基于注意力特征融合的无人机多目标跟踪算法
作者:
作者单位:

北京工业大学信息学部

作者简介:

通讯作者:

中图分类号:

V279;TP391

基金项目:

国家自然科学基金(基金号61171119)


UAV Multi Target Tracking Algorithm Based on Attention Feature Fusion
Author:
Affiliation:

Beijing University Of Technology, Faculty of Information Technology

Fund Project:

National Natural Science Foundation of China(No.61171119)

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    随着无人机技术的不断发展,无人机多目标跟踪已成为无人机应用的关键技术之一。针对无人机视频中的复杂背景干扰、遮挡、视点高度和角度多变等问题,提出一种基于注意力特征融合的无人机多目标跟踪算法。首先,将改进的卷积注意力模块引入残差网络,建立了三元组注意力特征提取网络;其次,在特征金字塔网络的结构上加入了新的特征融合通道,设计了多尺度特征融合模块,增强了模型对多尺度目标的特征表达能力;最后,根据目标的重识别特征匹配与检测框匹配得到目标轨迹。仿真实验结果表明,该算法有效提升了无人机多目标跟踪的精度,具有较好的鲁棒性。

    Abstract:

    With the development of UAV technology, multi-target tracking of UAV video has become one of the key technologies in the application of UAV. Aiming at the problem of complex background interference, occlusion, variable viewpoint of height and angle in UAV multi-target tracking video, a multi-target tracking algorithm based on attention feature fusion is proposed. Firstly, the improved convolution attention module is introduced into the residual network, and a three-tuple attention feature extraction network is established. Secondly, a new feature fusion channel is added to the structure of the feature pyramid network, and a multi-scale feature fusion module is designed to enhance the model"s ability to express the features of multi-scale targets. Finally, the target trajectory is obtained by target Re-Identification feature matching and bounding box matching. The simulation results show that the algorithm effectively improves the accuracy and robustness of the unmanned multi-target tracking.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2021-06-24
  • 最后修改日期:2021-10-11
  • 录用日期:2021-10-27
  • 在线发布日期: 2021-12-01
  • 出版日期: