基于虹膜识别的身份认证系统算法研究

【作者】 刘洋；

【导师】 彭文达； 李霞；

【作者基本信息】 中国科学院研究生院（西安光学精密机械研究所） ， 光学工程， 2009， 博士

【摘要】 本文主要工作是虹膜识别系统的算法研究,虹膜是位于瞳孔与巩膜之间的环状区域,虹膜识别具有比指纹更高的安全性成成为生物识别领域近年来研究的重点。论文主要针对虹膜识别算法部分的虹膜定位,特征提取,虹膜分类,虹膜系统安全性等四个方面做深入的分析研究。1、在虹膜定位研究中,本文提出基于特定感兴趣区采样的虹膜内外圆定位算法,快速准确地实现了虹膜内外圆定位。本文同时提出方向可控,标准差可调的不对称Asymmetry Canny边缘提取算子,通过方向开关的控制,有选择地提取虹膜感兴趣边缘,屏蔽无规则边缘干扰,有效实现了眼睑定位,睫毛干扰屏蔽等虹膜图像预处理工作。2、在特征提取算法中,本文构造出边界清晰,过渡带近似理想的滤波通道,可以更加准确地提取所需频段特征信息,滤除噪声干扰,为虹膜识别提供有效的特征数据。另外,本文通过实验对虹膜特征谱所在的有效频段及方向性做了详细的分析,为特征提取时滤波器参数的设置提供了有效实验依据。3、针对虹膜分类算法,本文提出了一种基于虹膜完整相位信息,权重可调的相位相关分类算法。实验表明,本文提出的分类算法比Daugman的Hamming距离髯法增大了一倍空间,并且更完整地保留了虹膜的原始特征信息,更能体现虹膜特征之间的差异性,分类效果好、鲁棒性强。4、针对虹膜识别系统的安全性,本文提出了以节省网络资源为目标的虹膜特征码混沌加密系统,以及便于算法升级和设备维护的虹膜图像复合混沌加密系统。实验表明以上两种系统密钥空间大,抗攻击能力强,加密效果好。本文最后的结论与展望部分对全文工作进行了概括性总结,列举了虹膜识别系统在理论和应用上现已取得的各种成就,以及虹膜识别系统亟待进一步发展和完善的地方。本文所做的这些工作是提高虹膜识别系统性能的有益尝试,是对现有虹膜识别算法的丰富与补充,本文提到的算法思想都经过了实验验证,并且取得了较好的实验效果。更多还原

【Abstract】 This paper mainly discusses the algorithm of iris recognition system. Iris is a circle region between the pupil and the sclera. Because iris recognition has better security than the fingerprint, it has become a key point in the field of biometric in recent years. This paper deeply analyzes four aspects of the algorithm of iris recognition, that is, iris location, feature extraction, iris classification and security of the iris system.First , In the research of iris location ,this paper proposes "Asymmetry Canny Edge Extraction Operator" ,which has the functions of direction controlling and adjustable asymmetric of standard deviation . By the controlling of the direction switch, it can selectively extract the interested edge of iris, and avoid the interference of ruleless edge to finish the iris pre-processing work, including the iris location, eyelids location, eyelashes interference shield etl.Second, In feature extraction, the paper constructs a rectangular filter channel which has the characters of clear border and relatively ideal transition band. It could accurately extract the wanted frequency information, get rid of noise interference, and provide reliable data for iris recognition. In addition, the paper fully analyzes the effective frequency and direction for the spectrum of the iris by experiment, which can provide available experimental reference for the filter parameter design.Third, For the iris classification, the paper proposes an adjustable weight of phase correlation classification algorithm, which is based on the iris full phase information. The experiment shows that classification distance is proposed in this paper, it can extend double space than the Hamming distance of Daugman, remain even more original features of the iris information and reflect the difference between features in a more effective way. So the experiment is better, and the robust is strong. Fourth,For the security of iris recognition system, this paper proposes two kinds of chaos encryption system. One is ’Iris character code chaos encryption system’, and another is ’Composite chaos encryption system’. The purpose of the first system is to save the resources of network, and the second system is convenient for algorithm upgrading and equipment maintenance. The experiment shows that the algorithms analyzed above have large key space, better attack resistance and better encryption performance.In the last part, the paper concluded the achievements that had bee(?) made in iris recognition , which includes the theory and application, and also analyses the key points that the iris recognition system needed to further develop and improve.All of the work which had been done is not only the useful attempt ion to improve the performance of iris recognition system, but also the rich and reinforcement to the existing iris recognition algorithm. All of the algorithms which had been mentioned in the paper have been tested in experiment, and achieve a better result.更多还原