基于Hu矩和支持向量机的人脸与非人脸分类识别  

FACE AND NON-FACE RECOGNITION BASED ON HU-MOMENT AND SUPPORT VECTOR MACHINES

作　　者：白怀文[1] 王暄[1] 

机构地区：[1]陕西师范大学物理学与信息技术学院,陕西西安710062

出　　处：《计算机应用与软件》2011年第7期237-239,共3页Computer Applications and Software

摘　　要：区域特征和轮廓特征是表征灰度人脸图像的有效方法,而图像矩能够从整体上描述物体的几何特征。在考虑H u矩在图像形状模式表征的优势和支持向量机的分类能力的基础上提出了一种基于Hu矩和支持向量机的人脸检测方法,首先提取人脸与非人脸的Hu矩,作为图像特征,然后将特征送入支持向量机进行学习。训练采用了单样本和双样本两种方法,最后用训练得到的模型对测试样本进行测试,实验结果表明,该方法是可行的。Regional characteristics and contour feature are the effective methods to represent grey face image,and the image moment can describe the geometrical characteristics of the object entirely.Considering the advantages of Hu-moment for depicting the shape-model of images and the fine classification capacity of support vector machine,we proposed a novel method for face and non-face classification on Hu-moment and SVM basis.Firstly,the Hu-moment invariants of image（face or non-face samples） are extracted,and then the features of Hu-moment are used to train the SVM.Two means,the single samples and twin samples,were adopted as the training way in our paper.Finally,further examination on test samples was done with the model derived from the training.As the experiment shows,this proposed approach is effective.

关 键 词：HU矩 支持向量机 人脸与非人脸分类 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]