracking, deep learning

中图分类号: