摘要
SLAM技术一直以来都是移动机器人实现自主导航和避障的核心,为提高特征提取和匹配的鲁棒性、抑制累积误差造成的位姿漂移,提出一种基于ORB-SLAM的算法。算法对Kinect采集得到的RGB数据进行特征点提取与匹配,在保证精度的同时,有效减少了配准算法的迭代次数,并与Kinect深度数据计算移动机器人在相邻图像间的位姿相结合;采用回环检测方法对后端进行优化,得到全局一致的轨迹与三维地图。实验结果表明该方法能有效提高SLAM系统的鲁棒性和地图构建的准确性。
SLAM has always been the core issue of autonomous navigation and obstacle avoidance for mobile robots.In order to improve the robustness of feature extraction and matching and to suppress pose drift caused by accumulated errors,an algorithm based on ORB-SLAM is proposed.The algorithm extracts and matches feature points from RGB data acquired by Kinect,which effectively reduces iteration times of registration algorithm while ensuring accuracy,and combines with Kinect depth data to calculate the pose of mobile robot between adjacent images.The backend is optimized by loop closure detection method to obtain globally consistent trajectory and 3D map.Experimental results show that the method can effectively improve the robustness of SLAM system and the accuracy of mapping.
作者
富裕
刘振宇
FU Yu;LIU Zhenyu(School of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,China)
出处
《微处理机》
2019年第4期51-56,共6页
Microprocessors
