PCA在人脸识别中改进算法.pdfVIP

第6卷 第1期 54 中国科技论文在线 SCIENCEPAPER ONLINE 2011 年 1 月 PCA 在人脸识别中的改进算法 黄 昉 ，张宝昌，刘金琨 （北京航空航天大学自动化与电气工程学院，北京 100191 ） 摘 要：基于传统的主成分分析(PCA)方法，为使PCA不再局限于满足高斯分布，提出了改进的PCA人脸识别方法。 改进的PCA方法先对训练图像集进行分块，对分块得到的子训练图像集利用传统PCA进行分析，得到多个投影矩阵， 通过投影矩阵将训练图片和测试图片投影到特征空间。此外，为了利用样本之间的相关性，进一步提出了基于距离矩 阵的扩展方法，使得对训练样本分块时丧失的信息得以保留，进一步提高了系统的性能。在Feret人脸库上的试验结果 表明，所提出的方法在识别性能上明显优于传统的PCA方法，识别率得到了提高。 关键词：主成分分析；特征抽取；推广的PCA ；特征矩阵；人脸识别 中图分类号：TP391 文献标志码：A 文章编号：1673－7180(2011)01 －0054 －5 Promoted PCA and its application in human face recognition Huang Fang, Zhang Baochang, Liu Jinkun (School of Automation Science and Electrical Engineering, Beihang University, Beijing 100191, China) Abstract: Principal component analysis (PCA) is one of accepted and important technique for feature extraction widely used in the areas of images recognition such as human face recognition. Promoted PCA, a human face recognition technique based on PCA, is presented, in order that PCA is not only limited to Gaussian distribution. First, the original training set was divided into some sub-sets. Second the well-known PCA method was directly used to these sub-sets obtained corresponding transformed matrixes. Through these matrixes, training set and testing set were mapped in feature space. In addition, in order to make use of the correlation among samples, a promoted method based on distance-matrix is presented. With this method, the information lost during the