基于Adboost与空间支持向量域分类的人脸检测
Face detection based on Adboost and Space Support Vector Domain classifier
杨定礼 1赵正敏 2白秋产1
作者信息
- 1. 淮阴工学院电子与电气工程学院,江苏淮安,223003
- 2. 淮阴工学院电子与电气工程学院,江苏淮安,223003;南京大学声学研究所现代声学重点实验室,江苏南京210093
- 折叠
摘要
为了在AdaBoost算法基础上进一步提高人脸检测率,提出首先运用AdaBoost算法对样本进行训练得到T个分类器,然后通过空间支持向量域分类(SSVDC)方法找到T个分类器的超球半径以及球心.同时,为了提高检测速度,首先对彩色图像进行肤色分割,去掉背景以及非肤色区域,然后计算所测样本的对应T个分类器的特征值,并计算其到各个超球球心的距离,并根据其与超球半径的关系来判断是否为人脸.在ORL人脸库、YALE人脸库以及CMU+MIT人脸库中进行实验.实验结果表明:本文算法比AdaBoost算法具有更高的检测速度与检测率,检测率可达到94.4％.
Abstract
In order to improve the face detection rate on the basis of the AdaBoost algorithm,a new approach is presented based on Space Support Vector Domain Classifier (SSVDC).Firstly,N classifiers are obtained through the AdaBoost algorithm,then hyper-sphere center and radius of those classifiers are calculated based on SSVDC.Skin color segmentation is applied to remove the background and non skin color region in order to improve the face detection speed.Then the distance from the test sample to the sphere center is used to determine whether it is human face.Based on face databases of the ORL,YALE,and CMU+MIT,experiments demonstrate that this algorithm detects face faster than AdaBoost algorithm and detection rate can reach 94.4％.
关键词
空间支持向量域分类器/AdaBoost/肤色分割/人脸检测Key words
SSVDC/AdaBoost/skin color segmentation/face detection引用本文复制引用
基金项目
教育部资助项目(20093218120021)
出版年
2013
NETL
NSTL
维普
万方数据