中国历史悠久，祖先留下了丰富宝贵的文化遗产。然而很多文物由于各种原因破裂成许多碎片。文物碎片模型的边界特征线提取可以广泛应用于文化遗产保护工作，如文物绘制，碎片拼接等，同时，点云模型的边界特征线提取在其他领域也有着广泛的应用。本文分析了文物碎片的特点，针对文物碎片表面粗糙，形状复杂，存在风化磨损等特点，提出两种文物碎片点云模型边界特征线的算法。具体的说，本文的主要贡献有：1) 对三维模型边界特征线提取当前主流的算法进行回顾，并对这些算法的优势和不足进行了分析比较。本文对文物碎片点云模型的特点和特征提取所可能遇到的问题进行了分析和总结。2) 提出了一种文物碎片点云模型的边界特征线提取算法。算法首先计算采样点的特征值，根据特征值对采样点聚类成若干个特征点簇；然后在这些特征点簇上构造特征线片断，并对这些线片断进行平滑去噪处理；最后把这些特征线片断通过算法连接起来，实现对点云模型的边界特征线提取。3) 提出了一种全自动的文物碎片点云模型的边界特征线提取算法。算法首先对碎片模型进行拉普拉斯平滑处理；然后把点云分割成若干个采样点集合，每个采样点集合表示模型上的一个面；最后从各个集合中提取边缘点，实现点云模型的边界特征线自动提取。4) 实现了本文中提出的这两个算法，使用不同类型的文物碎片点云模型对算法进行了测试。同时，本文对这两个算法的优势和需要改进的地方进行了分析。

China has more than 5,000 years of history, the ancestor left rich and valuable historical and cultural heritage. Due to the nature or human force， many of the artifacts were fractured into pieces of fragments. The feature lines extraction of fractured artifacts is a very important part in preservation of cultural relics, such as reassembling of fractured artifacts, artifacts rendering. At the same time, the extraction of feature lines of point cloud is very widely used in other fields. Two algorithms of feature lines extraction from point cloud are presented, especially focus on the eroded, rough fragments of fractured artifacts. The mainly contributions of this paper are presented as following:1) Review the state of art research and algorithms on feature lines extraction from 3D model problem. The challenges of feature lines extraction on fractured artifacts are discussed.2) Present an algorithm of feature lines extraction from point cloud of fractured artifacts. Firstly, the feature weight of each sample point of point cloud is computed via eigenvalue decomposition. Secondly, the sample points are clustered based on their weight and several local weighted minimum spanning trees are constructed following some smoothing processes. Thirdly, these separated trees are connected to generated completed and smoothed feature lines of the fractured artifacts. 3) Present an automatic algorithm of feature lines extraction from point cloud of fractured artifacts. Firstly, the fragments are smoothed via Laplace operator to reduce the noise. Then, several feature point groups are clustered via a modified cluster algorithm. Finally, the completed, smoothed feature lines are extracted from these feature groups. This algorithm is more reliable than previous one.4) Some kinds of data are used to test the proposed algorithms. The results demonstrate that the proposed algorithms are of robustness and efficiency to feature lines extraction from point cloud of eroded, rough fragment of fractured artifacts.