视觉里程计综述 |
投稿时间:2019-11-13 修订日期:2020-04-17 点此下载全文 |
引用本文: |
摘要点击次数: 487 |
全文下载次数: 0 |
|
基金项目:国家自然科学基金(61773219, 61701244),国家重点研发计划重点专项课题(2018YFC1405703 |
|
中文摘要:视觉里程计(Visual Odometry)作为视觉同步定位与地图构建技术(Visual Simultaneous Localization and Mapping)的一部分,主要通过相机传感器获取一系列拥有时间序列图像的信息,从而预估机器人的姿态信息,建立局部地图,也被称为前端,已经被广泛应用在了多个领域,并取得了丰硕的实际成果。对于无人驾驶,全自主无人机,虚拟现实和增强现实等方面有着重要意义。本文在介绍了经典视觉里程计技术框架模块中的各类算法的基础上,对近年来新颖的视觉里程计技术(VO)的研究和论文进行了总结,按照技术手段不同分为两大类——多传感器融合的视觉里程计(以惯性视觉融合为例)和基于深度学习的视觉里程计。前者通过各传感器之间的优势互补提高VO的精度,后者则是通过和深度学习网络结合。最后比较视觉里程计现有算法,并结合VO面临的挑战展望了视觉里程计的未来发展趋势。 |
中文关键词:视觉里程计 多传感器融合 深度学习 |
|
A survey of visual odometry |
|
|
Abstract:Visual odometry(VO) which is a part of Visual Simultaneous Localization and map building technology is mainly used to estimate the robot's pose and the local map through a series images captured by camera, also known as the front end, VO has been widely used in many fields and achieved fruitful practical results. It is of great significance for unmanned driving, autonomous drones, virtual reality and augmented reality. Based on the introduction of various algorithms in the framework module of the classical visual odometer technology, this paper summarizes the research and papers on the novel visual odometer technology (VO) in recent years, according to various technical means the novel methods are divided into two categories - multiple sensor fusion of visual odometer (Take VIO as an example) and the visual odometry which is based on the deep learning. The former improves the accuracy of VO by complementing the advantages of various sensors, while the latter combines with deep learning network. Finally, the existing algorithms of visual odometer are compared, meanwhile the future development trend of visual odometry is forecasted based on the challenges faced by VO. |
keywords:Visual odometry Multi-sensor fusion Deep learning. |
查看/发表评论 下载PDF阅读器 |