【作者】 舒展；

【导师】 朱明富；

【作者基本信息】 华中科技大学 ， 系统工程， 2006， 硕士

【摘要】 指纹分类是自动指纹鉴别系统中的一个重要组成部分,一直以来都是国内外研究人员研究的热点和难点。在自动指纹识别系统的研究中,指纹分类提供了指纹数据库的一个重要索引机制,具有非常重要的意义。通过指纹分类,可以立即排除不同类别的指纹。一个准确而连贯的分类可以显著地减少大型指纹数据库的匹配时间,因此设计一个性能良好的指纹分类系统将大大提高整个指纹自动识别系统的使用效率,方便整个指纹识别的特征数据分类存储和数据库检索。模拟自然界生物进化过程的遗传算法,不受限制性条件约束,具有隐含并行性和全局解空间搜索等显著特性。遗传算法提供了一种求解复杂系统优化问题的通用框架,它不依赖于问题的具体领域,对问题的种类有很强的鲁棒性,所以广泛应用于很多学科。本文在简要介绍指纹分类和遗传算法的理论基础之上,针对遗传算法在优化过程中存在的问题,将一种改进的遗传算法与原本的遗传算法指纹分类相结合,设计出一种基于遗传算法的指纹分类算法。同时,通过构造二级指纹分类策略,显著地改善了指纹分类效率。主要内容如下:1.对国内外的指纹分类算法进行综述,并通过对各种指纹分类算法进行比较分析,提出所要研究的问题。2.在原有的基于遗传算法的指纹分类的基础上,应用一种加入BP算子的遗传算法对原本的分类算法进行改进。3.在深入分析指纹自动分类的一般原理和方法的基础上,探讨和实现了一种二级指纹自动分类算法。更多还原

【Abstract】 Fingerprint classification(FC) which has been the hot point and hard point of researchers in national or aboard is an important composing of Automatic Fingerprint Recognition System(AFRS). In the research of AFRS, FC provides an important reference mechanism making very important sense. With the fingerprint classification, it can exclude fingerprint of the different class. As a nicety and coherent classification can observably reduce the matching time of huge fingerprint database, designing a FC system with great capability will enormously improve the use efficiency of the AFRS and the store of character data and searches of database will be more convenient.Genetic Algorithm(GA) which simulates the processes of biological evolution possesses some outstanding characters, such as kicking over the traces of confine, implicit parallelism and global searching ability. GA which independent of the idiographic filed of problems, provide a common frame to solve the optimize of complex system. GA used in a lot of subject because of it has a strong quality of rush-stick.After introducing the FC and GA, a FC algorithm based on GA which is consist of a improved GA and the basic FC algorithm based on GA was designed. At the same time, a two-level FC strategy which observably improves the efficiency of FC is constructed. The main contributions of this thesis can be summarized as follows:1. After reviewing the FC algorithm in national and aboard and comparing the different FC algorithm, the topic of our research is bringing forward.2. An improvement of the born classification is designed by adding a joined BP operator GA.3. Based on the analysis of FC, a two-level FC is designed.更多还原