基于表面贴合的不同视点传感器点云拼合方法研究  

Research on Point Cloud Registration about Sensor from Different View Points Based on Surface Fit

作　　者：陈天凡[1,2] 高诚辉[1] 何炳蔚[1] 

机构地区：[1]福州大学,福州350108 [2]福州职业技术学院,福州350108

出　　处：《中国机械工程》2012年第24期2968-2971,共4页China Mechanical Engineering

基　　金：国家自然科学基金资助项目(51175087)

摘　　要：针对视觉传感器对物体表面点云数据采集的离散性,提出基于表面贴合的点云精确拼合方法。先对不同视点传感器获得的点云进行初始配准拼合,再以包容球方法选取点云重合表面区域的三处局部表面,用三维二次曲面方程拟合;旋转变换采用四元数法,经非线性最小二乘优化得到精确拼合的平移和旋转参数。该方法具有较快的收敛速度和较高的拼合精度,实例验证了所提方法的可行性。Focusing on discreteness of point cloud data acquisition from an object surface,a method of point cloud accurate registration was proposed based on surface fit. Firstly, initial point cloud regis- tration about sensor from different view points was completed. Then, selecting three local surface from the surface overlap areas of point cloud by contain sphere, the local surface fitting with three--dimen- sional quadratic surface equation,the rotation transformation using quaternion method,the translation and rotation parameters about accurate registration were obtained by means of nonlinear least square optimizing method. The algorithm has faster convergence and higher registration precision. The exper- imental results indicate the effectiveness of the proposed method.

关 键 词：视觉传感器 点云拼合 表面贴合 局部表面 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]