您的位置:首页  > 论文页面

深度学习在视觉SLAM研究中的应用综述

发表时间:2019-12-31  浏览量:2447  下载量:299
全部作者: 敬学良,王晨升,杨光,李彦江
作者单位: 北京邮电大学自动化学院
摘 要: 目前,采用视觉传感器的同步定位与地图构建(simultaneous localization and mapping,SLAM)逐渐成为SLAM研究中的重点方向。视觉SLAM技术在室内自主导航和虚拟现实等领域也得到了应用。本文首先介绍了视觉SLAM的基本结构,并分析了传统特征点法和直接法的局限性;然后重点对视觉SLAM中采用深度学习方法的视觉里程计(visual odometry,VO)和回环检测的最新研究成果进行了综述,并将深度学习方法和传统方法进行了简要对比;最后对视觉SLAM的发展趋势进行了展望。
关 键 词: 人工智能;深度学习;综述;视觉同步定位与地图构建(SLAM);视觉里程计;回环检测
Title: Application review of deep learning in visual SLAM research
Author: JING Xueliang, WANG Chensheng, YANG Guang, LI Yanjiang
Organization: School of Automation, Beijing University of Posts and Telecommunications
Abstract: At present, simultaneous localization and mapping (SLAM) using visual sensors has gradually become a hot research topic in SLAM. Visual SLAM technology has also been applied in domains such as indoor autonomous navigation and virtual reality (VR). Firstly, the basic structure of visual SLAM is introduced in this paper, and the limitations of traditional feature-based method and direct method are analyzed. Then, the latest research results of visual odometry (VO) and loop closure detection using deep learning method in visual SLAM are emphatically reviewed, and the deep learning method and traditional method are briefly compared. Finally, the development trend of visual SLAM is proposed.
Key words: artificial intelligence; deep learning; review; visual SLAM (simultaneous localization and mapping); visual odometry; loop closure detection
发表期数: 2019年12月第6期
引用格式: 敬学良,王晨升,杨光,等. 深度学习在视觉SLAM研究中的应用综述[J]. 中国科技论文在线精品论文,2019,12(6):872-878.
 
6 评论数 0
暂无评论
友情链接