基于曲面边缘提取的航空叶片点云配准
Aircraft Blade Point Cloud Registration Based on Surface Edge Extraction
陈伟 1刘瑾 1杨海马 2潘方超 1赵红壮 1张锐1
作者信息
- 1. 上海工程技术大学电子电气工程学院,上海 201600
- 2. 上海理工大学光电信息与计算机工程学院,上海 200093
- 折叠
摘要
针对航空叶片点云配准过程中,特征提取困难导致的配准时间长、精度低等问题,提出一种基于Alpha Shape算法的点云边缘轮廓提取方法,将源点云使用PCA主成分分析法进行坐标系变换,以变换后的坐标系的Y-Z平面作为投影面,将曲面点云投影为平面点云,使用Alpha-Shape算法获得其平面边界,将平面边界坐标与原有曲面点云坐标进行比对,获得三维边缘点坐标,从而得到曲面边缘点云.在粗配准阶段,将目标点云进行下采样,对曲面边缘点云和目标点云用FPFH算法进行特征描述,将边缘点云和下采样后的目标点云使用随机一致性算法(SAC-IA)作点云配准,使用粗配准得到的参数矩阵将源点云变换到符合ICP算法要求的初始位置,最后使用基于列文伯格-马夸尔特(LM)优化的ICP算法完成精配准.结果表明,相比于SAC-IA+ICP算法,精度提高了54％,速度提升了17％.
Abstract
For the problems of long registration time and low accuracy caused by the difficulty of feature extraction during the registration of aviation blade point cloud,a method of point cloud edge contour extraction based on Alpha Shape algorithm is proposed.Using PCA principal component analysis to transform coordinate system of source point cloud,the Y-Z plane of the transformed coordinate system is used as the projection plane,the surface point cloud is projected into planar point cloud,and its planar boundary is obtained by Alpha-Shape algorithm,Compare the planar boundary coordinates with the original surface point cloud coordinates,and obtain the 3D edge point coordinates,then obtain the surface edge point cloud.In the coarse regis-tration stage,downsampling the target point cloud,the surface edge point cloud and the target point cloud are characterized by the FPFH algorithm,using the random consistency algorithm(SAC-IA)to register the edge point cloud and the down-sampled target point cloud,and the source point cloud is transformed to the initial position that meets the requirements of the ICP algo-rithm using the parameter matrix obtained from the coarse registration,Finally,the ICP algorithm based on Levenberg-Mar-quardt(LM)optimization is used to complete the precise registration.The results show that compared with SAC-IA+ICP al-gorithm,the accuracy is improved by 54％and the speed is increased by 17％.
关键词
边缘提取/点云配准/航空叶片/列文伯格-马夸尔特优化/主成分分析Key words
edge extraction/point cloud registration/aviation blade/Lewenberg Marquardt optimization/principal component analysis引用本文复制引用
出版年
2024
NETL
NSTL
万方数据