基于PCA、LDA和LR融合算法的人脸图像识别研究  

Research on Face Image Recognition Based on PCA,LDA and LR Fusion Algorithm

作　　者：王一朵 吕卫东[1] 胡陈陈 郑江怀 Wang Yiduo;Lv Weidong;Hu Chenchen;Zheng Jianghuai(Lanzhou Jiaotong University,College of Mathematics and Physics,Lanzhou 730070,China)

机构地区：[1]兰州交通大学数理学院,甘肃兰州730070

出　　处：《科学技术创新》2022年第22期72-75,共4页Scientific and Technological Innovation

摘　　要：为了让人脸算法能在降低计算复杂程度的前提下有着更高的识别准确率,尝试一种将PCA、LDA、LR融合的算法,首先利用PCA、LDA方法对人脸数据进行降维,再利用LR分类器进行混合人脸识别。这种融合算法改善了PCA和LDA这两种方法在光照不均匀时图像识别率低和无法求出最佳投影方向的问题,从而能够解决随着人脸数据的集中以及人脸样本类别的增多,而识别的有效性反而下降的问题;而LR分类方式作为传统机器学习中的一个十分重要且典型的分类模型,其算法本身简单易懂,而且分类过程准确高效,加上在人脸识别分类上的应用较少,有一定的研究意义。本文尝试在Wild数据库上利用python软件进行仿真模拟实验,发现该融合方法的识别准确率和传统的PCA算法以及KNN分类方法相比有显著的提高。In order to make face algorithm have higher recognition accuracy under the premise of reducing the complexity of calculation,this paper tries a fusion algorithm of PCA,LDA and LR.Firstly,PCA and LDA methods are used to reduce the dimension of face data,and then LR classifier is used for mixed face recognition.This kind of fusion algorithm improves the problem that PCA and LDA have low image recognition rate and can not find the optimal projection direction when the illumination is not uniform,so as to solve the problem that with the concentration of face data and the increase of face sample categories,the effectiveness of recognition decreases.LR class ification is a very important and typical classification model in traditional machine learning.Its algorithm itself is simple and easy to understand,and the classification process is accurate and efficient.In addition,it is rarely applied in face recognition classification,so it has certain research significance.This paper tries to use Python software to conduct simulation experiments on Wild database,and finds that the recognition accuracy of the fusion method is significantly improved compared with the traditional PCA algorithm and KNN classification method.

关 键 词：主成分分析(PCA) 线性判别分析(LDA) 逻辑回归(LR) 人脸图像识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]