基于标志点的多视点云优化拼接方法

Method Multi-view Point-clouds Optimal Registration Based on Marked Points

针对基于标志点序列扫描点云拼接过程中出现的累积误差问题,提出一种利用光束法平差方法计算最优全局点位的拼接优化算法。该方法以扫描特征点在全局坐标系下的位置为初始值,以反投映残差最小化为目的,构建优化代价函数,通过迭代的方法计算特征点的全局最优点位,最后以该最优点位为基准,对序列点云进行空间位置调整。实际测量结果表明,该方法可以明显减少拼接过程中所产生的累积误差,具有提高拼接精度的作用。

In order to resolve the practical problem about measurement accumulating error in the multi-view point cloud registration based on marked points, a registration optimization algorithm of calculating the optimal overall point location by means of bundle adjustment is proposed to improve registration accuracy. The method takes the scanned feature point location in global coordinate system as the initial value and the back projection residual error minimum as the matching point to construct the optimization cost function, and calculate the globle optimal point location of feature points by iteration method. And then the optimal point cloud location is taken as reference to perform the space location adjustment of sequence point clouds. The experimental results show the proposed algorithm can evidently reduce the accumulated error caused by registration and improve tbe registration accuracy.