一种基于曲率的点云自动配准算法

Automatic registration algorithm for scattered point clouds based on curvature feature

针对兵马俑破碎俑片虚拟复原过程中拼接效率低的问题,提出了一种新型的基于曲率的散乱点云数据自动配准算法。该算法利用MLS表面计算出两组点云中每个点的曲率,提取局部曲率变化最大的特征点,并计算曲率的Hausdorff距离来获得初始匹配点,然后根据初始匹配点之间极大极小曲率的相似度函数,采用粒子群优化算法确定精确匹配点。最后用四元组法求得坐标变换实现粗配准,并且用迭代最近点算法提高配准精度。实验验证了算法的有效性和稳定性。

Due to the bottlenecks such as low efficiency in the process of the virtual recovery of the Terra-Cotta Warriors broken fragments, this paper proposed a novel curvature-based point cloud automatic registration method. Firstly,it estimated curvature of each point by using MLS surface, then extracted the feature points based on the maximum local change of curvature. Secondly,it obtained the initial matching points by computing the Hansdorff distance of curvature, and acquired the accurate matching points by the PSO algorithm based on the maximum, minimum curvature similarity function. Finally, estimated the coordinate transform by the quaternion to achieve coarse registration, and improved the registration accuracy by the ICP algo- rithm. The experiments show the effectiveness and stability of this algorithm.