摘要
传统的迭代最近点(Iterative Closest Point,ICP)算法对点云配准产生初始位置敏感,易陷入局部最优,采用群智能优化算法可以有效解决这一问题,但同时会带来计算量较大、搜索效率不高的问题。为此,该文提出了一种二阶振荡的人工蜂群算法点云配准方法,通过对输入点云的均匀采样,并基于邻域半径约束的固有形状特征点(Intrinsic Shape Signature,ISS)提取简化点云,通过改进的二阶振荡人工蜂群算法完成对点云较好的初始配准,得到空间变换矩阵参数。最后通过近邻搜索法(k-Dimension tree,k-d tree)加速对应点查找,以提高点云ICP精细配准的效率。通过对不同初始位置的点云库模型和场景数据进行的配准实验表明,相比传统的配准方法和改进的群智能优化策略,该算法抗噪性好,配准精度高,鲁棒性强。
Traditional iterative closest point(ICP)algorithm is sensitive to the initial position generated by point cloud registration and is prone to local optimization.Swarm intelligence optimization algorithm can effectively solve this problem,but at the same time,it brings about large computational cost and low search efficiency.Therefore,we propose a second-order oscillation point cloud registration method of artificial bee colony.Through uniform sampling of input point clouds and extraction of simplified point clouds based on neighborhood radius constraint intrinsic shape signature(ISS),the improved second-order oscillating artificial bee colony algorithm is used to complete the initial registration of the point cloud,and the space transformation matrix parameters are obtained.Finally,the nearest neighbor search method(k-Dimension tree,k-d tree)is applied to accelerate the corresponding point search to increase ICP fine registration efficiency.The registration experiments on the point cloud database model and scene data at different initial locations show that compared with the traditional registration method and the improved swarm intelligent optimization strategy,the proposed algorithm has excellent anti-noise performance,high registration accuracy and strong robustness.
作者
马卫
李微微
MA Wei;LI Wei-wei(School of Hotel Management,Nanjing Institute of Tourism and Hospitality,Nanjing 211100,China)
出处
《计算机技术与发展》
2023年第6期79-87,共9页
Computer Technology and Development
基金
江苏省高校“青蓝工程”中青年学术带头人项目(QLDT2021)
江苏省社科应用研究精品工程课题(22SYB-117)
科研创新团队资助项目(2021KYTD04)。
关键词
点云配准
人工蜂群算法
二阶振荡
特征提取
配准优化
point clouds registration
artificial bee colony algorithm
second-order oscillation
feature extraction
registration optimization
作者简介
通讯作者:马卫(1983-),男,副教授,博士,研究方向为群智能优化、进化计算和计算机视觉。
