基于PCA的眉毛识别方法研究  

Eyebrow Recognition Based on Principal Component Analysis

作　　者：李玉鑑[1] 王利娟[1] 

机构地区：[1]北京工业大学计算机学院,北京100022

出　　处：《计算机工程与科学》2008年第11期28-30,37,共4页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60775010);北京市属市管高等学校"中青年骨干教师培养计划"项目PHR(IHLB)

摘　　要：本文分析了人类的眉毛作为一种独立生物特征进行身份鉴别的可能性,提出了利用PCA方法进行眉毛识别的基本思路。通过32个人的小规模眉毛图像数据库,在直接利用纯眉毛图像或间接利用其傅里叶变换构造特征向量的两种情况下,分别研究了PCA方法在眉毛识别中的识别率与所采用的归一化方法和信息阈值之间的关系。实验结果表明,PCA方法的最高识别率在不采用傅里叶变换时为60.00%,而在采用傅里叶变换时为90.63%。This paper analyzes the possibility that human eyebrow works as an independent biometric for personal identification,and proposes the idea that PCA may be applied to eyebrow recognition. Using a small-scale database of eyebrow images taken from 32 persons, the problem of how PCA-based eyebrow recognition accuracy is related to the size-normalized methods and the information threshold has been studied respectively in the two cases： one constructing feature vectors from pure eyebrow images, the other from their Fourier Transform. The experimental results show that the highest accuracies are respectively 60. 00% in the first case and 90. 63% in the second one.

关 键 词：生物特征识别 眉毛识别 主成分分析 傅里叶变换 特征提取 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]