一种面向船体扫描的多站点云鲁棒配准技术  

A Multi-site Cloud Robust Registration Technique for Hull Scanning

作　　者：胡敏捷 杨荣淇 谢洪[2] 马琦 HU Minjie;YANG Rongqi;XIE Hong;MA Qi

机构地区：[1]上海船舶研究设计院 [2]武汉大学

出　　处：《计量与测试技术》2024年第5期54-58,共5页Metrology & Measurement Technique

摘　　要：为实现船舱外壳的点云配准,本文提出了一种基于回环约束的多站点云配准方法。首先,使用ISS从场景中提取关键点,将颜色信息与形状信息的直方图相融合,得到船舱点云外壳的CSHOT特征描述符;然后,基于回环的粗配准,构建全连接图,并利用GROR,结合最大一致性集约束和闭环约束移除误匹配;最后,采用全局的精配准LUM算法完成位姿优化。实验表明:该方法的平均RMSE为0.0077m,对船舱外壳的点云配准取结果较好。In order to realize point cloud registration of cabin shell,a multi-site cloud registration method based on loop constraint is proposed in this paper.Firstly,ISS is used to extract key points from the scene,and the histogram of color information and shape information is fused to obtain the CSHOT feature descriptor of the cabin point cloud shell.Then,based on the rough registration of the loop,the fully connected graph is constructed,and GROR is used to remove the mismatching by combining the maximum consistent set constraint and the closed-loop constraint.Finally,the global LUM algorithm is used to complete the pose optimization.The experiments show that the average RMSE of this method is 0.0077m,and the result is good for the point cloud registration of the cabin shell.

关 键 词：船舱外壳 CSHOT 最大一致性集约束 闭环约束 多站点云配准 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]