摘要
在特征提取过程中,样本图像特别容易受到某些外部条件变化的干扰,如光照变化、拍摄角度以及表情姿态等,而这些变化大都依赖人脸的局部信息.为了提高算法在这些条件变化下的人脸验证效果,提出了基于局部投影信息的客户相关算法.新算法在充分利用局部保持投影和鉴别性局部保持投影的优势上,进一步结合客户相关方法,将不同类别之间的差异表示得更加清楚,得到更具有判别能力的投影向量.通过在XM2VTS数据库上进行的人脸验证实验,证明了新算法在人脸验证方面的优势.
As image data are easily affected by the illumination, facial expression and poses, traditional subspace learning often fails to estimate reliably the low-dimensional. To improve the result of face verification, the algorithms of CSLPP and CSDLPP were proposed, which combine the merits of LPP, DLPP and client specific technique. Client specific subspace could clearly describe the difference among different classes and obtain the projecting vector of more robust discriminant ability. The experiment results obtained on the facial databases XM2VTS show the effectiveness of the proposed methods on face verification.
作者
许常青
XU Changqing(College of Internet of Things Engineering, Jiangsu Vocational College ofInformation Technology, Wuxi 214000, China)
出处
《河南教育学院学报（自然科学版）》
2018年第4期34-38,共5页
Journal of Henan Institute of Education(Natural Science Edition)
基金
江苏高校品牌专业建设工程资助项目(PPZY2015C239)
