双Kinect联合扫描点云配准方法  

A Registration Method for Point Clouds Obtained by Two Kinects

作　　者：鲁栋栋 吴以凡[1] 张桦[1] 

机构地区：[1]杭州电子科技大学计算机应用技术研究所,浙江杭州310018

出　　处：《杭州电子科技大学学报（自然科学版）》2014年第6期61-65,共5页Journal of Hangzhou Dianzi University：Natural Sciences

基　　金：国家自然科学基金资助项目(61202093);浙江省3D产业关键技术重点科技创新团队资助项目(2011R50009)

摘　　要：大型场景或者大尺寸物体的三维重建无法由单台设备完成,而由多台设备采集得到的点云位置偏移较大,一般的点云配准方法无法达到三维重建所需的精度。该文提出了一种基于Kinect标定的点云配准方法。首先通过棋盘格标定算法对两台Kinect进行标定,然后利用标定得到的相机内外参数计算不同视角下点云数据间的运动参数,接着根据得到的运动参数,对原始采集的点云数据进行初始配准。在此基础上,采用改进的迭代最近点算法,对初始配准后的点云数据进行精确配准。实验结果表明,与直接采用改进的ICP算法相比,该方法配准结果无明显的错位现象,且计算速度提高了15%左右。3 D reconstruction of the large-scale scene or the large size of the object can＇t be done by a single device, while the positions of the point clouds obtained by the multiple devices have a large offset, and existing methods can＇t reach the precision of 3D reconstruction. This paper describes a registration method for point clouds obtained by calibrated Kinects. The method uses chessboard calibration algorithm to calibrate two Kinects for purpose of obtaining the Kinects＇s intrinsic and extrinsic parameters, which are used to calculate the motion parameters between the multi-perspective point clouds obtained by calibrated Kinects. According to the motion parameters, the initial registration is accomplished, so that one can achieve the precise registration with using the improved iterative closest point（ICP） algorithm. The experimental results show that this method lowers the dislocation phenomenon and improve the calculation speed by 15%.

关 键 词：Kinect标定 棋盘格标定 点云配准 迭代最近点算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]