视觉及其融合惯性的SLAM技术发展综述

Review on SLAM Technology Development for Vision and its Fusion of Inertial Information

当前视觉同时定位与建图(Simultaneous localization and mapping,SLAM)技术已经成为导航定位领域重要的导航方式之一。为提升视觉SLAM全天候多场景的综合性能,融合多种信息的视觉SLAM技术得到了快速发展。本文系统综述了近年国内外视觉SLAM的研究成果。首先归纳整理了视觉及视觉/惯性SLAM中的前端里程计、后端优化、回环检测和地图构建关键环节。在纯视觉SLAM技术分析方面,围绕传统基于几何变换和现代基于深度学习的两大类视觉SLAM方法开展讨论,并梳理了相关具有代表性的视觉SLAM技术。在详细阐述视觉/惯性组合导航的多信息融合SLAM方法基础上,分析了近年基于异源图像的视觉及其与惯性组合的SLAM导航技术进展。最后对视觉SLAM的未来发展方向进行了针对性展望。

Visual simultaneous localization and mapping(SLAM)technology has become one of the important navigation methods in the field of navigation and positioning.In order to improve the comprehensive performance of visual SLAM in all-weather and multi-scene environments,the visual SLAM technology integrating various information has been developed rapidly.This paper systematically reviews the research results of visual SLAM in recent years.First,the key links of front-end odometry,back-end optimization,loop closure detection and map construction in visual and visual/inertial SLAM are summarized.Second,in the analysis of pure visual SLAM technology,the two main types of visual SLAM methods based on traditional geometric transformation and modern deep learning are discussed,and related representative visual SLAM technologies are sorted out.Third,based on the detailed description of the multi-information fusion SLAM method of visual/inertial combined navigation,the progress of SLAM navigation technology based on heterogeneous image vision and its combination with inertial sensor in recent years is analyzed.Finally,the future development direction of visual SLAM is prospected.