项目名称: 安全人脸识别系统中的特征保护方法研究
项目编号: No.61262056
项目类型: 地区科学基金项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 齐永锋
作者单位: 西北师范大学
项目金额: 43万元
中文摘要: 人脸识别在司法、金融和人机界面等方面具有广泛的应用前景和理论研究价值,一直以来是模式识别领域的研究热点。但是提取特征的鉴别性和对特征的保护是制约其推广的瓶颈。本项目拟采用随机投影方法产生可撤除的人脸模板,以保护存在于系统模块中的人脸特征;通过采用多数字水印技术,确保在传输通道中人脸特征的真实性。主要创新研究包括:(1)基于局部多值微分模式的光照不变人脸特征提取方法研究;(2)基于随机投影的双向二维可撤除人脸特征提取方法研究;(3)基于脆弱数字水印和鲁棒数字水印的双水印人脸特征真实性验证方法研究。通过对存在于系统模块中人脸特征的可撤除性研究和在传输信道中人脸特征的真实性认证研究,本项目拟构建多层次的人脸特征保护方案,为生物特征识别方法的推广提供理论支持与技术基础。
中文关键词: 人脸识别;特征提取;特征保护;;
英文摘要: Face recognition has broad application prospects and theoretical research value in terms of judicial, financial and human-machine interface, it has been a hot research field of pattern recognition. But the discriminablity of the feature extracted and the feature protection are bottleneck restricting the promotion. By adopting random projection method to obtain cancelable face template which can protect face feature in thesystem module, and multiple digital watermarking technilogy is adopted to ensure the integrity of face feature in transmission channel. The main innovations include: (1) the research of illumination invariantface feature extraction based on multiple local derivative pattern ; (2) the resaerch of cancelable face feature extraction based on two directional two dimensinal random protection; (3) the research of face feature integrity verifiction in transmission channel based on fragile and robust watermarking.By study the cancelability of face feature in system modual and the integrity verifiction in transmission channel, this project is going to propose a multiple level face feature protection scheme which would provide theoretical support and technology base for the promotion of biometric identification methods.
英文关键词: Face recognition;Feature extraction;Feature protection;;