摘要
在逆向工程中,为处理庞大的三维点云数据,重建物体的表面,提出一种基于多视角的改进ICP算法。通过采集多视角下的点云数据,利用Delaunay三角剖分以及深度值信息对相邻两组点云的重合部分进行提取,根据三角面片重心与待测物体重心之间的距离将获得的点云数据进行分类配准,计算最优的旋转矩阵与平移向量,提高ICP算法的效率。实验结果表明,该算法能够提高配准精度,缩短配准时间,具有良好的稳定性。
In the reverse-engineering, to reconstruct the surface of objects, a huge number of three-dimensional point clouds must be dealt with. An improved ICP algorithm on different viewpoints therefore was presented. To extract the overlapping por- tions of the point clouds, the point clouds on different viewpoints were acquired, the Delaunay triangulation and the depth value information were used. To get the best rotation matrix and the translation vector, a classification of the point clouds according to the distance between the barycenter of each triangular patch and that of the object was made, and then the efficiency of the ICP algorithm was improved. Experimental results show that the algorithm can improve the registration accuracy and shorten the registration time, which has good stability.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第11期3928-3932,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61379080)
关键词
双目相机
多视角
点云配准
最近点迭代
三角剖分
binocular camera
different viewpoints~ point cloud registration
ICP algorithm
Delaunay triangulation