摘要
视觉同时定位与地图构建(V-SLAM)在机器人、无人机导航、自动驾驶等领域有着广泛的研究。直接法V-SLAM基于环境亮度不变性假设,跟踪相机的位姿并构建环境地图。针对直接法V-SLAM,首先简述其基本原理;然后分析、比较几种具有代表性的直接法V-SLAM系统;最后讨论直接法的优缺点和发展趋势,并进行了总结和展望。
Visual simultaneous localization and mapping has a wide range of research in the fields of robot,unmanned aerial vehicle navigation,automatic driving,et al.Direct-method V-SLAM,based on the assuption of photometric consistency,tracks the camera's position and orientation and builds an environment map.For the direct-method V-SLAM,this paper summarized the basic principles,and introduced some state-of-art direct-method V-SLAM systems with analysis and comparison.Finally,it discussed the advantages and disadvantages of direct-method V-SLAM and some research tendency,as well as making conclusion.
作者
潘林豪
田福庆
应文健
邱千钧
Pan Linhao;Tian Fuqing;Ying Wenjian;Qiu Qianjun(College of Weapon,Naval Engineering University,Wuhan 430033,China;Naval Representative Office in the 2nd Academy of CASA,Beijing 100854,China)
出处
《计算机应用研究》
CSCD
北大核心
2019年第4期961-966,共6页
Application Research of Computers
基金
海军工程大学自主立项基金资助项目
关键词
计算机视觉
同时定位与地图构建
直接法
运动推断结构
多视图几何
computer vision
simultaneous localization and mapping
direct-method
structure-from-motion
multiple-view geometry