摘要
针对佩戴口罩或者墨镜的交通违法人员难以采用常规的人脸识别方法识别的难题,采用深度卷积对抗神经网络生成外部组件的可视化特征向量的集合,结合特定人脸特征,形成多状态自适应的人脸特征表示。该技术实现从对象到场景的人脸特征分层表示,实现配搭口罩或者墨镜等场景下交通违法人员的人脸识别功能,提升执法的效率。
Aiming at the difficulty of face recognition for traffic offenders wearing masks or sunglasses, a deep convolutional adversative neural network is used to generate a set of visual feature vectors of external components, and a multistate adaptive face feature representation is formulated by combining specific face features. This technology realizes the hierarchical representation of facial features from objects to scenarios, and implements the face recognition function of traffic offenders in scenarios wearing masks or sunglasses, and thus improves the efficiency of law enforcement.
作者
杜翠凤
温云龙
李建中
DU Cuifeng;WEN Yunlong;LI Jianzhong(GCI Science&Technology Co.,Ltd.,Guangzhou 510310,China)
出处
《移动通信》
2019年第9期75-78,85,共5页
Mobile Communications
基金
广州市科技计划项目产业技术重大攻关计划(201802010042)
关键词
深度卷积对抗神经网络
多状态
自适应
人脸识别
deep convolutional adversative network
multi-state
adaptive
face recognition
