摘要
为了提高光照人脸识别正确率,针对传统特征提取算法存在的不足,提出一种子块加权保持近邻嵌入算法和相关向量机相融合的光照人脸识别算法。首先,将人脸图像划分成多个子图,然后,再采用保持近邻嵌入算法对各子块提取特征信息,并进行加权连接一个特征矩阵,最后,输入到相关向量机中进行分类识别,并采用ORL和Yale人脸库对算法的性能进行测试。仿真结果表明,其算法不仅提高了人脸识别的正确率,而且加快了人脸识别的速度,对光照变化具有较好的鲁棒性。
In order to improve the accuracy of illumination face recognition,an illumination face recognition based on block weighted neighborhood preserving embedding and relevance vector machines is proposed for the deficiency of traditional feature extraction algorithm in this paper.Firstly,the image is divided into many blocks,and neighborhood preserving embedding algorithm is used to extract features of blocks.Then they are weighted to form a feature matrix.Finally relevance vector machine is used to build face classifier,and ORL and Yale face databases are used to test the performance.The simulation results show that the proposed algorithm not only improves face recognition rate,but also accelerates face recognition speed,and it also has good robustness for illumination changes.
出处
《微型电脑应用》
2015年第7期62-65,6,共4页
Microcomputer Applications
关键词
人脸识别
保持近邻嵌入
子块加权
相关向量机
Face Recognition
Neighborhood Preserving Embedding
Block Weighted
Relevance Vector Machine