摘要
针对目前人脸识别在样本量、鲁棒性等方面所受的局限,提出一种基于CNN和SVM的人脸识别模型。通过构建CNN模型进行训练,对原始图像提取来自CNN不同深度的特征图,并将其进行加权融合。将融合后的特征作为最终特征输入SVM多分类器进行分类。实验结果显示,对于小样本数据集以及面部遮挡、光照变化数据集,特征的融合对模型精度提升明显,且与传统模型比较本方法识别精度更高。
In view of the limitations of face recognition in terms of sample size and robustness,a face recognition model based on CNN and SVM is proposed.By constructing CNN model for training,feature maps from different depths of CNN are extracted from the original image,and weighted fusion is performed.The fused features are input into SVM multi classifier as final features for classification.The experimental results show that for small sample data sets,face occlusion and illumination change data sets,the fusion of features can significantly improve the accuracy of the model,and the recognition accuracy of this method is higher than that of the traditional model.
作者
梁晴晴
LIANG Qingqing(Hebei University of Economics and Business,Shijiazhuang 050061,China)
出处
《现代信息科技》
2020年第19期62-65,70,共5页
Modern Information Technology
关键词
卷积神经网络
支持向量机
特征融合
人脸识别
convolution neural network
support vector machine
feature fusion
face recognition