Abstract: In some engineering practices, it is necessary to match a Spot Array [P*] which is most similar to an unknown Spot Array [P] from a set [S] that includes multiple similar known Spot Arrays and match pairs of samples between the two Spot Arrays. The key challenge to realize the task is how to match the known [P*] with the most similar to unknown [P.] This is a new novel challenge somewhat and it has been researching barely in theory. A described algorithm based on geometrical features analysis in graphics is explored in this paper. Among the algorithm, its first step is to construct each Spot Array into a Simple Graph as its outline graphic and constructs a chained construction for each at the same time, it is possible to get the pairs of samples matching finished between the two Spot Arrays by utilizing these chained constructions. In the end, the similarity between the unknown Spot Array and each known Spot Array can be calculated by using their geometrical features, then matching up one pair which most similar. It is not affected by the effects of both graphical rotation and size scaling. By several experiments show that this algorithm can fastly and precisely finish the task in comparing the graphical similarity and samples matching.

Key words: comparison in spot array, comparison in spot array similarity, comparison in graphical outline, matching spot array sample

摘要： 在工程实践中，需要从具有多个相似点阵的已知点阵（Spot Array）集[S]中匹配出一个与未知点阵[P]最相似的点阵[P*]，然后对两个点阵做样本点匹配。完成这个任务的关键挑战在于如何匹配出与未知点阵[P]最相似的已知点阵[P*]。这个问题比较新颖，目前少有理论研究，文中探索出一种基于图形的几何特征分析的描述算法，算法首先将每个点阵构建成一个唯一的简单图（Simple Graph）轮廓图形，同时为每个图形构建一个链式结构，然后利用轮廓图形的几何特征计算未知点阵与各已知点阵的相似程度，匹配出相似度最高的一个，最后利用链式结构完成两个点阵间的样本点匹配。该算法不受点阵的坐标系旋转和尺度缩放的影响。通过实验表明，该算法能够快速、准确地完成点阵相似度比较和样本点匹配任务。

关键词: 点阵比较, 点阵相似度比较, 图形轮廓比较, 点阵样本点匹配