基于三维正态分布变换的地面与SLAM点云配准

Terrestrial and SLAM point cloud registration based on 3D normal distribution transform

地下空间信息采集作为城市信息化、数字化建设的基础工作日趋重要,但单一平台传感器难以获取全要素位置信息,为此,本文将地面与移动激光扫描系统应用于地下空间测量领域。首先,针对场景特征相似、SLAM激光点云定位精度较差的问题,将其按照采集时间进行分段;其次,以初始点云块为基准,利用改进的三维正态分布变换(3D-NDT)算法实现局部点云块间的精确配准;然后,针对多平台点云的空间基准、观测视角等差异,通过构建距离和角度为约束条件的仿射不变量实现多平台点云的初步对齐,为其精配准提供良好的初始位置,并利用NDT算法实现多平台点云基准统一;最后,应用于地下管廊和矿山数据采集。结果表明,该方法对初始位置相差较大、特征较少的激光点云有良好的适用性,能够获取地下空间完整的位置信息。

Underground space information collection is becoming increasingly important as the foundation of urban informatization and digitalization construction,but it is difficult for single sensor to obtain the whole location information,the terrestrial and mobile laser scanning system based on simultaneous localization and mapping(SLAM)technology is applied to the field of underground space measurement in our study.Firstly,aiming at the problem of similar scene features and poor positioning accuracy of SLAM laser point cloud,it is segmented according to the acquisition time,and the 3 D normal distribution transform(3 D-NDT)algorithm is improved to achieve accurate registration between local point cloud blocks.Then,according to the differences in the spatial reference and observation angle of the multi-platform point cloud,the coarse registration is realized by constructing the affine invariant with the distance and angle as the constraint condition,which provides good initial position for precise registration and realizes the unification of multi-platform point cloud.Finally,the method is applied to the underground pipe gallery and mine,which shows that the method has good applicability to laser point cloud with large difference in initial position and few features,and can obtain complete position information of underground space.