摘要
传统迭代最近点(Iterative Closest Point,ICP)算法在进行点云配准时,若点云初始位置相差较大时,容易陷入局部最优,同时,该算法无法解决部分重叠的点云的配准问题。鉴于此,提出了一种改进的ICP激光点云精确配准方法。首先通过对两片点云的主成分分析并矫正主轴方向以完成初始配准,获得一个较好的初始位置。然后利用2次搜索最近距离来获取各点的概率值,并将其嵌入到最小二乘函数中来改进ICP算法,以达到对部分重叠的点云进行配准的目的。实验结果表明,在不同重叠度的数据下,提出的方法的配准误差分别为0.307 8 mm、0.287 2 mm;运行时间仅为4.4 s、4.2 s。该方法可以对初始位置相差较大且具有部分重叠的点云进行精确配准,同时提高运行效率并对噪声具有相应的鲁棒性。
The traditional iterative closest point(ICP)algorithm can easily fall into local optimum when the initial position of point cloud is quite different.At the same time,this algorithm cannot solve the registration problem of partially overlapped point cloud.In view of this,a laser point cloud precise registration method with improved ICP is proposed.Firstly,the principal components of two point clouds are analyzed and the principal axis direction is corrected to complete the initial registration to obtain a better initial position.Then use two search nearest distances to obtain the probability values of each point and embed them in the least squares function to improve the ICP algorithm to achieve the purpose of registering partially overlapping point clouds.The experimental results show that the registration errors of the proposed method are 0.3078 mm and 0.2872 mm under the data of different overlapping degrees and the running times are only 4.4 s and 4.2 s.This method can accurately register point clouds with large initial position differences and partial overlap,while improving the operating efficiency and making the noise with related robustness.
作者
李慧慧
刘超
陶远
LI Huihui;LIU Chao;TAO Yuan(Anhui University of Science and Technology,School of Geomatics,Huainan Anhui 232000,China)
出处
《激光杂志》
北大核心
2021年第1期84-87,共4页
Laser Journal
基金
国家自然科学基金(No.41704008)。
关键词
点云配准
ICP
主成分分析
重叠度
概率值
point cloud registration
ICP
principal component analysis
overlap degree
probability value
