Abstract: A novel method of fast face detection algorithm based on Harr feature cascade strong classifier and skin color model was proposed. By using cascade strong classifier, human faces were detected rapidly based on the characteristics of the face Haar. The face region to sentence was selected by using skin color clustering model. The experiment led up to the fact that false detection rate is low and the speed for detection is fast and the algorithm is valuable in practical use.

Key words: Haar wavelet transform , Face detection , AdaBoost cascade classifier , Feature scaling and translating

摘要： 提出了基于Haar特征级联强分类器和肤色模型校验的快速人脸检测算法。利用基于人脸Haar特
征的级联强分类器快速检测人脸,得到待判人脸区域,其中可能含有非人脸误检区域;利用肤色聚类模型对待判人
脸区域进行筛选,过滤误检的非人脸区域。实验结果表明,该算法误检率低,检测速度快,在实际应用中有一定的价
值。

关键词: Haar小波变换 , 人脸检测 , AdaBoost级联分类器 ,  , 特征缩放与平移