基于SLAM中视觉与IMU传感器融合方法综述

Overview of Visual and IMU Sensor Fusion Methods in SLAM

同步定位与建图(Simultaneous Localization And Mapping,SLAM)技术已经成为移动机器人领域重要的导航方式,视觉传感器由于便宜的价格和丰富的图像信息,让视觉同步定位与建图(Visual Simultaneous Localization And Mapping,V-SLAM)技术受到研究者广泛关注。为提升V-SLAM的综合性能,多种传感器信息融合的V-SLAM得到了快速发展。归纳并整理几何变换法的V-SLAM中各关键环节,详细阐述了SLAM中视觉与IMU传感器融合方法,对松融合与紧拟合两种主流方法进行论述,并对一些具有代表性的传统视觉惯性融合的SLAM方案进行讨论分析。最后,总结了视觉惯性融合的SLAM方案的关键问题和未来的发展方向。

Simultaneous Localization and Mapping(SLAM)has become an important navigation method in the field of mobile robotics.Visual sensors,due to their low cost and rich image information,have attracted extensive attention in Visual SLAM(V-SLAM)research.To improve the overall performance of V-SLAM,V-SLAM with sensor fusion of multiple sensors has been rapidly developed.This article summarizes and organizes the key components of geometric transformation-based V-SLAM,and provides a detailed explanation of the fusion methods between visual and IMU sensors in SLAM.It discusses the two mainstream methods of loosely-coupled and tightly-coupled fusion,and analyzes some representative traditional visual-inertial fusion SLAM solutions.Finally,the article summarizes the key issues and future development directions of visual-inertial fusion SLAM solutions.