基于人耳检测的身份识别方法研究

VIP免费

3.0 陈辉 2024-11-19 4 4 1.14MB 46 页 15积分

侵权投诉

摘要

随着社会科技的发展与进步，人们对于保密工作的需求越来越高。各种卡业

务、密码业务等已经发展的相对比较成熟。但是这些保密业务的安全级别还是有

待增强。伴随着人们对安全鉴定技术的要求的提高，生物技术被渐渐重视起来。

而近年来，生物特征识别也越来越引起人们的关注。人耳识别作为模式识别领域

的一个较为年轻的研究方向，对模式识别的发展和应用起着至关重要的作用。目

前，在国内研究方面的科研机构和人员还处于一个相对比较初级的阶段，具有巨

大的潜力和广阔的应用和发展空间。在国外，生物识别技术发展的比较早，但是

对于人耳识别的研究相对较少。

目前，对于人耳识别的技术还没有一套完整的理论体系，因此，其必然依赖

于传统的鉴别技术、图像处理技术和模式识别技术等。同时，人耳作为一种特殊

的生物特征有其固有的生理结构和位置，必然有其独特的处理方法和识别技术与

之相适应。本文采用在边缘检测基础上做特征提取，在这种混合算法的基础上做

人耳匹配等进行探索和研究，这是生物识别领域的新思路，目前在国内外还没有

类似的研究。在文中详细的介绍了人耳图像的滤波方法，包括 Wiener 滤波、自适

应滤波和 Garbor 滤波三种滤波方法，在比较中找出最适合人耳图像的滤波器；边

缘检测是本文的基础，在文中分别介绍了 Canny 算子边缘检测和罗盘算子边缘检

测，曾经有媒体称罗盘算子是优于 Canny 边缘检测的、性能最好的边缘检测方法，

本文将对这种方法做详细阐述；文中的特征提取方法采用 Hu 七阶不变矩方法，并

在人耳图像上做 Canny 边缘检测和七阶矩特征提取组合操作，保存其特征值。

一套完整的人耳自动识别系统一般包括以下几个步骤：人耳图像采集、图像

预处理、图像的特征提取、人耳图像的识别。本文重点讨论图像的特征提取部分。

该部分实质上分为两个部分：图像的边缘检测和特征提取。为了提高对动态人耳

图像的识别率，本文将采用几种边缘检测和特征提取方法做人耳识别，比较它们

的匹配率和识别效率。

本文的识别工作将在 DSP 实验平台上完成，采用 CCD 摄像头采集图像。该平

台是实时显示图像并得到识别结果的系统，因此我们可以很直观的看到人耳身份

识别效果。

关键字：人耳识别边缘检测特征提取模式识别 TMS320DM642

DSP

ABSTRACT

With development and progress of social science and technology, people are

getting higher and higher demand of the security work. A variety of cards, passwords

has been the development of relatively mature. However, the level of security of such

confidential business should be enhanced. With the improvement of security

authentication technology, biological technology has been gradually pay attention. In

recent years, biometrics is increasingly concerned. Ear recognition as a young research

orientation in pattern recognition filed, pays a vital role for the development and

application of pattern recognition. Currently, the domestic research institutions and

personnel of scientific research are still in a relatively early stage, and have great

potential and wide application and development space. In the foreign, the development

of biometric is relatively early, but recognition of the human ear is relatively little.

Currently, the technology for ear recognition does not have a complete theoretical

system. Therefore, it must rely on traditional identification technology, image

processing and pattern recognition technology. The human ear as a special biological

characteristics which own the inherent physical structure and location, must have the

unique corresponding identification technology. This paper makes feature extraction

from the human ear based on edge detection, explore and research human matching ear.

This is a new idea in the field of biometric identification. It is no similar study in

domestic and abroad. This paper introduce the filter method of human ear, include

Wiener filtering, Adaptive filtering, Garbor filtering, discover the most suits filter of

human ear image; edge detection is the basis for this paper, introduce the Canny

operator edge detection and Compass operation. Some media said that Compass

operator is better than the Canny operator in edge detection performance, this paper will

elaborate on this. In this paper, feature extraction method use Hu seven-order moment

invariant method, extract combination operation of Canny edge detection and

seven-order moment invariant in human ear image and preserve their characteristic

values.

A complete set of human ears recognition system generally include the following

steps: ear image acquisition, image preprocessing, image feature extraction, ear image

recognition. This paper focuses on the section of image feature extraction. This section

is essentially divided into two parts: image edge detection and feature extraction. In

order to improve the recognition rate of the dynamic human ear image, this paper will

use several edge detection and feature extraction method to do human ear recognition,

compare their matching rate and recognition efficiency.

The identification work will be completed in the DSP experimental platform, the

images are collected by CCD camera. The platform is a real-time display the image, get

the recognition results, therefore, we can see that the human ear identification results is

clear.

Keyword: human ear recognition, edge detection, feature extraction,

pattern recognition, TMS320DM642 DSP

中文摘要

ABSTRACT

第一章绪论······················································································· 1

§1.1 选题背景及意义······································································· 1

§1.2 人耳识别方法的国内外现状························································ 3

§1.3 人耳识别过程中亟待解决的问题·················································· 6

§1.4 本文的主要工作和内容安排·································································6

第二章图像预处理·············································································· 8

§2.1 引言·······················································································8

§2.2 人耳图像的灰度转换································································· 8

§2.3 图像滤波处理·········································································· 9

§2.3.1 维纳滤波和自适应滤波···················································· 10

§2.3.2 Gabor 滤波····································································12

§2.4 本章小结··············································································· 13

第三章边缘检测················································································14

§3.1 引言····················································································· 14

§3.2 边缘检测步骤·········································································14

§3.2.1 增强算法·······································································15

§3.2.2 边缘检测·······································································17

§3.2.2.1 Canny 算子边缘检测··············································· 18

§3.2.2.2 罗盘算子边缘检测·················································· 21

§3.3 本章小结··············································································· 22

第四章特征向量提取··········································································23

§4.1 引言····················································································· 23

§4.2 不变矩理论············································································ 23

§4.3 实验结果分析·········································································24

§4.4 本章小结··············································································· 26

第五章实验系统和实验结果分析·························································· 27

§5.1 引言····················································································· 27

§5.2 TMS320DM642 系统···································································27

§5.3 软件环境··············································································· 29

§5.3.1 软件环境·······································································29

§5.3.2 程序设计·······································································29

§5.4 实验结果··············································································· 31

§5.5 实验结果分析·········································································36

§5.6 本章小结··············································································· 36

第六章总结与展望·············································································37

§6.1 总结····················································································· 37

§6.2 展望····················································································· 37

参考文献····························································································38

在读期间公开发表的论文和承担科研项目及取得成果·································· 42

致谢·································································································· 43

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

15 积分 4人已下载

立即下载 VIP免费下载

摘要：

摘要随着社会科技的发展与进步，人们对于保密工作的需求越来越高。各种卡业务、密码业务等已经发展的相对比较成熟。但是这些保密业务的安全级别还是有待增强。伴随着人们对安全鉴定技术的要求的提高，生物技术被渐渐重视起来。而近年来，生物特征识别也越来越引起人们的关注。人耳识别作为模式识别领域的一个较为年轻的研究方向，对模式识别的发展和应用起着至关重要的作用。目前，在国内研究方面的科研机构和人员还处于一个相对比较初级的阶段，具有巨大的潜力和广阔的应用和发展空间。在国外，生物识别技术发展的比较早，但是对于人耳识别的研究相对较少。目前，对于人耳识别的技术还没有一套完整的理论体系，因此，其必然依赖于传统的鉴别技术...

展开>> 收起<<

基于人耳检测的身份识别方法研究.pdf

共46页,预览5页

还剩页未读，继续阅读

基于人耳检测的身份识别方法研究

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

推荐作者

热门标签

举报选择: