CAJ | 학술논문

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。
유우시각신식흔용역수도외계배경인소적영향，인차기우시각적이동궤기인동보정위여지도구건문제소선취적특정점요구구유교고적정학도화량호적로봉성。침대단목SLAM문제，제출일충기우확전잡이만려파기적단목시각SLAM산법。해산법채용SURF특정점，결합반향심도고계법，응용확전잡이만려파기융합SURF특정신식여궤기인위자신식완성SLAM과정。방진실험결과표명，재미지실내결구화배경하，해산법운행가고，정위정도고。
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.