一种新的多姿态人脸图像识别方法  

A New Multi Pose Face Image Recognition Method

作　　者：徐胜超 XU Shengchao(School of Data Science,Guangzhou Huashang College,Guangzhou 511300)

机构地区：[1]广州华商学院数据科学学院,广州511300

出　　处：《计算机与数字工程》2023年第2期468-471,478,共5页Computer & Digital Engineering

基　　金：国家自然科学基金项目(面上项目)“大数据平台计算安全保障机制研究”(编号:61772221);广州华商学院校内导师制科研项目“基于多模态人脸识别的人证合一系统的关键技术研究”(编号:2021HSDS15)资助。

摘　　要：由于常规已有方法未能对人脸图像进行去噪处理,导致人脸结果不准确,识别时间较长,为此结合局部奇异值特性,提出一种新的多姿态人脸图像识别方法。该方法首先通过小波包去噪方法对多姿态人脸图像进行阈值去噪,对人脸图像进行一层小波变换,去噪处理人脸图像,形成多个子图像,分别分解子图像和训练图像奇异值,在此基础上获取局部和整体特征向量,将两者进行结合,提取人脸图像的姿态信息,完成多姿态多人脸识别。仿真实验测试表明,该方法能够以较高的精度和较快的速度完成多姿态人脸图像识别,具有较好的性能。Because the conventional existing methods fail to denoise the face image,resulting in inaccurate face results and long recognition time,a multi pose face recognition method is proposed by integrating the local singular value characteristics.Firstly,the wavelet packet denoising method is used to denoise the multi pose face image.The face image is denoised by a layer of wavelet transform.The face image is denoised to form multiple sub images.The singular values of the sub image and the training image are decomposed respectively.On this basis,the local and overall feature vectors are obtained,and the two are fused to extract the pose information of the face and complete multi pose and multi face recognition.Simulation results show that the proposed method can complete multi pose face recognition with high accuracy and fast speed,and has good performance.

关 键 词：局部奇异值特性 多姿态 人脸识别方法 训练图像 奇异值分解 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]