计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (7): 20-28.DOI: 10.3778/j.issn.1002-8331.1712-0279

• 热点与综述 • 上一篇    下一篇

单目视觉里程计研究综述

祝朝政1,何  明1,杨  晟2,吴春晓1,刘  斌1   

  1. 1.中国人民解放军陆军工程大学 指挥控制工程学院,南京 211117
    2.河海大学 计算机与信息学院,南京 211100
  • 出版日期:2018-04-01 发布日期:2018-04-16

Survey of monocular Visual Odometry

ZHU Chaozheng1, HE Ming1, YANG Sheng2, WU Chunxiao1, LIU Bin1   

  1. 1.College of Command Control Engineer, Army Engineering University, Nanjing 211117, China
    2.College of Information and Computer, HoHai University, Nanjing 211110, China
  • Online:2018-04-01 Published:2018-04-16

摘要: 单目视觉里程计不仅能为移动机器人提供导航避障等功能,在无人驾驶等领域中也有更广阔的应用价值。剖析了视觉里程计的基础原理,研究了单目视觉里程计技术的国内外现状;同时,对ORB-SLAM2、DSO等典型单目视觉里程计进行深入分析和比较。针对当前视觉里程计研究中普遍关注的鲁棒性和实时性等问题,探讨了下一步研究工作的发展方向和趋势。

关键词: 视觉里程计, 特征点法, 直接法, 视觉SLAM

Abstract: Monocular Visual Odometry can not only provide functions such as navigation and obstacle avoidance for mobile robots, but also can be applied to the current hot autopilot and other fields. This paper analyzes the basic principle of Visual Odometry, and studies the research status of monocular Visual Odometry technology at home and abroad. At the same time, the classical monocular Visual Odometries are analyzed and compared in detail, such as ORB-SLAM2, DSO and so on. Aiming at the issues of robustness and real-time which are generally concerned in the current visual odometry research, the development directions and trend of the next research work are discussed.

Key words: monocular Visual Odometry, feature method, direct method, visual SLAM