基于断裂面邻域特征的文物碎片拼接

Reassembly method of cultural relic fragments based on the neighborhood characteristics of fracture surface

为解决传统拼接算法在断裂部位受损情况下存在拼接误差大且耗时的问题,本文提出一种基于断裂面几何特征的破损文物碎片自动拼接算法。首先,定义碎片模型邻域特征参数,提取断裂面特征点,依据最小二乘法原理构造曲率特征参数对特征点集进行优化;然后,为解决稀疏点云特征难以匹配的问题,定义特征点间相对距离和相对夹角作为特征描述符,依据集合相似理论对特征点进行相似性度量,提取断裂面特征点对匹配集,并利用随机抽样一致性算法剔除误匹配点对,筛选出最优匹配集;最后,采用奇异值分解法计算旋转、平移矩阵,利用基于K-D树改进的迭代最近点算法实现碎片的精确拼接。实验结果表明:与传统的拼接算法相比,本文特征点少,特征描述符简单,鲁棒性强,有效提高了碎片拼接的准确性和效率。

To solve the problems concerning large errors and long times required in traditional stitching algorithms in the case of a damaged fracture location,this paper proposes an automatic splicing algorithm based on the geometric characteristics of the fracture surface.Here,the neighborhood feature parameters of the fragment model are defined,the feature points of the fracture surface are extracted,and the curvature feature parameters are constructed according to the principle of the least-squares method to optimize the feature points set.Subsequently,to solve the problem regarding the difficulty in matching the features of a sparse point cloud,the relative distance and relative angle between the feature points are defined as feature descriptors.According to the set similarity theory,the feature points of the fracture surface are measured via similarity measurement,and the matching set of the feature points of the fracture surface is extracted.Following this,a random sampling consistency algorithm is used to eliminate the mismatched points and select the optimal matching set.Finally,singular value decomposition(SVD)is used to calculate the rotation and translation matrix,and an improved iterative nearest point algorithm based on a K-D tree is used to achieve accurate splicing of the fragments.The experimental results showed that,when compared with the traditional reassembly algorithm,the algorithm proposed in this paper has fewer feature points,simpler feature descriptors,and higher robustness;it also more effectively improves the accuracy and efficiency of fragment reassembly.