航空学报 > 2022, Vol. 43 Issue (9): 25882-025882   doi: 10.7527/S1000-6893.2021.25882

无人机自主降落标识检测方法若干研究进展

赵良玉1, 李丹1, 赵辰悦1, 蒋飞2   

  1. 1. 北京理工大学 宇航学院, 北京 100081;
    2. 武警研究院, 北京 100012
  • 收稿日期:2021-05-31 修回日期:2021-06-21 出版日期:2022-09-15 发布日期:2021-07-09
  • 通讯作者: 赵良玉,E-mail:zhaoly@bit.edu.cn E-mail:zhaoly@bit.edu.cn
  • 基金资助:
    国家自然科学基金(12072027,11532002)

Some achievements on detection methods of UAV autonomous landing markers

ZHAO Liangyu1, LI Dan1, ZHAO Chenyue1, JIANG Fei2   

  1. 1. School of Aerospace Engineering, Beijing Institute of Technology, Beijing 100081, China;
    2. Institute of People's Armed Police, Beijing 100012, China
  • Received:2021-05-31 Revised:2021-06-21 Online:2022-09-15 Published:2021-07-09
  • Supported by:
    National Natural Science Foundation of China (12072027, 11532002)

摘要: 为了进一步促进中国海/陆空跨域协同技术的研究和发展, 综述了无人机自主降落标识检测方法的国内外最新研究成果。首先, 在分析视觉引导无人机自主降落流程的基础上, 简要总结了常用的基于图像分割、基于分类器和基于深度学习的标识检测方法。然后, 介绍了无人机自主降落于静平台和车辆、舰艇等动平台的国内外若干研究团队及成果, 并对团队采用的降落标识及检测方法进行了梳理。最后, 围绕动平台及复杂环境下的标识检测和相关软件算法、硬件设备、多传感器融合等讨论了当前存在的难点和可行的解决方案, 对未来克服人工标识依赖性, 采用深度学习思想进行非合作环境下的安全区域检测方法进行了展望。

关键词: 跨域协同, 无人机, 自主降落, 标识检测, 深度学习

Abstract: To further promote the research and development on sea/land-air cross-domain collaborative technology in China, the main research achievements and the latest progresses of autonomous landing marker detection methods for the Unmanned Aerial Vehicle (UAV) are reviewed. Firstly, following the introduction of vision guided UAV autonomous landing, the marker detection methods based on image segmentation, classifier and deep learning are discussed. Secondly, the overseas and domestic research teams and achievements of autonomous landing of UAV on static and moving platforms such as vehicles and ships are introduced. Landing markers and detection methods used by the teams are summarized. Finally, a number of key technical issues and feasible solutions for further investigations are discussed in terms of marker detection on moving platforms and in complex environments, system software algorithms, hardware equipment and multi-sensor fusion. How to overcome the dependence on artificial marker and use deep learning ideas for detection of safe landing areas in non-cooperative environments in the future are also discussed.

Key words: cross-domain collaboration, UAV, autonomous landing, marker detection, deep learning

中图分类号: