基于虹膜身份识别算法的研究

发布时间：2018-02-12 16:17

本文关键词： 虹膜识别虹膜定位特征提取特征编码匹配　出处：《东北石油大学》2014年硕士论文　论文类型：学位论文

【摘要】：在当代社会科技快速发展，交通通信极其便利的大环境下，身份鉴别是我们日常生活中经常遇到的问题，如何快速高效准确的实现身份鉴别是一个火热且极具潜力的研究课题。在众多的生物识别算法中虹膜识别是一种新兴的生物识别技术，基于虹膜来识别身份的技术具有如下几个方面的优势：唯一性、稳定性、非侵犯性以及携带方便等，它的这些优势使得基于虹膜识别在身份识别领域中占据着重要的位置，大量的科研学者投入无数的时间与精力到虹膜身份识别的研究中，使得虹膜识别身份的研究在这几年得到了快速的发展。本文研究的基于虹膜识别身份算法主要在虹膜内外边界点确定、虹膜图像归一化处理、虹膜图像的特征提取与编码以及特征编码匹配等几个方面进行了学习研究，最后通过判断特征编码匹配度与阈值大小关系进而实现身份的认证。其中在虹膜的定位过程中，本文采用了内外边缘分离的虹膜圆定位算法，内边缘定位运用最小二乘法圆拟合算法，考虑到内边缘定位精度不高的问题，对虹膜内边缘运用圆周上的差分运算来实现虹膜内边缘的精确定位，而虹膜的外边缘定位则运用经典的霍夫变换与边缘检测相结合算法。在归一化阶段本文采用了直角坐标系到极坐标的变换算法，将环形结构虹膜图像转换为极坐标下的矩形图像，方便下阶段的虹膜特征提取与识别。虹膜特征提取算法现阶段应用最多的是基于二维Gabor滤波器特征提取，但是由于二维Gabor滤波器宽受到限制，而一维Log Gabor滤波器能很好的克服二维Gabor的缺点与不足，，但是一维LogGabor在处理虹膜图片这样二维图片时会丢失全局的信息，针对此种情况本文提出应用二维的Log Gabor滤波器进行虹膜特征提取，构造四尺度四方向多通道Log Gabor滤波器组，利用该滤波器组对虹膜的纹理进行特征提取并且在滤波结果中选取能够较好的表现虹膜纹理的图片进行相位编码。在匹配阶段本文提出了最佳移位差方和距离标准偏差匹配方法，并将本文提出的算法与传统的汉明距离匹配算法相比。然后应用MATLAB的GUI模块简单设计了虹膜身份识别界面，利用本文提出的算法对CASIA数据库内的虹膜图像进行类内与类间试验，实验结果证明本文提出的最佳移位差方和距离标准偏差匹配算法具有更低的错误拒绝率与错误接受率，也就是本文提出的算法拥有更好的虹膜识别率，能够取得更好的识别效果。
[Abstract]:In the great environment of the rapid development of science and technology in contemporary society and the convenience of transportation and communication, identity identification is a problem we often encounter in our daily lives. How to realize identification quickly, efficiently and accurately is a hot and potential research topic. Iris recognition is a new biometric technology in many biometric recognition algorithms. The iris recognition technology has the following advantages: uniqueness, stability, non-invasiveness and portability. These advantages make iris recognition play an important role in the field of identity recognition. A large number of researchers have invested innumerable time and energy into the research of iris identification, which has led to the rapid development of iris identification in recent years. In this paper, the iris identification algorithm based on the iris inner and outer boundary point determination, iris image normalization, iris image feature extraction and coding, and feature coding matching are studied. Finally, identity authentication is realized by judging the relationship between the matching degree of feature coding and the threshold value. In the process of iris localization, the iris circle localization algorithm with the separation of inner and outer edges is used in this paper. In the inner edge location, the least square circle fitting algorithm is used to locate the iris inner edge accurately, considering the low accuracy of the inner edge location, the difference operation on the inner iris edge is used to realize the accurate location of the iris inner edge. However, the external edge of iris is located using the classical Hough transform and edge detection algorithm. In the normalization stage, this paper adopts the transformation algorithm from rectangular coordinate system to polar coordinate system. The iris image of annular structure is converted into rectangular image in polar coordinates, which facilitates iris feature extraction and recognition in the next stage. At present, the iris feature extraction algorithm is mostly used based on two-dimensional Gabor filter. However, because the width of two-dimensional Gabor filter is limited, the one-dimensional LogGabor filter can overcome the shortcomings and shortcomings of two-dimensional Gabor well. However, one-dimensional LogGabor will lose global information when dealing with two-dimensional images such as iris images. In this paper, a two-dimensional Log Gabor filter is proposed for iris feature extraction, and a four-scale four-direction multi-channel Log Gabor filter bank is constructed. The filter bank is used to extract the feature of iris texture and to select the image which can represent the iris texture in the filtering result for phase coding. In the matching stage, we propose the best shift difference square and distance scale. Quasi-deviation matching method, The proposed algorithm is compared with the traditional hamming distance matching algorithm. Then the iris identification interface is designed by using the GUI module of MATLAB, and the Iris image in CASIA database is tested by the proposed algorithm. The experimental results show that the proposed algorithm has lower error rejection rate and error acceptance rate, that is, the proposed algorithm has better iris recognition rate. Can obtain better recognition effect.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.41

【参考文献】