基于改进PCA和支持向量机的掌纹识别  

Palmprint Recognition Based on Improved PCA and SVM

作　　者：李昆仑[1] 张亚欣[1] 刘利利[1] 耿雪菲[1] 

机构地区：[1]河北大学电子信息工程学院,保定071000

出　　处：《计算机科学》2015年第B11期146-150,共5页Computer Science

基　　金：国家科技支撑计划项目(2013BAK07B04);河北省自然科学基金项目(F2013201170);河北省高等学校科学技术研究重点项目(ZD2014008)资助

摘　　要：掌纹识别是一种新兴的生物特征识别技术。掌纹识别是用掌纹特征(包括人眼可见的和不可见的)来进行身份鉴别的一种方法。其中掌纹特征提取和掌纹特征匹配是掌纹识别研究的关键部分和核心内容。在特征提取方面,给出了两种改进的特征提取方法。先对掌纹图像进行傅里叶变换,再对变换后的图像进行主成分分析;针对掌纹图像的特点,对PCA进行改进,设计了适用于掌纹图像的分块主成分算法。将一整幅掌纹图像分为若干子块图像,在此基础上进行主成分分析。通过实验验证了改进的特征提取方法可以提高识别准确率。在特征识别方面,模版匹配虽然在一定程度上计算量小,准确率高,但容易陷入小样本问题。因此通过训练SVM分类器,进行掌纹识别。实验证明该方法有较好的可行性。Palmprint recognition is an important part of biological feature recognition. Feature extraction and feature recognition are the main content of palmprint recognition. This paper made the improvement to the PCA algorithm based on principal component analysis. At first, palmprint image is processed by Fourier transform, and then the principal component analysis is used. Another method is that the palmprint image is processed by block principal component analysis. The improved feature extraction method was verified by the experiment. The results show that it can improve the recognition accuracy rate. In the aspect of feature recognition, although to a certain extent, template matching has small amount of calculation and high accuracy,it is easy to fall into the small sample size problem. This paper completed the palmprint recognition by training SVM classifier. Experimental results show that the method has better feasibility.

关 键 词：掌纹识别 主成分分析 改进的主成分分析 傅里叶变换 SVM分类器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]