摘要
针对人脸姿态、光照和表情等各方面原因引起人脸识别率不高的问题,提出了一种基于单样本特征点形变成冗余样本的压缩传感人脸识别方法.将人脸图像信号进行小波变换得到系数的稀疏表示,采用高斯随机测量矩阵进行测量得到离散人脸单样本,基于特征点形变人脸三维模型生成冗余样本,通过稀疏特征点正交匹配追踪非线性重建算法重建冗余图像进行人脸识别.仿真实验结果表明,所提出的算法相对于同类算法,时间复杂度较低、精确度较高、鲁棒性较强,且随复杂环境变化,其优势更明显.
In allusion to the problem of face inaccurate recognition for various reasons of pose,illumination and expression,a method of face recognition based on compressed sensing was proposed. The sparse representation of face image was obtained by wavelet transform,the method of Gauss random measurement matrix was used to obtain the discrete face sample,the redundant samples were generated based on the feature point and the3 D model of the feature points. The face was recognized by the redundant image reconstructed by the sparse feature point orthogonal matching pursuit algorithm. The simulation results showed that the presented approach time complexity was lower,the accuracy was higher,and the robustness was stronger by compared other similar methods. And its advantage was more obvious in complex environment.
出处
《佳木斯大学学报（自然科学版）》
CAS
2015年第6期855-857,共3页
Journal of Jiamusi University：Natural Science Edition
基金
公安部科技创新项目(2013YYCXHNST035)
湖南省科技厅计划项目(2013GK3088)
湖南省哲学社会科学基金项目(11YBA123)
湖南省教育厅教改项目(湘教通[2014]247号-620)
关键词
压缩感知
稀疏特征
单样本
冗余样本
人脸识别
compressed sensing
sparse feature
single sample
redundant sample
face recognition
作者简介
周建华（1974-），男，湖南宁乡人，湖南警察学院信息技术系副教授，湖南大学博士研究生，主要从事压缩感知及图像处理研究．
