基于肤色分割和AdaBoost算法的人脸检测  

Face Detection Based on Skin Color Segmentation and AdaBoost Arithmetic

作　　者：雍宝虎[1] 李岚[2] 邓勇[1] 

机构地区：[1]江苏大学电气信息工程学院,镇江202013 [2]江南大学物联网工程学院,无锡214122

出　　处：《科学技术与工程》2013年第11期3119-3122,3225,共5页Science Technology and Engineering

基　　金：江苏省自然科学基金资助项目(04KTB310171);江苏大学高级人才启动基金(05TDG029)资助

摘　　要：针对复杂背景和可变光照下的静态彩色图像人脸检测,提出了一种基于肤色分割和AdaBoost算法的人脸检测方法。首先把彩色图像转换到YCbCr空间。接着应用自适应光线补偿算法对图像进行光线补偿。再结合形态学、几何约束等方法排除背景干扰、进行肤色区域分割。其次应用改进的AdaBoost算法对分割出的区域进行验证,进而精确定位人脸。实验表明:该方法检测率高、适应性好、鲁棒性强,对人脸检测有较强实用性。Focused on face detection of static color image under complex background and variable illumination, a method combining adaptive light compensation and AdaBoost algorithm is proposed. First, conversion of input im- age from RGB to YCbCr, then, the algorithm worked by doing light compensation using self-adaptive compensation algorithm, and then used morphologic methods, geometric restrict to make possible face area segmented and located the candidate＇ s face area. At last used improved AdaBoost arithmetic to test and indicate. The experiments show that the method has high accuracy, good adaptability, strong robustness to face detection problem.

关 键 词：人脸检测 光线补偿 ADABOOST算法 权值更新规则 多窗口策略 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]