面向车身虚拟匹配的非均匀点云数据配准算法

Study on Inhomogeneous Point Cloud Registration Method for Digital Matching of Auto Bodies

车身点云扫描的质量检测方式为薄板件的虚拟匹配提供了数据基础。针对薄板件虚拟匹配过程的点云数据配准问题,提出了一种考虑车身复杂曲面特征的点云主平面拟合与精确配准流程与方法。首先在曲率化离散方法基础上,提取薄板件数模的非均匀化点云,采用基于主成分分析法的数模与实测点云主平面的构建方法,结合其主向量的旋转重合,通过迭代搜索获得均方根误差最小的最佳初始配准位置。进一步,利用改进的ICP算法对点云进行精确配准。采用车门内板单件的实测案例对提出方法进行验证,实验结果表明本方法较传统ICP算法将配准精度提升50%以上,为车身试制过程的薄板件精度评价与虚拟匹配分析提供了有效的方法基础。

Point cloud scanning inspection of the auto body provides the necessary data for digitalmatching of sheet metal parts.Aiming at the registration problem of point cloud data of sheet metal parts,this paper proposes a method of point cloud main plane extraction and accurate registration approach considering the complex form of the auto body.Firstly,based on the curvature of the discrete method,the heterogeneous point cloud is extracted.Then,the principal component analysis is used for principal plane construction.Combining with its main vector rotation,the MSE error obtained by iterative search is minimized for the best location for initial registration.Afterwards,an improved ICP algorithm is proposed for the second step of accurate point cloud registration.In this paper,the proposed method is verified with the measured data of the inside door panel.Results showed that the registration accuracy was improved by more than 50%compared with the traditional ICP algorithm,which provides an effective method for the quality evaluation and digital matching analysis of the sheet metal parts in the launch process of the auto body.