计算机工程与应用 ›› 2022, Vol. 58 ›› Issue (10): 178-184.DOI: 10.3778/j.issn.1002-8331.2011-0343

• 模式识别与人工智能 • 上一篇    下一篇

无人机目标实时自适应跟踪系统

严飞,马可,刘佳,刘银萍,夏金锋   

  1. 1.南京信息工程大学 自动化学院,南京 210044
    2.江苏省大气环境与装备技术协同创新中心,南京 210044
    3.南京信息工程大学 大气物理学院,南京 210044
  • 出版日期:2022-05-15 发布日期:2022-05-15

UAV Target Real-Time Adaptive Tracking System

YAN Fei, MA Ke, LIU Jia, LIU Yinping, XIA Jinfeng   

  1. 1.School of Automation, Nanjing University of Information Science & Technology, Nanjing 210044, China
    2.Jiangsu Collaborative Innovation Center for Atmospheric Environment and Equipment Technology, Nanjing 210044, China
    3.School of Atmospheric Physics, Nanjing University of Information Science & Technology, Nanjing 210044, China
  • Online:2022-05-15 Published:2022-05-15

摘要: 随着无人机(unmanned aerial vehicle,UAV)在航拍、空中侦察等相关领域被广泛应用,对于无人机的智能化需求逐渐提高。目标跟踪具有信息量大、实时性高等优点,能够为无人机的智能飞行提供大量且实时的外部信息。进行低开销、低功耗的无人机目标跟踪系统的研究,对无人机智能化进程的加速具有深远意义。为更好解决跟踪过程中出现的对背景复杂、光照变化、目标较小、移动速度快等情景的目标跟踪准确度和实时性等问题,提出一种基于FPGA的自适应跟踪算法。将视频RGB数据流进行灰度转换;将模板数据和搜索区域的像素灰度值保存于片内RAM,搜索区域内遍历寻找与模板相似度最高的区域;当背景或光线发生变化,采用更新模板的方式,进一步提高跟踪准确度。仿真结果显示所提跟踪算法具有较好的实时性,跟踪重叠率在95%以上,跟踪速率在50?frame/s以上,有较好的跟踪准确度和实时性。

关键词: 现场可编程门阵列(FPGA), 目标跟踪, 自适应, 图像储存, 图像处理

Abstract: As unmanned aerial vehicle(UAV) is widely used in aerial photography, aerial reconnaissance and other related fields, the demand for intelligence of UAV gradually increases. Target tracking has the advantages of large amount of information and high real-time performance, which can provide a lot of real-time external information for the intelligent flight of UAV. The research of UAV target tracking system with low cost and power consumption has far-reaching significance for the acceleration of UAV intelligent process. In order to better solve the problems of target tracking accuracy and real-time performance in the process of tracking such situations as complex background, light change, small target and fast moving speed, this paper proposes an adaptive tracking algorithm based on FPGA. Firstly, the video RGB data stream is converted into gray scale. Secondly, the template data and the pixel gray value of the search area are saved in RAM, and the region with the highest similarity to the template is traversed within the search area. Finally, when the background or light changes, the template is updated to further track the target accurately. Simulation results show that the proposed tracking algorithm has good real-time performance, with tracking overlap rate above 95% and tracking rate above 50 frame/s, with good tracking accuracy and real-time performance.

Key words: field programmable gate array(FPGA), target tracking, self-adaptive, image storage, image processing