改进ICP算法在点云配准中应用.pdfVIP

System Simulation Technology Application (Volume 14) Application of Improved ICP Algorithm in Point Cloud Registration Wanxin XU, Hongli XU College of Computer Information Technology, Beijing Jiaotong University, Beijing, China, 100044 Email: xwx611@126.com Abstract: In order to solve range image registration problem of 3D model reconstruction. In this paper, the basic principle of the ICP algorithm are introduced briefly. Improved technologies of each stage of ICP algo- rithm are introduced in detail. And, two range images with the different viewpoints are registered by choosing re-sampling method and divided the longest shaft K-D-tree method and curvature feature based on traditional ICP algorithm, realize the perspective of different images quickly registration process. The accuracy of the algorithm will not be affected to improve the matching speed. The algorithm is verified in the applications. Keywords: Iterative closest point; range image; registration; K-D-tree 改进的ICP 算法在点云配准中的应用 徐万鑫，许宏丽 北京交通大学计算机与信息技术学院，北京，中国，100044 Email: xwx611@126.com 摘 要：为了解决三维模型重建中的距离图像配准问题。在介绍ICP 算法的基本原理基础上，分析了 对于ICP 算法的各个阶段的改良方案，进一步地，在传统的ICP 算法的基础上，利用重采样方法进行 数据精简，结合分割最长轴的K-D-tree 方法和曲率特征，实现了不同视角下距离图像的快速配准过程。 改进后的算法在准确度基本不受影响的情况下，提高了配准速度，进行了仿真实验，验证了算法的有 效性和稳定性。 关键词：最近点迭代；距离图像；配准；K-D-tree 1 引言 许多学者对点云配准方法进行了大量的研究，目前 应用最为广泛的是由 Besl[1] 等人提出的迭代最近点 随着三维激光扫描技术的发展，三维点云数据在逆 （Iterative Closest Point，ICP ）算法。该方法寻找两片 向工程、可视化技术、虚拟现实技术、机器视觉等领域 点云之间的对应点对之间的关系，通过迭代方法计算两 具有十分广泛的应用。由于激光扫描设备在扫描时受到