摘要
在通过生物特征对人脸认证识别时,针对支持向量数据描述存在不能紧密包裹、没有合适拒识机制、正确识别率不能逼近100%的问题,提出了一种具有合适拒识机制的高正确识别率分类器设计算法-基于同类特征点集和包裹点集的同类特征区域紧密包裹曲面的求解算法,设置所有紧密包裹面之外的公共区域为分类器的拒识区域,用ORL人脸库和扩展Yale B人脸库各自作对比实验表明,本文的方法在较小拒识率情况下,分类器正确识别率能逼近100%.
In face recognition by biometrics,there are some problems in support vector data description,such as not encapsulating tightly,without proper rejection mechanism,and the correct recognition rate can not approach 100%,a high correct recognition rate classifier design algorithm based on similar feature set and wrapped point set is proposed. The algorithm sets all the common areas outside the closely wrapped surface as the rejection area of the classifier. The comparison experiments between ORL face database and Yale B face database show that the correct recognition rate of the classifier can approach 100% under the condition of small rejection rate.
作者
孙玥
杨国为
SUN Yue;YANG Guo-wei(School of Electronic Information,Qingdao University,Qingdao 266071,China;College of Information Engineering,Nanjing Audit University,Nanjing 210000,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2020年第3期656-661,共6页
Journal of Chinese Computer Systems
基金
国家重点研发计划项目(2017YFC080-4000)资助
国家自然科学基金面上项目(61772277)资助
江苏省基础研究计划项目(BK20171494)资助.
关键词
生物特征
支持向量数据描述
人脸识别
拒识机制
紧密包裹曲面
同类特征区域
biometrics
support vector data description
face recognition
refusal mechanism
closely enveloped surface
similar feature regions
