【作者】 李玉珍；

【导师】 王宜怀；

【作者基本信息】 苏州大学 ， 软件工程， 2005， 硕士

【摘要】 主曲线(principal curves)是高维空间中数据集合的一维光滑曲线。主曲线的理论与应用研究已经得到了众多计算机科学工作者的关注。寻找主曲线的有效算法和进行主曲线的应用研究，已经成为模式识别领域值得关注的热点之一。 本文的工作主要分为两个部分，第一部分提出了一种自适应的简明主曲线算法，第二部分给出了主曲线在汉字骨架提取中的应用。在论文的第一部分，首先对给定的数据集设计了一种提取数据集主成分特征的算法，随后，以此为基础，提出了一种自适应的主曲线算法，给出了算法的设计思想与过程描述。主曲线算法由两个层次构成：第一层次是基于向量量化器(即GL_算法)，其计算生成了离散数据集的多边形主曲线；第二层次将自组织拓扑映射与向量量化器相结合，最终生成主曲线。本文的主曲线算法继承了HS型主曲线算法和K型主曲线算法的主要优点，降低了一般主曲线算法的复杂度，使其变得更简洁明了。论文的第二部分是在总结已有汉字骨架提取的各种算法基础上，加以综合改进，同时应用第一部分的主曲线理论与算法，给出了一套完整的提取汉字骨架并将汉字骨架显示出来的算法，包括手写体汉字和各种印刷体汉字的骨架。该部分的算法可以分成三个层次：其一是对汉字进行预处理：包括细化给定的汉字、提取细化后的汉字特征和汉字笔段、完成生成汉字笔划等算法；其二是依据所提取的汉字特征，对畸形笔划进行优化处理。其三是将光学汉字骨架进行数学转换至欧氏空间上，并依据汉字笔划自身规范性，将本文第一部分的主曲线算法应用于欧氏空间的汉字骨架上，对汉字骨架进行优化处理，同时将主曲线优化模拟后的汉字骨架通过计算机屏幕输出再现出来。 主曲线理论、算法及其应用正在成为我国计算机科学界关注的研究方向之一。本文的工作将作为该项研究的起点。尤其是主曲线算法，在当前国内的计算机研究领域具有一定的先导性价值。将主曲线理论应用汉字骨架优化具有一定的实际价值。更多还原

【Abstract】 Principal curves are one-dimensional curves of data set in multidimensinal space. The theory and applicational study have been concerned with many computer researchers. Finding an effective algorithm of principal curves and studying its application have been the one point of animated thesis in pattern identify.Two parts compose this thesis. The first part expounds a self-adapted and concise principal curves algorithms. The second part is its application to Chinese character skeletonization. In the first part of my thesis, it above all designs an algorithm, called PCA, or picking up principal characteristic of data set. Based On it, a simpler principal curves algorithm is designed. The algorithm can be demarcated two segments. The first segment bases on the theory of generalized vector quantizer called refined GL_algorithm. It computes polygonal line principal curves for discrete data set. The other segment combines the method of self-organized topological mapping with refined GL_algorithm to design the principal curves algorithm. This principal curves algorithm inherits main virtues of the principal curves of HS-type and K-type, and it is simpler than HS-type and K-type. In the second part of my thesis, it is an application of the principal curves to Chinese characters skeletonization, including the characters skeletonization of handed type and any printing type. This part can be demarcated three layers. The first layer is called pretreatment. It is combined of many algorithms, including thinning Chinese characters, picking up the characteristic structure and segment of Chinese characters, creating stroke of Chinese characters . The second layer mostly includes the algorithms of optimizing stroke of Chinese characters. The third layer is to swith skeletonization of Chinese characters, to improve the principal curves algorithm, to optimize the characters skeletonization with principal curves, and to output the characters skeletonization.更多还原