基于对偶四元数描述的LiDAR点云解析配准算法

LiDAR Point Cloud Registration Algorithm Based on Dual Quaternion

点云配准技术在空间三维建模中具有重要作用,传统的ICP算法存在初始位置要求高、迭代不收敛、计算量大等问题,解析法直接解算出变换参数,无需迭代,计算效率较高。在简要介绍对偶四元数的基础上,利用对偶四元数能够描述坐标系间的旋转与平移,推导出点云配准新算法。实验结果表明,该算法具有解析法相对于迭代法的优点,同时也解决了变换参数不同时求解的问题,可以避免误差累积问题的出现,理论上可以进一步提高配准精度,在以后的研究中具有一定的实用价值。

Based on the brief introduction of dual quaternion which can be used todescribe the rotation and translation between different coordinates systems,this paper proposes a new registration algorithm of LiDAR point cloud based on dual quaternion.Experimental results indicate that this algorithm has the advantages of analytical method and it can solve the problem that the transform parameters cannot make solutions at the same time,which avoids the accumulation of errors and improves the accuracy of registration in theory.It also has some practical value in future studies.