一种基于改进投影方法的点云拼接算法

Point-clouds registration algorithm based on improve projection

针对传统ICP(iterative closest point)算法在收敛速度或拼接精度上无法同时满足实时测量要求的问题,提出了一种适用于单、双目结构光三维测量系统的快速且高精度的点云拼接算法。该算法首先将参考视点云上的采样点反向投影到目标视点云的2D成像平面上,然后再将2D反向投影点正向投影到目标视点云上,展示了一种由目标视投影点到采样点法线上的投影直至收敛于线—面交点的迭代过程。实际测量结果表明,该算法在保证拼接精度的同时显著提高了收敛速度,是一种具有很高实用价值的拼接算法。

According to the traditional ICP algorithm in convergence speed or the registration accuracy failed to meet the demands of the real-time measurement,this paper described an accurate and efficient algorithm of point cloud registration that suitable for single and binocular structured light three-dimension measurement system.First the reference point cloud sampling point back projected to the destination point cloud of the 2D image plane,and then forward-projected the source point to the destination surface.The paper showed that iterative projections of the projected destination to the normal vector converge to the intersection point.Actual measurement result shows that the method to ensure the accuracy of registration while the convergence rate is significantly pricey,and it is a high practical of registration algorithm.