大型工件位姿估计中的稀疏点云配准方法

针对飞机、轮船、汽车等大型设备制造领域中大尺寸工件的精确位姿检测的问题,提出了一种用于估计具有自由曲面特征的工件空间位姿的稀疏点云数据自动配准算法。为了提高点云配准的精度和速度,着重将配准过程分为粗略配准和精确配准两个阶段实现。鉴于稀疏点云数据中的各点曲率具有旋转平移不变性,设计基于曲率特征的点云粗略配准算法;为弥补现有ICP算法的不足,提出基于距离约束的四叉树逼近算法。仿真实验表明,完成点云配准后,稀疏点云中各个点位的平均误差为0.0168 mm,均方差为0.0095 mm,说明该算法能够自动、准确、快速地实现点云数据配准。

A registration algorithm of sparse point cloud in pose estimation for large work-piece

Currently, there is a technical trouble on pose detection of large work-piece in the field of large equipment manufacturing such as aircraft, ships, and automobile. In order to resolve this problem, aiming at the registration of large work-piece pose estimation, a registration algorithm of sparse point cloud for estimating pose of large work-piece is proposed in this paper. The registration algorithm includes two main parts： the coarse registration and the fine registration. Firstly, curvature which is proved to be invariant with translation and rotation can be used to perform the coarse registration. Secondly, a quadtree algorithm based on distance constraint is introduced so as to complete the fine registration of sparse point cloud. At last, experimental results show that the average error of position between the corresponding points is 0. 0168 mm and mean-square deviation of that is 0. 0095 mm. The proposed algorithm which is suitable to process sparse point cloud can be used to realize automatic, accurate and rapid registration.