基于全局和局部特征集成的低分辨率人脸识别方法  

Low-resolution face recognition methods based on global and local feature integration

作　　者：孙立强[1] 刘晓霞[1] 刘文学[1] SUN Liqiang;LIU Xiaoxia;LIU Wenxue(Qingdao Vocational and Technical College of Hotel Management,Qingdao 266100,China)

机构地区：[1]青岛酒店管理职业技术学院,山东青岛266100

出　　处：《无线互联科技》2024年第23期129-132,共4页Wireless Internet Science and Technology

摘　　要：低分辨率图像中人脸特征模糊不清,导致人脸识别结果准确率较低,因此,文章提出基于全局和局部特征集成的低分辨率人脸识别方法。该方法采用主成分分析法提取低分辨率人脸图像的全局特征,通过LBP算法提取低分辨率人脸图像的局部特征,构建一个包含特征集成模块与分类识别模块的卷积神经网络,输入提取的全局和局部特征,经过学习输出低分辨率人脸识别结果。实验结果表明,该方法使低分辨率人脸识别结果的正确率高达94%,是有效且优越的。Due to the blurred face features in low resolution images,the accuracy of face recognition results is low,and the low resolution face recognition method based on global and local feature integration is proposed.Principal component analysis method to extract the global features of low resolution face image,through the LBP algorithm of local features,build a convolution neural network containing feature integration module and classification recognition module,input extraction of global and local features,after learning output low resolution face recognition results.The experimental results show that the accuracy of the low-resolution face recognition results under the design method is up to 94%,and the proposed method is effective and superior.

关 键 词：全局特征 局部特征 特征集成 低分辨率 人脸识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]