基于布谷鸟优化的三维点云配准算法  

A 3D POINT CLOUD REGISTRATION ALGORITHM BASED ON CUCKOO OPTIMIZATION

作　　者：马卫[1,2] Ma Wei(State Key Laboratory for Novel Softuware Technology,Nanjing Uninersity,Nanjing 210093,Jiangsu,China;School of Hotel Management,Nanjing Institute of Tourism and Hospitality,Nanjing 211100,Jiangsu,China)

机构地区：[1]南京大学计算机软件新技术国家重点实验室,江苏南京210093 [2]南京旅游职业学院酒店管理学院,江苏南京211100

出　　处：《计算机应用与软件》2020年第12期216-223,272,共9页Computer Applications and Software

基　　金：国家自然科学基金项目(61321491,61272219,61100110);国家高技术研究发展计划项目(2007AA01Z334);江苏省科技计划(BE2011058,BY2012190);计算机软件新技术国家重点实验室创新基金重点项目(ZZKT2013A12);江苏省高校自然科学基金项目(17KJB520013,15KJB520017)。

摘　　要：针对ICP算法在点云配准时存在对初始位置敏感、易陷入局部最优的缺陷,提出一种基于布谷鸟优化算法的点云配准方法。对输入点云的均匀采样,并基于领域半径约束的固有形状特征点(Intrinsic Shape Signature,ISS)提取进一步简化点云。通过布谷鸟算法莱维飞行全局搜索更新策略完成对点云较好的初始配准,得到空间变换矩阵参数。利用k-D树(k-Dimension tree)近邻搜索方法加快对应点的搜索速度,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进行配准实验,结果表明该算法能有效克服ICP算法的缺陷,其全局搜索性能与寻优精度更具优势,抗噪性好,配准精度高,鲁棒性强。The Iterate Closed Point(ICP)algorithm is sensitive to the initial position when it is registered in the point cloud,and it is easy to fall into the local optimum.We propose a point cloud registration algorithm based on cuckoo optimization.It simplified the point cloud by uniformly sampling the input point cloud and extracting the Intrinsic Shape Signature(ISS)based on the domain radius constraint.Then we used the Cuckoo algorithm Lévy flights global search update strategy to complete the better initial registration of the point cloud and get the spatial transformation matrix parameters.The k-D tree(k-Dimension tree)neighbor search method was used to accelerate the corresponding point search so as to improve the efficiency of point cloud ICP fine registration.The registration experiment of the point cloud library model with different initial positions shows that this algorithm can effectively overcome the defects of the ICP algorithm,and its global search performance and optimization accuracy are more advantageous,with good noise immunity,high registration accuracy,and strong robustness.

关 键 词：布谷鸟算法 全局优化 点云配准 粗精配准 ICP算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]