摘要
提出了一种基于奇异熵与随机森林的人脸识别方法。该方法以奇异熵来描述人脸特征。首先在整个人脸图上进行奇异值分解,应用整体奇异熵提取人脸全局特征。然后将人脸图像分成均匀子块,在各子块中进行奇异值分解,应用局部奇异熵提取人脸局部特征。之后将整体奇异熵和局部奇异熵融合形成最终分类特征。最后通过随机森林分类器对其进行分类。在Yale人脸库上的实验表明,该方法对表情、光照具有鲁棒性,且有较高的识别率和较短的识别时间。
A face recognition method based on singular entropy and random forest is presented.This algorithm ex-tracts the facial feature by utilizing the singular entropy.Firstly,the singular value decomposition is performed on the whole face image,and the global feature is extracted through the whole singular entropy.Then the face image is divide into homogeneous sub-blocks.The singular value decomposition is performed on each sub-block,and the local features is extracted by the local singular entropy.Then the global singular entropy and local singular entropy are fused to form the final classification features.Finally , the random forest classifier is employed to classify the final fea-tures.Experimental results on Yale face database demonstrate that the proposed approach not only has high recognition rate and shorter recognition time but also has certain robustness to the expression and to the influence of light.
出处
《软件》
2016年第2期35-38,共4页
Software
基金
河南省教育厅科学技术研究重点项目(12B520039)
关键词
人脸识别
奇异值分解
奇异熵
随机森林
Face recognition
Singular value decomposition(SVD)
Singular entropy
Random forest