摘要
笔者设计了一种基于横向几何特征的人脸识别方法。先通过Dlib等软件工具从图片中提取一系列人脸特征点,然后计算这些人脸各特征点之间的横向距离,并以这些距离之间的比例值组成特征向量。用这些特征向量作为样本结合标签来训练搭建好的全连接网络。从本次对网络的训练结果来看,笔者所设计的人脸识别方法具有较好的识别效果。
The author designs a face recognition method based on transverse geometric features.Extracting a series of facial feature points from images through Dlib software tools,and then calculating the transverse distances between the feature points of face,and the values of the ratio between those distances are grouped as feature vector.These feature vectors used as sample combined with labels are used to train the full-connected network.According to the training results of the network of this time,the face recognition method designed in this paper has a good recognition effect.
作者
杨龙生
Yang Longsheng(School of Electronics and Information Engineering,Shanghai University of Electric Power,Shanghai 200090,China)
出处
《信息与电脑》
2018年第16期91-92,共2页
Information & Computer
关键词
几何特征
人脸识别
全连接网络
geometric features
face recognition
full-connected network
