摘要
研究了基于灰度图像的人脸检测问题 .采用小波变换方法提取人脸特征 ,大大地降低了特征矢量的维数 .使用交叉检验方法有效地解决了支持向量机训练时的参数估计问题 ,所设计的系统可以分别进行离线训练和在线检测 ,并且具有再学习的功能 .实验结果表明 ,该系统具有较高的正检率和较低的虚警率 。
A gray-image face detection system is presented. By using wavelet transform, the dimension of feature vector is extremely decreased. The parameter of support vector machine is generally selected by experience. This paper shows that cross validation method can effectively solve the parameter estimation problem. The designed system can realize offline training, online detection and progressive learning. Experiment results and empirical comparisons show that the system can approach high detection rate and low false alarm. In addition, the performance of the system can be improved by further learning.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2002年第9期947-950,共4页
Journal of Xi'an Jiaotong University
