基于人脸比例特征提取与匹配的身份鉴别  

Human recognition based on facial proportion feature extraction and matching

作　　者：张棋森 肖香苏 喻晓斌 张焱 Zhang Qisen;Xiao Xiangsu;Yu Xiaobin;Zhang Yan(College of Automation,Chongqing University of Posts and Telecommunications,Chongqing,400065,China)

机构地区：[1]重庆邮电大学自动化学院,重庆400065

出　　处：《电子测量技术》2020年第1期137-140,共4页Electronic Measurement Technology

基　　金：重庆市技术创新与应用示范产业类重点研发(cstc2018jszx-cyzdX0131);重庆市留创计划创新类项目(cx2018128)资助。

摘　　要：针对人脸识别身份鉴别中的小样本问题,提出基于人脸比例特征提取与匹配的身份鉴别方法。首先基于图像灰度分布统计特征方差分析为阈值对图像进行二值化处理,然后基于主动形状模型算法实现人脸检测与特征点定位,构建脸型、鼻型、眼型共7项人脸比例特征并作为身份鉴别特征向量,最后利用基于层次分析法的相似度匹配模型实现身份鉴别。在Yale和MD图像集上的测试表明,所提方法适用于复杂背景下人脸面部特征的提取和身份鉴别应用。Aiming at small sample challenge in face recognition human recognition, a human recognition method is studied based on the facial proportion feature extraction and matching. Firstly, binary processing the image based on the threshold of the calculated character variance of the image′s grey degree distributing;then face localization and facial features detection are carried out based on the active shape model algorithm, and seven kinds of facial proportion features, involving facial, nose and eyes, are calculated and used as the characteristic vector for human recognition;finally, a similarity matching model with analytic hierarchy process is employed for human recognition. The experiment results on Yale and MD image sets show that the proposed method is applicable for facial feature extraction and human recognition under complex background.

关 键 词：身份鉴别 人脸识别 比例特征 主动形状模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]