基于颜色分割和HOG学习机制的人脸检测

Face Detection Based on Color Segmentation and HOG Feature

人脸检测在信息安全和处理等应用领域起到很重要的作用。提出了一种鲁棒的人脸检测算法。该算法结合了基于颜色分割的粗定位过程和基于Histogram of Oriented Gradient(HOG)特征的人脸精确定位过程。粗定位利用人脸的颜色特征,采用基于YCbCr空间的颜色分割方法,获得图像中有可能包含人脸的图像子区域;精确定位则采用HOG特征,并利用SVM进行分类,得到人脸的准确位置。实验证明我们的方法对不同亮度、各种年龄及性别、表情变化及部分遮挡的人脸都具有很强的鲁棒性。

Face detection plays an important role in information security and other applications.In this pa-per,an efficient novel approach is proposed to achieve automatic face detection.The detection method combines rough color segmentation with accurate face location based on HOG feature.The rough locating stage can obtain possible region of face using color segmentation based on YCbCr space.The exact locating stage,by HOG fea-tures and using Support Vector Machine（SVM）,searches faces in these possible regions.Experimental results show that the proposed approach can achieve robustness to illumination,scale,viewpoint change and even partial occlusion.