中文摘要：

多边形的骨架线是对其形态的准确描述，多尺度的骨架线表达在图形分析如多边形的形态恢复和多边彤的模式识别等方面具有重要作用。利用双缓冲区变换、障碍距离变换和Voronoi图技术，提出一种新的多边形多层次骨架线提取方法，并利用复杂湖泊多边形、简单多边形、含岛洞多边形、噪声边缘多边形等不同形状的多边形测试和比较该方法提取的骨架线的连通性、拓扑一致性和对图形旋转的灵敏性等。试验结果表明，本文方法在上述特性上优于现有算法。

英文摘要：

Polygon skeleton is generally recognized as a good shape descriptor. The multiscale representation of polygon skeletons makes it convenient for image analyzing, such as polygon shape recovering, pattern recognition etc. A new method was presented to extract multiscale skeletons based on double buffering transforming, distance transformation with obstacles and Voronoi diagram. The method was tested by extracting various skeletons from polygons in different shapes and compared the skeletons qualities from the following indexes, connectivity, topological consistency, and graphics rotating sensitivity. These experimental polygons include complex lake polygon, simple polygon, polygons with holes and noisy polygons. Result shows that according to the above criteria, our method outperforms existing methods to a certain extent.