基于激光雷达的多视角点云配准方法  

Multiview Point Cloud Registration Method Based on Laser Radar

作　　者：耿磊[1] 曹春鹏 肖志涛[1] 张芳[1] Geng Lei;Cao Chunpeng;Xiao Zhitao;Zhang Fang(Tianjin Key Laboratory of Optoelectronic Detection Technology and System,School of Life Sciences,TianGong University,Tianjin 300387,China;Tianjin Key Laboratory of Optoelectronic Detection Technology and System,School of Electrical and Electronic Engineering,TianGong University,Tianjin 300387,China)

机构地区：[1]天津工业大学生命科学学院天津市光电检测技术与系统重点实验室,天津300387 [2]天津工业大学电气与电子工程学院天津市光电检测技术与系统重点实验室,天津300387

出　　处：《激光与光电子学进展》2022年第12期466-473,共8页Laser & Optoelectronics Progress

基　　金：天津市自然科学基金(18JCYBJC15300);天津市高等学校创新团队培养计划(TD13-5034)。

摘　　要：传统的货车点云配准算法通过寻找点云间的关键特征实现点云配准,这种方法效率较低,且点云之间存在重复的场景、噪声点,配准算法寻找的关键特征往往是不准确的。对此,提出了一种基于激光雷达的多视角点云配准方法。所提方法将惯性测量单元引入点云配准,在不依赖点云数据的情况下完成对应点云位姿矫正,然后采用随机采样一致性算法拟合局部平面搜索最近点,融入最近点迭代算法,快速寻找对应点集,实现点云精确配准。在采集的卡车数据集上进行实验,所提方法可在4 s内完成配准,平移误差最大不超过0.01 m,旋转误差控制在0.1°以内。实验结果表明,所提方法在货车点云配准中具有良好的配准效率及配准精度,具有较高的适用性。The conventional truck point cloud registration algorithm enables point cloud registration by determining the key features among point clouds. However, this method is inefficient because of the presence of repeated scenes and noise points among point clouds. Furthermore, the key features obtained using this algorithm are often inaccurate. Therefore, herein, a multiview point cloud registration method based on laser radar is proposed. In the proposed method, the inertial measurement unit is introduced into point cloud registration to complete the pose correction of the corresponding point cloud without relying on point cloud data. Then, the random sampling consensus algorithm is used to fit the local plane to determine the nearest point, which is integrated with the nearest point iterative algorithm to rapidly identify the corresponding point set and realize the accurate registration of the point cloud. The proposed method is verified via an experiment on a truck data set. The proposed method can complete the registration within 4 s, the maximum translation error is 0. 01 m, and the rotation error is within 0. 1°.Experimental results confirm that the proposed method exhibits good registration efficiency and accuracy in truck point cloud registration and shows high applicability.

关 键 词：遥感 测量 激光雷达 点云配准 自动装货 惯性测量单元 最近点迭代算法 

分 类 号：P237[天文地球—摄影测量与遥感]