一种解决人脸识别误匹配灾难问题的方法  

A METHOD TO SOLVE THE PROBLEM OF FACE RECOGNITION MISMATCHING DISASTER

作　　者：张永宏[1] 曹健[1] 王丽华[1] 

机构地区：[1]南京信息工程大学信息与控制学院,江苏南京210044

出　　处：《计算机应用与软件》2014年第9期177-180,共4页Computer Applications and Software

基　　金：江苏省"六大人才"高峰高层次人才项目(2010-JXQC-132);教育部留学回国人员科研启动基金项目(教外司留[2010]609号);江苏省"青蓝工程"中青年学术带头人资助项目(20101005)

摘　　要：人脸识别过程中由于外界环境、样本数量以及扰动条件等因素的影响,分类器的分类错误率会出现急剧增加,导致误匹配灾难现象的出现。针对这一问题,采用一种改进型感知器与预识别算法,在人脸无拒识闭集识别的基础上改善和解决了这一问题。在VC++平台设计出的实时人脸识别系统上经过多次测试,结果表明,所设计的分类算法具有很好的自适应性与鲁棒性,误匹配率低、速度快,便于实际应用。Due to the influence of external environment,the number of samples and the disturbance conditions in process of facerecognition,the error rate of classifiers will increase dramatically,which results in the mismatching disaster situation.To solve this problem,we adopt a modified perceptron and pre-recognition algorithm,improve and solve this problem on the basis of face recognition without closedand rejection set.After several tests made on real-time face recognition system designed on VC ＋＋ platform,the results show that theclassification algorithm designed in this paper has good adaptability and robustness,low mismatching rate,fast speed and is convenient forpractical application.

关 键 词：人脸识别 分类器 误匹配灾难 感知器 预识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]