摘要
针对近红外人脸识别对表情和姿势变化缺乏足够鲁棒性的问题,提出一种基于Contourlet变换、非负矩阵分解NMF(Non-negative Matrix Factorization)与支持向量机(SVM)的近红外(NIR)人脸识别算法。该算法首先对NIR人脸图像进行Contourlet变换,接着用NMF进行分解,取其系数矩阵的一阶统计量作为特征数据,然后利用SVM进行分类与识别。实验结果表明,该算法具有较高的识别率,而且对人脸表情和姿势变化具有较强的鲁棒性。
In response to the problem that existing near infrared （ NIR ） face recognition methods lack insufficient robustness to the varietiesof facial expression and gesture, we describe a new NIR face recognition algorithm which is based on Contourlet transform, non-negativematrix factorisation （ NMF） and support vector machine （ SVM） .In the proposed method, Contourlet transform is applied to NIR face image, and then NMF is utilised to decompose each sub-image derived.The first-order statistics of the coefficient matrixes is taken for the features data.Then, SVM is employed in classification and recognition steps.Experimental results show that the proposed approach has higherrecognition rate, and has higher robustness to the varieties of facial expression and gesture as well.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第12期229-232,共4页
Computer Applications and Software
基金
陕西省自然科学基础研究计划项目(2009JM8003)
