基于纹理分析和小波变换的虹膜识别算法研究.pdf
《基于纹理分析和小波变换的虹膜识别算法研究.pdf》由会员分享,可在线阅读,更多相关《基于纹理分析和小波变换的虹膜识别算法研究.pdf(58页珍藏版)》请在得力文库 - 分享文档赚钱的网站上搜索。
1、 论文题目:基于纹理分析和小波变换的虹膜识别算法研究 专 业:信号与信息处理 硕 士 生:宋 洁 (签名) 指导教师:吴冬梅 (签名) 摘 要 随着信息技术的高速发展,人们对身份识别的准确性、实用性和安全性等方面的要求越来越高。传统的身份识别方法已经不能满足现代社会的要求。近些年,基于生物识别技术依靠自身独特的优越性得到了迅速的发展和应用。在现有各类生物识别技术中,虹膜具有非侵犯性、唯一性、稳定性等优点,利用它进行的身份识别具有更高的准确率,目前虹膜识别技术被认为是最具有广阔应用前景的生物识别技术之一。 一般的虹膜识别系统主要包括了虹膜图像采集、虹膜图像预处理、虹膜特征提取以及虹膜特征匹配四个
2、部分。其中虹膜定位和特征提取是最为关键的部分。 根据虹膜固有的环状特征和灰度变化特征,本文将虹膜定位分成内边缘定位和外边缘定位两个部分。首先定位内边缘,采用了阈值分割和曲线拟合。先根据灰度直方图分割出瞳孔区域,利用二值化初步地将瞳孔分割出来,再在二值化图像中寻找最大的连通块消除睫毛及部分较暗区域的干扰,最后再在二值化图像中找出瞳孔的边缘点,采用最小二乘法估计瞳孔边缘的半径和圆心参数,就得到了虹膜内边缘的准确定位。然后定位外边缘,采用了微积分算子,根据虹膜内边界参数缩小外圆圆心的范围,可以避免盲目搜索,提高定位速度。采用本算法的定位准确率达到了 98.28%。 在虹膜特征提取过程中,针对当前存在
3、的特征提取方法的不足之处,在以下三个方面进行了改进: (1)对保留纹理区域的改进; (2)沿两个方向提取纹理特征信号; (3)用一阶导数的模极大值比用二阶导数的过零点检测函数 ()f x 突变点的效果更好。 本文先对纹理特征的选取进行改进,然后沿两个方向提取特征,再采用高斯函数的一阶导数作为小波对特征信号作小波变换后,利用模极大值的位置判断奇异点位置,再进行特征编码,最后比较特征编码的相似性,从而实现了匹配识别。采用本算法的识别率为 96.89%。 本文算法都在 VS2008 平台上使用中科院自动化所提供的 CASIA1.0 虹膜数据库进行了测试,结果表明,本文的算法不仅能有效地进行虹膜识别,
4、并具有很好的识别效果。 关 键 词: 虹膜识别;虹膜定位;小波变换;特征提取 研究类型: 理论研究 Subject : Study on Algorithm of Iris Recognition Based on Texture Analysis and Wavelet Transform Specialty : Signal and Information Processing Name : Song Jie (Signature) Instructor: Wu Dongmei (Signature) ABSTRACT With the rapid development of infor
5、mation technology, people have increasingly high requirements in identification accuracy, practicality and safety. The traditional identification methods cant satisfy the requirements of the modern world. In recent years, based on biological recognition technology on its own unique advantages obtain
6、ed a rapid development and application. In the existing each kind of biometrics, iris is non-invasive, uniqueness and stability etc, and use it to identify themselves with higher accuracy, at present the iris recognition technology is considered the most has the broad application prospect of biometr
7、ics one. General iris identification systems mainly include iris image acquisition, iris image preprocessing, iris feature extraction and iris feature matching four parts. Among them prevailing localization and feature extraction is the most important part. According to the ring feature and gray var
8、iation characteristics of iris inherent, this paper will divide iris position into two parts: inside edge localization and outer edge localization. First, the inside edge of the positioning threshold segmentation and curve fitting. According to gray histogram segment pupil area, using binary prelimi
9、nary will pupil segmentation out, then in pupil binary image in search for the biggest connecting block eliminate eyelash and part of the darker areas in the interference, and finally in binary image finds the pupil edge points, the least-square method estimates the radius and circle parameters of t
10、he pupil edge, get the accurate positioning of iris inside edge. Then, positioning the outside edge, adopted calculus operator, according to iris inner boundary parameter narrow within the scope of the centre of a circle, can avoid blind search, improve localization speed. Using this algorithm posit
11、ioning accuracy reached 98.28%. In the iris feature extraction process, in view of the current existing feature extraction method of the shortcomings, in the following three aspects are improved: (1) the improvement to retain texture area; (2) postpone two direction extract texture feature signal; (
12、3) use a derivative modulus maxima has better result than second derivative with zero detect mutating of function ()f x . In this paper, first, the selection of the texture characteristics was improved, and then postpone two direction characteristics extracted, and then using a derivative of Gaussia
13、n function as wavelet to characteristic signal make wavelet transformation, using modulus maxima location estimation singularity position, and feature encoding, and finally the similarity of comparative feature codes, thus fulfilling the matching recognition. Using the algorithm for the recognition
14、rate reached 96.89%. This algorithm in VS2008 platform used CASIA1.0 provided by the Chinese academy of sciences automation iris database was tested, and results show that this algorithm can not only effectively iris identification, but also has good recognition effect. Key words: Iris recognition I
15、ris location Wavelet transform Feature extraction Thesis : Theoretical study 目 录 I 目 录 1 绪论 .1 1.1 虹膜识别技术的研究背景和意义 .1 1.2 虹膜识别系统简介 .5 1.3 虹膜识别技术的国内外研究和应用现状 .6 1.3.1 虹膜识别技术的国内外研究历史和现状 .6 1.3.2 虹膜识别技术的国内外应用 .9 1.4 论文组织结构 .10 2 虹膜定位 . 11 2.1 三种经典的虹膜定位算法 . 11 2.1.1 Daugman 的虹膜定位算法 . 11 2.1.2 Wildes 的虹膜定位
16、算法 .12 2.1.3 王蕴红、谭铁牛等的虹膜定位算法 .13 2.1.4 三种定位算法的比较 .14 2.2 改进的虹膜定位算法 .15 2.2.1 虹膜内边缘定位 .15 2.2.2 虹膜外边缘定位 .17 2.3 定位结果及分析 .18 2.4 本章小结 .20 3 虹膜图像归一化与图像增强 .21 3.1 虹膜图像归一化 .21 3.2 虹膜图像增强 .23 3.3 本章小结 .26 4 虹膜特征提取和匹配 .27 4.1 纹理分析方法 .27 4.1.1 统计法 .27 4.1.2 结构法 .29 4.1.3 用空间自相关函数作纹理测度 .30 4.1.4 频谱法 .31 4.2
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 基于 纹理 分析 变换 虹膜 识别 算法 研究
限制150内