摘要：

碎片拼接是古陶瓷修复的关键工作, 针对古陶瓷碎片形状随机、数量大、表面纹理弱且存在局部缺损而导致算法的精度较低、匹配时间较长等问题, 提出一种基于凹凸性和转向角的古陶瓷碎片二次配算法。在提取古陶瓷碎片轮廓曲线的基础上, 通过先后使用粗匹配和细匹配的二次匹配组合实现碎片的两两精确匹配。一次粗匹配先通过多边形逼近碎片轮廓曲线, 以降低轮廓的复杂性, 再提取多边形的顶点凹凸性和顶点转向角构建一次轮廓特征集合, 最后利用凹凸互补性和遍历顶点对齐的双模态特征初次匹配算法来寻找大致匹配段, 并得到粗匹配点集。二次细匹配先随机选取粗匹配点集中的任意相邻两点点对来提取碎片轮廓片段, 以减少轮廓点数量并提高算法效率, 再计算轮廓片段的轮廓转向角以提取二次轮廓特征集合, 最后利用基于粒子群优化的二次匹配来搜索精确匹配段, 并得到细匹配点集。实验结果表明, 该算法对二维古陶瓷碎片的拼接效果较好, 且具有较强的鲁棒性, 拼接误差不超过2%, 运行时间效率相比已有算法提高了8%~20%。

关键词: 碎片拼接, 二次匹配算法, 轮廓提取, 凹凸性, 转向角, 粒子群优化

Abstract:

Fragment splicing is the key work in ancient ceramic repair. To address the problems of the random shapes, large number, weak surface textures, and local defects of ancient ceramic fragments (which result in a low accuracy and long matching time of the algorithm), this paper proposes a quadratic matching algorithm of ancient ceramic fragments based on concavity-convexity and steering angle. On the basis of extracting the contour curves of ancient ceramic fragments, the algorithm realizes the two-pair matching of fragments by combining coarse matching and fine matching. First coarse matching: The algorithm approximates the fragment profile curve through the polygon first to reduce the complexity of the profile. Then, it extracts the vertex concavity-convexity and vertex steering angle of the polygon to construct the first contour feature set. Finally, it uses the initial matching algorithm of the dual-modal features of concave-convex complementarity and traversal vertex alignment to determine the approximate matching segment, and the coarse matching point set is obtained. Quadratic fine matching: First, the algorithm selects any two adjacent points in the coarse matching points selected randomly to extract the fragment contour fragment. This reduces the number of contour points and improves the efficiency of the algorithm. Then, it calculates the contour steering angle of the contour fragment to extract the secondary profile feature set. Finally, it uses the quadratic matching based on Particle Swarm Optimization (PSO) to search for the exact matching segment and obtain the fine matching point set. The experimental results reveal that the algorithm displays a good stitching effect and strong robustness for the splicing of two-dimensional ancient ceramic fragments. The splicing error is less than 2%, and the running time efficiency is high(8%-20% higher than that of the existing algorithm).

Key words: fragment splicing, quadratic matching algorithm, contour extraction, concavity-convexity, steering angle, Particle Swarm Optimization(PSO)