摘要
通过分析已有的掌纹识别方法和特征提取所面临的问题,提出了一种新的掌纹识别算法——直接监督保局投影(DSLPP)。该算法在传统的保局投影(LPP)算法中加入类别信息,同时对角化XLXT和XDXT,可以直接达到保局投影算法的最优准则,并且无须在原始高维数据(如原始图像)上先进行任何特征提取或降维处理。在PolyU掌纹库中进行实验,与Eigenpalm、Fisherpalm和LPP算法相比具有较高的识别速度和识别率;当掌纹库中图像总数为600张,共100人,每人用5张掌纹图像作为训练样本,1张掌纹图像作为测试样本时,可以达到100%的识别率。
By analyzing the existing methods of palmprint recognition and the problem of feature extraction, this paper pro- posed a new palm image feature extraction and recognition method based on direct DSL PP. This algorithm added the class information into the traditional LPP and diagonalized both XLXrand XDX^T simultaneously. Achiered the optimize Fishers criterion directly, without any feature extraction or dimensionality reduction steps on high-dimensional data ( such as raw images). Tested and evaluated the proposed method using the PolyU palmprint database. Experimental results show that DSLPP is more powerful than Eigenpalm, Fisherpalm and LPP algorithm for palm feature extraction and recognition, and the recognition rate is 100% using the total images of 600 , a total of 100 people, each with five palmprint image as a training sample, a palmprint image as a test sample.
出处
《计算机应用研究》
CSCD
北大核心
2009年第7期2777-2779,共3页
Application Research of Computers
基金
浙江省科技厅新苗人才计划基金资助项目(2007G60G2070010)
关键词
掌纹识别
主成分分析
线性判别分析
保局投影
监督保局投影
直接监督保局投影
palm recognition
principal component analysis (PCA)
linear discriminant analysis (LDA)
locality preservingprojections (LPP)
supervised locality preserving projections (SLPP)
direct supervised locality preserving projections(DSLPP)