基于手部特征的多模态生物识别技术研究
基于手部特征的生物多模态技术研究
摘 要
随着科学技术与时俱进的脚步，人们对于安全问题的关注性也越来越高，生物识别的技术优势越加突出了其重要性，生物特征识别就是指进行身份的判断时利用人类所特有的生理特征和行为特征。但同时基于单模态的身份识别技术却有着一些缺点，比如防伪性能差，易被窃取等，本文就这一问题而提出了多模态生物识别技术，它就是通过利用多种生物特征的融合技术进行身份识别和身份验证，主要优势在提高识别效率，抗噪性、普适性等方面都有所改善。虽然近年来多模态生物识别技术已经得到了多方面各个层次的关注和重视。但是，目前该技术仍然不是十分成熟，需要我们在以后的工作中进一步研究和探讨。
我们将具有丰富特征的人体手部作为本论文的研究重点，本文研究的手部的生物特征主要是基于掌纹特征和手形特征的，他们作为生物识别特征具有稳定性和唯一性，本文主要基于小波变换对生物特征进行分析，基于小波变换的生物特征提取方式，接着运用了改进后的典型相关分析方法，对所提取的手形和掌纹两种生物特征的特征向量在特征层的融合方法，此种方法主要是通过引入矫正系数，通过对投影方向上的特征矢量的调整，使准则函数值达到最大的要求。
本文主要利用人手部的生物特征识别技术，分析不同种方式下的掌纹特征的提取，使掌纹和手形的这两种生物特征模态在特征层进行融合，经过实验数据得出多模态生物特征的识别率要高于任一模态下的生物识别率，实现了多模态生物融合的初步探索和尝试，为我们后续的研究和发展提供了一定的了理论依据。
关键词 多模态生物识别；矫正系数；特征层融合；小波变换
-I-
Based on the biological characteristic of hand
multi-modal technology research
Abstract
With the pace of science and technology advancing with the times, people are more and more concerned about security issues. Biological recognition technology advantages more prominent its importance, biological feature identification is to identify the physiological characteristics and behavior characteristics of human beings when they are used for identification. But at the same time, there are some disadvantages in the single mode based identification technology, for example, poor security performance, easy to be stolen, etc. In this paper, we put forward a multi modal biometric technology. It is through the use of a variety of biological features of the fusion technology for identification and authentication, major advantages in improving the recognition efficiency, noise immunity, universality and other aspects have been improved. Although in recent years the modal biometric technology has been more and more attention of all levels. However, the technology is still not very mature, we need further research and discussion in future work.
We will have rich characteristics of the human hand as the focus of this paper, The hand biometric is mainly based on Palmprint and hand geometry, they are as a biometric characteristic with stability and uniqueness, this paper is mainly based on wavelet transform analysis on biological characteristics of proposed biometric ,The extraction method of biological features is based on Wavelet Transform, then puts forward the improved method of canonical correlation analysis, Two kinds of fusion method of feature vector and palmprint features in the hand shape biological feature level, this method is mainly by introducing the correction coefficient, to adjust the projection direction of the feature vector, the criterion function value reached the maximum.
-II-
In this paper, The use of biometric identification technology of the hand,We put forward the two modes of the palmprint and hand shape in feature level fusion, through experimental data obtained, multi-modal biometric recognition rate is higher than either mode of biological recognition rate, achieve and attempt the multimodal biometric melting of preliminary exploration ,to provide certain theoretical basis for our further research and development.
Keywords multimodal biometric recognition；correction coefficient；feature fusion；wavelet transform
-III-
目 录
摘 要 ........................................................................................................................... I Abstract ........................................................................................................................ II
第1章 绪 论 .............................................................................................................. 1 1.1 研究的背景及意义 .......................................................................................... 1 1.2 生物识别技术概述 .......................................................................................... 2 1.2.1 各种生物特征的分类和比较 .................................................................. 2 1.2.2 多模态生物特征识别 .............................................................................. 4 1.3 国内外研究现状 .............................................................................................. 5 1.4 论文研究的主要内容 ...................................................................................... 6 第2章 多模态生物识别中信息融合理论与方法 .................................................... 7 2.1 信息融合概述 .................................................................................................. 7 2.2 多模态生物特征识别研究 .............................................................................. 7 2.2.1 模式源的选择 .......................................................................................... 8 2.2.2 多模态识别中信息融合层次 .................................................................. 8 2.3 多融合方案确定 ............................................................................................ 13 2.4 本章小结 ........................................................................................................ 15 第3章 手部图像特征提取 ...................................................................................... 16 3.1 手部特征识别现状分析 ................................................................................ 16 3.2 手形轮廓特征提取与掌纹预处理 ................................................................ 17 3.2.1 手形轮廓特征提取 ................................................................................ 17 3.2.2 掌纹预处理 ............................................................................................ 20 3.3 基于小波变换图像特征提取 ........................................................................ 21 3.3.1 小波变换 ................................................................................................ 21 3.3.2 依据小波分解原理处理图像 ................................................................ 24 3.3.3 构造小波能量特征 ................................................................................ 26 3.3.4 特征匹配 ................................................................................................ 27 3.4 基于2DPCA的特征提取 ................................................................................ 27 3.5 本章小结 ........................................................................................................ 28 第4章 基于改进的多模融合识别 .......................................................................... 29 4.1 典型相关分析的基本思想 ............................................................................ 29
- IV -