非负矩阵分解与线性判别分析算法相融合的彩色人脸识别  

Non negative matrix factorization and linear discriminant analysis for face recognition algorithm of color combination

作　　者：李岚[1] 蔺国梁[1] 

机构地区：[1]兰州文理学院数字媒体学院,兰州730000

出　　处：《自动化与仪器仪表》2015年第4期196-198,共3页Automation & Instrumentation

摘　　要：针对彩色人脸中的环境影响,提出了一种新的识别方法。首先将彩色人脸图像在YCb Cr空间进行转换,增强了图像的聚类性;然后将转换后的图像进行NMF,得到对应基图像的分解系数;再经过LDA算法,找到投影矩阵Wopt,使得类间距最大,类内距最小,从而确定图像所属的类。实验验证此方法有效地提高了识别的准确率和效率。A new recognition method is proposed for the impact of color face environmental. Firstly, the color face image is converted into YCbCr space and enhanced the clustering. Then, the corresponding decomposition coefficient is obtained based on the NMF of transformed images. At last, the class is determined the image belonging to through the LDA algorithm and the projection matrix Woptis composed which make the class distance is maximum and the distance within the class is minimum. Experiments show that the accuracy and efficiency of recognition effectively is improved.

关 键 词：非负矩阵分解 彩色图像 YCBCR空间 线性判别分析 投影矩阵 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]