摘要
为了提高激光点云的配准精度和效率,解决两片点云之间存在尺度变换的配准问题,提出了一种基于有向包围盒的尺度点云配准算法。首先,分别生成两片点云的空间有向包围盒,利用两个包围盒对应边的比值计算尺度因子。然后,将目标点云包围盒进行尺度放缩,再利用包围盒对应顶点的关系计算旋转矩阵。同时,引入点云的单位向量和,以单位向量和之间余弦相似度最大为准则,选择正确的旋转矩阵。最后,为了实现精确配准,将尺度因子引入点到面迭代最近点(Iterative Closest Point, ICP)算法中,利用加权最小二乘法求解变换参数。实验结果表明,在点云之间存在数据缺失、噪声干扰和尺度变换的情况下,所提算法可以实现快速精确配准,且具备良好的稳健性。
To improve the accuracy and efficiency of laser point cloud registration and solve the problem of scale transformation between two-point clouds, a scale point cloud registration algorithm based on oriented bounding box is proposed. Firstly, generate the space oriented bounding boxes of two-point clouds respectively, using the ratio of the corresponding edges between two bounding boxes to calculate the scale factor. Then, scaling the target point cloud bounding box and using the relationship of corresponding vertices between two bounding boxes to calculate the rotation matrix. At the same time, the unit vector sum of the point cloud is introduced, and the correct rotation matrix is selected based on the principle of maximizing cosine similarity of the unit vector sum. Finally, in order to achieve precise registration, scale factor is introduced into the point to plane Iterative Closest Point(ICP) algorithm, solving transformation parameters by using the weighted least squares method. Experimental results show that the proposed algorithm can be quickly and accurately register in the case of missing data, noise interference and scale transformation in the point clouds, and has good robustness.
作者
王玉文
李珊君
杨赟秀
舒勤
WANG Yuwen;LI Shanjun;YANG Yunxiu;SHU Qin(College of Electrical Engineering,Sichuan University,Chengdu 610065,China;Southwest Institute of Technical Physics,Chengdu 610041,China)
出处
《激光杂志》
CAS
北大核心
2022年第1期12-18,共7页
Laser Journal
基金
国家重点研发计划(No.2017YFB0405101)。
关键词
机器视觉
激光点云配准
有向包围盒
尺度配准
最小二乘法
machine vision
laser point cloud registration
oriented bounding box
scale registration
least squares method
