摘要： 提出了一种基于曲率相似性的二维碎片轮廓匹配算法。该算法根据碎片轮廓上各点的曲率值来确定轮廓上的特征点，按照特征点对轮廓进行分段，将二维碎片的匹配简化为特征段的相似性度量与可匹配性验证。对可匹配的轮廓段进行欧氏变换以拼合两个轮廓曲线，自动地检测在当前变换下两个轮廓是否有重叠情况存在。没有重叠的两个轮廓组合起来更新为一个新的轮廓，进行再匹配直至重建过程完成。实例表明，算法是有效、稳定的。

关键词: 碎片匹配, 轮廓, 拼合, 曲率, 特征点

Abstract: This paper proposes an approach to the problem of object reconstruction from broken fragments of two-dimensional objects. Feature points on the contour are detected based on the curvature of every point. The feature point and its neighborhood points constitute a feature segment. Thus the contour matching task is reduced into similarity measure and feasibility verification of feature segments. The matchable contours are rotated and translated to fulfill fragments reassembly. Then the detection is carried out automatically, for checking whether there are overlaps between the two contours under this transformation. The contours without overlap are formed to a new contour, and the matching process is continued until the reconstruction is fulfilled. The experimental results demonstrate the algorithm is efficient and robust.

Key words: fragment matching, contour, reassembly, curvature, feature point

中图分类号: 

• TP311