摘要
论文提出一种基于二级神经网络的人脸检测算法,先用第一级神经网络筛选出可能的人脸区域,然后将该区域进行不同角度旋转,送入第二级神经网络,根据网络的输出值确定人脸的倾斜角度,最后用验证策略进行判定是否为人脸。对各种图像进行实验的结果表明,该算法对于检测正面端正人脸有较好的效果和较强的鲁棒性,检测正面多角度的人脸也很有效。
This paper presents a face detection system based on secondary neural networks.The first neural network is used to find areas that probably contain a face.Then these areas are rotated and sent to the second neural network.The orientation of the face is confirmed by the outputs of the second network.Finally,validation strategy is used to judge whether it is a true face.Experiment results show that the system is robust and can achieve high detection rate when detecting upright faces.It also does good to detect orientation faces.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第27期46-47,92,共3页
Computer Engineering and Applications
基金
国家自然科学基金项目(编号:60135010)资助
关键词
人脸检测
神经网络
多角度检测
验证策略
face detection,neural network,multi-orientation detection,validation strategy
