【作者】 常军民；

【导师】 陈国定；

【作者基本信息】 浙江工业大学 ， 控制理论与控制工程， 2005， 硕士

【摘要】 印鉴鉴别是计算机视觉和模式识别领域中一个比较活跃的研究课题。印章作为个人、企事业单位和政府的一种具有法律意义的标志和证据,在中国、日本和韩国等一些亚洲国家有着广泛的应用。目前,蓄意伪造印章的现象时有发生,有些伪造印鉴足以达到以假乱真的程度;同时,手工核对印章的低效率已成为信息管理系统的瓶颈。因此,开发印鉴自动鉴别系统有着广泛的应用前景。 印鉴鉴别可以分为印鉴预处理、配准、特征提取以及分类鉴别四部分,而配准和特征提取是整个印鉴鉴别的关键。国内外学者对此进行了深入的研究,从特征提取方法的角度来看,可以将其分为两大类:一类是基于配准技术而提取的印鉴几何特征,另一类是直接提取的印鉴代数特征。这两类方法在一定程度上都可以对印鉴进行鉴别,但由于真伪印鉴本身的差别很小,而且由于外在因素如盖印时印泥量、用力方向及大小等不同而导致的同一枚印章所盖印鉴之间的差异甚至超过真伪印鉴之间的差别。因此有效特征比较难于提取,从而导致识别率比较低。 对于印鉴鉴别这一较难的模式识别问题,传统的鉴别方法主要是针对印鉴的某一个方面提取印鉴的特征,使用单一的分类器进行判断,导致印鉴鉴别率不高。近年来,多特征多分类器融合方法成为模式识别研究的热点,该方法的优点是避免单一特征和单一分类器的局限性,多特征分类器的融合策略着重于“多特征和多分类器联合”在识别分类中的作用,降低“单一特征”和“单一分类器”对最终识别结果的更多还原

【Abstract】 Seal imprint verification is a very active and synthetic topic in the field of computer vision and pattern recognition. It is a legal sign and proof of person and government. It has extensive application in the Asian countries for example China, Japan and South Korea etc. At present making forgery seal imprint is often occurring, some forgery imprints have little difference with genuine imprints. What is more, low efficiency of handwork check imprints is becoming a bottleneck of information management, so exploitation seal identification system will have a very extensive application outlook.Major works of seal imprint verification include seal imprint preprocessing, registration, features extraction and classification etc. Particular, registration and features extraction are two vital parts of verification system. Scholars have deeply studied it home and abroad. From the angle of methods of features extraction, seal imprints verification are classified into two types, one is based on seal imprints geometry features and the other is based on seal imprints algebraic features and these methods are used to verified seal imprints in some degree. Due to little difference between genuine and forgery seal imprints and even between genuine and genuine seal imprints, effective features are very difficultly更多还原