基于深度学习的人脸遮挡检测方法  

Detection for masked face based on deep learning

作　　者：邵一鸣 孙红星[1] 陈虹羊 SHAO Yiming;SUN Hongxing;CHEN Hongyang(School of of Electronic and Information Engineering,University of Science and Technology Liaoning,Anshan 114051,China;School of Computer science and Software Engineering,University of Science and Technology Liaoning,Anshan 114051,China)

机构地区：[1]辽宁科技大学电子与信息工程学院,辽宁鞍山114051 [2]辽宁科技大学计算机与软件工程学院,辽宁鞍山114051

出　　处：《辽宁科技大学学报》2019年第6期454-461,共8页Journal of University of Science and Technology Liaoning

摘　　要：针对银行自动柜员机环境下不法分子通过遮挡面部实施违法犯罪行为的安全问题,采用将深度学习中MTCNN多任务卷积神经网络和ResNet残差分类网络相结合的方法,先利用MTCNN模型实现对图像中人脸的检测,再通过残差分类网络实现对已检测人脸有无遮挡的分类,最终实现人脸遮挡检测。实验证明本文方法优于传统方法,人脸遮挡检测率较高,具有一定的工程应用价值。To solve the security problem of criminals with face masks committing crimes in the environment of bank ATMs,a method was adopted by combining the multi-task convolutional neural networks(MTCNN)and the residual classification network(ResNet)in deep learning.First,the MTCNN model was used to detect the faces in the image,and then the residual classification network was used to classify covered and uncovered faces.Finally,the covered faces can be detected.The experiments show that this method is superior to the traditional methods in detecting masked faces with a high detection rate.Therefore,the new method has a certain value for engineering application.

关 键 词：深度学习 MTCNN多任务卷积神经网络 ResNet残差分类网络 人脸遮挡 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]