基于CLAHE的PCA-LDA典型地域人脸识别研究  

Study on Typical Regional Face Recognition Based on PCA-LDA with CLAHE

作　　者：何李 蒋行国 李嘉利 李德才 HE Li;JIANG Xingguo;LI Jiali;LI Decai(School of Automation and Information Engineering,Sichuan University of Science&Engineering,Yibin644000,China;Artificial Intelligence Key Laboratory of Sichuan Province,Yibin 644000,China)

机构地区：[1]四川轻化工大学自动化与信息工程学院,四川宜宾644000 [2]人工智能四川省重点实验室,四川宜宾644000

出　　处：《四川轻化工大学学报（自然科学版）》2023年第6期57-64,共8页Journal of Sichuan University of Science & Engineering(Natural Science Edition)

基　　金：四川省科技厅省院省校项目(2020YFSY0027);人工智能四川省重点实验室开放基金项目(2020RZJ03);四川轻化工大学人才引进项目(2019RC12)。

摘　　要：针对中国南方和北方典型地域人脸特征不明显而难以识别的问题,提出了一种基于限制对比度自适应直方图均衡化(CLAHE)的PCA-LDA算法。首先建立中国南北典型地域人脸数据集,利用CLAHE将人脸图像分块处理后,通过对比度拉伸实现局部信息增强与噪声抑制;然后,使用主成分分析(PCA)算法将高维的人脸图像映射到低维的空间,并生成最能体现样本地域特征的特征脸;之后,通过线性判别分析(LDA)算法寻找样本最佳投影方向进一步压缩维度;最后,采用支持向量机(SVM)分类器进行识别。最终结果表明,所提出的算法能够有效增强人脸特征并减少其他因素的干扰,对南方和北方典型地域识别率分别能达到64.0%和77.0%。Aiming at the problem that face characteristics of typical regions in southern and northern China are not obvious and difficult to identify,a PCA-LDA algorithm based on Contrast Limited Adaptive Histogram Equalization(CLAHE)has been proposed.Firstly,a typical regional face dataset in the south and north of China is established,and local information enhancement and noise suppression are achieved by contrast stretching after chunking the face images with using CLAHE.Then,the Principal Components Analysis(PCA)algorithm is used to map the high-dimensional face images to a low-dimensional space and generate the eigenfaces that can best reflect the regional characteristics of the sample.Subsequently,the Linear Discriminant Analysis(LDA)algorithm is used to find the best projection direction of the sample to further compress the dimensionality.Finally,the Support Vector Machines(SVM)classifier is used for recognition.The final results show that the proposed algorithm can effectively enhance face features and reduce the interference of other factors,achieving a recognition rate of 64.0%and 77.0%for typical regions in the south and north of China,respectively.

关 键 词：PCA-LDA 地域特征 特征脸 支持向量机 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]