改进的多视点云精确配准技术  

An Improved Multi-view Point Clouds Precise Registration Method

作　　者：唐辉[1] 吕乃光[1] 韩建栋[2] 

机构地区：[1]北京信息科技大学光电信息与通信工程学院,北京100192 [2]山西大学计算机与信息技术学院,山西太原030006

出　　处：《微计算机信息》2010年第27期165-167,共3页Control & Automation

基　　金：基金申请人:吕乃光;项目名称:现场大尺寸数字摄影测量的精度理论研究;颁发部门:国家自然科学基金项目(50675015)

摘　　要：结构光多视测量下的三维点云配准是逆向工程中的关键和难点。提出了一种基于曲率的改进ICP算法,在采用改进的主方向贴合法实现预配准的基础上,通过求取被测物的曲面特征点,使用曲率相似的点作为初始匹配点对,然后采用k-dtree来加速搜索最近点,应用最近点迭代(ICP)算法,实现精确配准。实验分析表明,该算法具有计算速度快,配准精度高的特点,具有良好的配准效果。The registration of 3-D point clouds under the structure lights of multi-views measuring is the key and difficult problem in reverse engineering. A registration method of an improved ICP （Iterative Closest Point） algorithm based curvature feature points is put forward. Affixed to the basis of an improved main direction in the use of pre-registration to achieve the legitimate, through the strike of the surface characteristics of required materials and the use of the point at the same curvature as the initial matched points, we can accelerate the process of searching for the nearest point by making use of the k-d tree, and the application of ICP algorithm brings in the achievement of accurate registration. Experimental analysis shows that the algorithm has the characteristics of speedy calculation as well as high precision and well matched good registration results.

关 键 词：点云配准 最近点迭代算法 主元分析法 曲率 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术]