基于CS-LBP特征融合的人脸智能识别方法研究  

Research on Intelligent Face Recognition Based on CS-LBP Feature Fusion

作　　者：张四平[1] 王梅[1] 胡念 Zhang Siping;Wang Mei;Hu Nian(School of software Hunan College of Information,Changsba Hunan 410200,China)

机构地区：[1]湖南信息职业技术学院软件学院,湖南长沙410200

出　　处：《信息通信》2020年第6期40-42,共3页Information & Communications

基　　金：湖南省教育厅科学研究项目(基金编号:19C1314)“基于云计算人脸表情特征提取与识别的研究”。

摘　　要：传统人脸智能识别方法存在识别正确率低的问题,为此提出了一种基于CS-LBP特征融合的人脸智能识别方法。通过分析人脸表情特征提取流程,设计CS-LBP特征融合目标计算,获取模糊特征LBP阈值。并结合不同LBP算子距离,均衡人脸图像灰度,以此提高人脸识别图像的清晰度。同时,绘制人脸识别标准纹路模型,计算经过识别后的人脸纹路与标准纹路的垂直距离,判断距离是否符合高斯单元线性组合规律,若符合规律可输出识别参数与结果,完成人脸的智能识别。采用设计实验的方式,验证了提出方法在实际识别应用中,可提高人脸图像识别正确的次数,证明了设计方法更具备实际应用价值。Traditional face intelligent recognition methods have the problem of low recognition accuracy.Therefore,a face intelligent recognition method based on CS-LBP feature fusion is proposed.Through analyzing the process of facial expression feature extraction,the CS-LBP feature fusion target is designed to obtain fuzzy feature LBP threshold.Combining with different LBP operator distance,the gray level of face recognition image is balanced to improve the clarity of face recognition image.At the same time,draw the standard pattern model of face recognition,calculate the vertical distance between the recognized pattern and the standard pattern,judge whether the distance accords with the Gaussian unit linear combination rule,if it accords with the rule,it can output recognition parameters and results,and complete the intelligent recognition of face.Experimental results show that the proposed method can improve the number of correct face image recognition in practical applications,and prove that the design method has more practical value.

关 键 词：CS-LBP 特征融合 人脸识别 智能识别 

分 类 号：TP039[自动化与计算机技术]