摘要
为了改善人脸识别方法只基于一种特征、识别方法单一造成的识别率低的问题。使用多种特征融合进行人脸识别,可以有效改善单一特征因光照、角度以及尺度变化对识别的影响,提高识别率。经过试验证实,将LBPH、SIFT以及通过卷积神经网络提取的VIPLFaceNet特征按照一定的权重进行组合时,可以有效的结合3种特征的识别特点,获得比单一特征更好的识别率。当VIPLFaceNet、SIFT和LBPH3种特征以4∶1∶5的权重进行融合时,可以获得95.35%的识别率,识别率明显提升。
In order to improve the problem of recognition rate which is based on only one feature, a single identification method,the writer uses a variety of feature fusion for face recognition, can effectively improve the recognition rate of a single feature, which will be affected by light, angle, and scale changes. The experiments show that when LBPH, SIFT, and VIPLFaceNet features are combined with a certain weight, a better recognition rate can be obtained than a single feature. When the three features of VIPLFaceNet, SIFT and LBPH are fused with a weight of 4∶1∶5, the recognition rate of 95.35% can be obtained, and the recognition rate is obviously improved compared with the single feature.
作者
DOGNERY SINALY SILUE
DOGNERY SINALY SILUE(School of communication and Information Engineering,Shanghai University,Shanghai 20044,1,China;Institute of Smart City,Shanghai University,Shanghai 200444,China)
出处
《电子测量技术》
2018年第20期142-146,共5页
Electronic Measurement Technology
关键词
人脸识别
特征融合
多特征
face recognition
fusion of features
multi-features
