袁夏, 赵春霞. 一种基于点—面匹配的3D-SLAM方法[J]. 机器人, 2011, 33(2): 215-221.
引用本文: 袁夏, 赵春霞. 一种基于点—面匹配的3D-SLAM方法[J]. 机器人, 2011, 33(2): 215-221.
YUAN Xia, ZHAO Chunxia. 3D-SLAM Based on Point-plane Matching[J]. ROBOT, 2011, 33(2): 215-221.
Citation: YUAN Xia, ZHAO Chunxia. 3D-SLAM Based on Point-plane Matching[J]. ROBOT, 2011, 33(2): 215-221.

一种基于点—面匹配的3D-SLAM方法

3D-SLAM Based on Point-plane Matching

  • 摘要: 提出一种基于点一面匹配的3维空间同步定位与3维地图创建(3D-SLAM)方法以解决3D-SLAM中的点云匹配问题.首先将3维空间中的6自由度(6DOF)匹配问题合理地简化成5DOF匹配问题,然后算法在激光雷达获取的每行数据中提取平面拐点,再通过区域生长的方式找到点云中的平面区域.通过计算平面的法向量,并比较两帧数据之间关联平面的法向量方向估计出旋转参数,然后算法利用一种改进的层次投影方法计算平多参数.在含有较高噪声的真实数据集上的实验证明该算法是有效的.

     

    Abstract: A point-plane based point cloud matching algorithm is proposed to deal with point cloud matching problem in 3D simultaneous localization and mapping(3D-SLAM).The 6DOF matching problem in 3D space are logically simplified as a 5DOF problem firstly.And then,the algorithm extracts break point in each row of laser data and employs an area growth method to find planes in point cloud.Normal vectors of planes are computed.The rotation of two frames can be estimated by comparing normal vectors of two associated planes in two frames.An improved leveled map algorithm is used to compute the translation parameter.Experiments on real data set containing high noise validate the proposed 3D-SLAM method.

     

/

返回文章
返回