基于点线特征的视觉同时定位与地图构建综述

A survey of visual simultaneous localization and mapping based on point-line features

同时定位与地图构建(simultaneous localization and mapping,SLAM)作为计算机视觉中的热门方向,在无人驾驶、移动机器人等领域中发挥着重要的作用。由于线特征在低纹理环境下的优势,越来越多的研究人员利用点线特征融合的方法提高SLAM系统的精度和鲁棒性。文中首先简要阐述了传统的点特征SLAM系统在低纹理环境下的局限性,并对现有的视觉SLAM综述文献进行了总结;随后,对经典的点线SLAM方案进行了介绍,并总结了点线特征融合在前端、后端、闭环检测中的研究进展;最后,对点线SLAM未来的发展方向进行了展望。

As a popular subject in computer vision,simultaneous localization and mapping(SLAM)plays an important role in self-driving,mobile robot and other fields.Due to the advantages of line features in low-texture environments,more and more researchers use point-line feature fusion methods to improve the accuracy and robustness.Firstly,this paper briefly describes the limitations of the traditional SLAM system with point feature in low texture environment,and summarizes the existing review;Then,the classical SLAM based on point-line features is introduced,and the research progress of point line feature fusion in front-end,back-end and closed loop is summarized;Finally,the future development direction of point-line SLAM is prospected.