基于SURF特征的单目视觉SLAM方法研究

A Research on Monocular Visual SLAM Based on SURF Feature

由于视觉信息很容易受到外界环境因素的影响,因此基于视觉的移动机器人同步定位与地图构建问题所选取的特征点要求具有较高的精确度和良好的鲁棒性.针对单目SLAM问题,提出一种基于扩展卡尔曼滤波器的单目视觉SLAM算法.该算法采用SURF特征点,结合反向深度估计法,应用扩展卡尔曼滤波器融合SURF特征信息与机器人位姿信息完成SLAM过程.仿真实验结果表明,在未知室内结构化环境下,该算法运行可靠,定位精度高.

Because the visual information is easily affected by external environment factors, therefore the selected feature points of mobile robot based on visual simultaneous localization and map building requires high stability and good robustness. For the problem of monocular visual mobile robot SLAM（Simultaneous Localization and Mapping）,a kind of mono-SLAM algorithm based on Extended kalman filter is proposed by using SURF（Speed Up Robust Features） feature points and the inverse depth method. The process of SLAM is completed by fusing the information of SURF features and robot information with EKF. The result of simulation experiment indicates that the proposed algorithm is feasible, and with high localization precision in indoor structured environment.