基于特征融合的指纹识别方法
VIP免费
摘 要
作为一种可靠的身份认证方式,指纹识别将生物特征、图像处理和模式识别
等方法有机的结合在了一起。鉴于信息产业化的发展和理论研究的丰硕成果,指
纹识别在人们的日常生活中有了越来越广泛的应用。目前的自动指纹识别系统
(Automated Fingerprint Identification System, AFIS)主要是基于指纹细
节点的匹配模型,但是由于图像采集环境等因素的影响,直接从指纹图像中精确
的提取细节点特征是十分困难的。为有效的减少伪细节点特征对算法的影响,通
常需要使用图像分割、指纹增强、脊线细化等方法对图像进行预处理。但这些图
像预处理方法要求的关键技术多、占用系统内存大,不利于指纹识别及时性的实
现。
基于图像纹理识别的方法是解决这一问题的有效途径。该类算法抗噪声能力
强,可直接对指纹进行特征提取,由于不需要复杂的图像预处理过程而显著的加
快了指纹识别的速度。此外,由于指纹图像纹理的唯一性,该类算法同时也具有
较高的可靠性。因此,指纹图像的纹理识别方法近年来迅速成为指纹识别领域的
一个热点研究方向。
但是,单一的图像纹理特征并不能充分描述图像的整体特性,因而通常不能
取得较好的准确识别率。提取图像的多种纹理特征来弥补彼此各方法的不足,使
用一种特征融合的方式进行指纹识别是解决这一问题行之有效的方法。此外,不
同类型的指纹图像在纹理走势上具有明显的差异,这种差异可以用 Henry 分类方
法将指纹图像分为五类:左箕型、右箕型、斗型、拱型、尖拱型。我们计算指纹
图像小波变换各子图的图像方差,从实验结果发现,纹型的差异直接体现在小波
变换的某一个子图上。这说明,把指纹图像进行二维离散小波分解,以特定纹型
为参考选择纹理最丰富的小波子图,一方面有利于提取到有效指纹特征,另一方
面也大大的降低了指纹识别的计算量。
基于上述讨论,本文提出了一种基于特征融合的指纹图像纹理识别算法。该
算法首先对指纹图像进行分类,然后对其进行三层离散小波分解,根据该指纹纹
型提取相应小波变换子图,并提取子图的灰度共生矩阵特征和改进不变矩特征。
最后,建立三层神经网络,使用无噪声和有噪声图像交替学习的方法进行网络训
练,待网络稳定后再将提取的特征向量输入建立的神经网络进行识别。建立指纹
识别数据库对本文算法进行测试,准确识别率、系统消耗时间等各项指标均得到
了较好的实验结果。
关键词:指纹 分类 小波变换 不变矩 神经网络
ABSTRACT
As a reliable personal authentication method, fingerprint recognition dynamic
combine biological characteristic, image processing and pattern recognition. With the
rapidly developing of information technology and the achievements of theotetical
research, fingerprint recognition have more and more application in people’s life.
Nowadays, automated fingerprint identification system largely base on the minutiaes
matching model. But because of the influence on the image acquisition environmental
factors, it is very difficult to extract the fingerprint minutiaes from the image directly.
Image segmentation, fingerprint enhancement, ridge thinning and the other methods
for image preprocessing can effectively reduce the false minutiaes. However, these
methods may require key technologies too much, and it is to the disadvantage of
fingerprint recognition timeless.
Fingerprint recognition based on the image texture is an effective way to solve
these problems. These recognition algorithm using image texture has the advantages
of strong ability of resisting noises, and as a result of without preprocessing before
feature extraction, the speed of fingerprint recognition on line will be increased
remarkably. For this reason, image texture recognition algorithm becomes a hot topic
in recent years.
However, a single image texture features does not adequately shows the image
properties, the result also can not perform very well. Multi-feature fuison using
differernt texture feature from the same image can resolve the trouble mentioned.
Furthermore, the different fingerprint has a diverse texture trend. According this
fingerprint characteristic, Henry fingerprint classification divide image to left loop,
right loop, whorl, arch and tented arch. We calculated the image variance in wavelet
sub-image that the image decomposed by 2D-DWT, analysis results shows that
wavelet sub-image has obviously discrimination accout of the diverse fingerprint
pattern. It is mean that searching for the sub-image in fingerprint recognition’s most
favor though specific fingerprint pattern is feasible. And also, this way can reduce the
comptuing effort.
For reasons outlined above, this paper proposed a fingerprint texture recognition
based on feature fusion. Firstly, the fingerprint image classification. And then, the
image is analyzed by wavelet for three layers, feature extraction based on GLCM and
advanced invariant moment from the wavelet sub-image which selected according to
the fingerprint pattern. Finally, we employ the three layers neural network to
fingerprint recognition which network training under noise free image feature
condition and nosie image feature condition. Experiment results demonstrate that the
less system consumes time and high correct recognition rate.
Key Words: fingerprint, classification, wavelet transform, invariant
moment, Neural network
目 录
摘 要
ABSTRACT
第一章 绪 论 ........................................................ 1
§1.1 选题的目的和意义 ........................................... 1
§1.2 指纹识别的基本原理 ......................................... 2
§1.2.1 指纹识别的过程 ........................................ 3
§1.2.2 性能评价指标 .......................................... 4
§1.3 指纹识别发展史和研究现状 .................................... 5
§1.4 本章小结 ................................................... 7
第二章 基本理论 ...................................................... 8
§2.1 小波变换 ................................................... 8
§2.1.1 一维小波变换 .......................................... 9
§2.1.2 二维小波变换 ......................................... 10
§2.1.3 二维多分辨分析 ....................................... 10
§2.2 灰度共生矩阵 .............................................. 13
§2.3 图像改进不变矩 ............................................ 15
§2.3.1 图像不变矩特征 ....................................... 15
§2.3.2 图像改进不变矩特征 ................................... 17
§2.4 神经网络 .................................................. 17
§2.4.1 神经网络简介 ......................................... 18
§2.4.2 神经网络结构及分类 ................................... 19
§2.5 本章小结 .................................................. 23
第三章 特征提取 ..................................................... 24
§3.1 指纹分类 .................................................. 24
§3.2 图像方差 .................................................. 28
§3.3 特征分析 .................................................. 29
§3.4 本章小结 .................................................. 32
第四章 实验及结果 ................................................... 33
§4.1 神经网络 ................................................... 33
§4.2 特征提取及网络训练 ........................................ 36
§4.3 实验结果及分析 ............................................ 38
§4.4 本章小结 .................................................. 40
第五章 总结与展望 ................................................... 41
§5.1 本文的主要工作和成果 ...................................... 41
§5.2 指纹识别发展前景的展望 .................................... 42
参考文献 ............................................................ 43
硕士期间发表的论文及承担项目 ........................................ 46
致 谢 ............................................................... 47
摘要:
展开>>
收起<<
摘要作为一种可靠的身份认证方式,指纹识别将生物特征、图像处理和模式识别等方法有机的结合在了一起。鉴于信息产业化的发展和理论研究的丰硕成果,指纹识别在人们的日常生活中有了越来越广泛的应用。目前的自动指纹识别系统(AutomatedFingerprintIdentificationSystem,AFIS)主要是基于指纹细节点的匹配模型,但是由于图像采集环境等因素的影响,直接从指纹图像中精确的提取细节点特征是十分困难的。为有效的减少伪细节点特征对算法的影响,通常需要使用图像分割、指纹增强、脊线细化等方法对图像进行预处理。但这些图像预处理方法要求的关键技术多、占用系统内存大,不利于指纹识别及时性的实现...
相关推荐
-
VIP免费2025-01-09 4
-
VIP免费2025-01-09 4
-
VIP免费2025-01-09 4
-
VIP免费2025-01-09 4
-
VIP免费2025-01-09 4
-
VIP免费2025-01-09 5
-
VIP免费2025-01-09 4
-
VIP免费2025-01-09 4
-
VIP免费2025-01-09 5
-
VIP免费2025-01-09 4
作者:牛悦
分类:高等教育资料
价格:15积分
属性:50 页
大小:981.13KB
格式:PDF
时间:2024-11-19