基于前后端图优化的RGB-D三维SLAM  

Three-dimensional RGB-D SLAM based on graph optimization

作　　者：邢科新[1] 竺海光 林叶贵[1] 张雪波[2] XING Kexin;ZHU Haiguang;LIN Yegui;ZHANG Xuebo(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023,China;China Institute of Robotics and Automatic Information System,Nankai University,Tianjin 300071,China)

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310023 [2]南开大学机器人与信息自动化研究所,天津300071

出　　处：《浙江工业大学学报》2018年第6期616-621,共6页Journal of Zhejiang University of Technology

基　　金：国家自然科学基金资助项目(61202203)

摘　　要：针对传统滤波方法在解决移动机器人同时定位与地图构建(SLAM)中存在的累积误差问题,将图优化方法应用于前端和后端优化中,以提高移动机器人位姿估计和建图的准确性。运用ORB算法进行图像的特征提取与匹配,将图优化的方法应用到PnP问题的求解中,实现了机器人位姿的准确估计。基于词典(Bag of words)的闭环检测算法来进行闭环检测,得到存在的大回环,同时利用相邻几帧的匹配关系实时检测邻近几帧之间可能存在的局部回环。用图优化的方法对这些回环进行优化,得到准确的运动轨迹和点云地图。实验结果表明:基于前后端图优化的RGB-D三维SLAM算法,在室内环境下具有良好的精度和实时性。In order to solve accumulated errors for mobile robot in Simultaneous Localization and Map Construction(SLAM)of the traditional filtering methods,the graph optimization methods is applied in front-end and back-end optimizations in order to improve the accuracy of the pose estimation and graph building.The ORB algorithm is adopted to extract the features and match the images.To realized the accurate estimation of robot pose,the graph-based optimization is applied to solve the PnP problem.The closed loop detection algorithm based on bag-of-words algorithm is used to detect the closed loopand the existing big loop can be tested.At the same time,the matching relationships of adjacent frames are used to detect the possible local loops between neighboring frames in real time..The graph optimization methods are used to optimize the local loops and the accurate motion trajectory and point cloud map can be obtained.The experimental results shows that the algorithm has good real-time and accuracy performance in the indoor environment.

关 键 词：图优化 三维SLAM PNP 视觉词典 回环检测 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]