利用轮廓特征的多视点几何数据配准

Registration of Multiple Range Data from Real Scenes Using 3D Contour Features

以三维激光扫描仪为获取数据平台,针对扫描过程中存在的数据配准问题,提出了利用轮廓特征的大规模三维场景多视点几何数据自动配准算法。首先采用自适应曲线拟合技术提取三维轮廓特征,在此基础上建立八叉树数据检索结构。引入马氏距离,以八叉树叶结点为匹配单元计算初始转换矩阵,并经ICRP算法逐步优化直至获得两视点的最佳配准。最后,采用就近原则,给出大规模三维场景多视点全局配准策略。并通过对室内、室外和古建等场景的三维配准实验,证明算法的鲁棒性。

Laser scanner captures range data of real scenes. There exists some problems of data processing; most important of them is how to align all range data into the same coordinate system. An algorithm of registration of multiple range data from real scenes was proposed using 3D contour features. Firstly, using self-adaptive curve fitting, 3D contour features were extracted, and 3 D contour features were split into a searching structure of octree. Secondly, using mahalanobis distance, the leaf nodes were matched between two scans to compute original transform matrix, and the transform matrix was refined step by step through ICRP, until the best transform matrix was obtained. Lastly, the nearby principle was employed, a new global registration strategy was given. The experiments of multiple range data registration from indoor scenes, outdoor scenes and ancient buildings were done, and the results show the proposed algorithm is robust.