基于YCbCr肤色检测与AdaBoost联级算法的嘴部特征定位

Human's mouth location based on YCbCr complexion detection and AdaBoost cascade connection method

为了结合面部其他融合特征研究疲劳状态模式识别,主要研究了人的嘴部特征的定位及状态分析。对基于YCb Cr肤色结合Ada Boost联级算法采集帧图像定位人脸,在人脸定位的基础上,基于三庭五眼的先验知识对嘴部特征进行粗定位,采用迭代法自适应阈值分割法进行嘴部特征状态分析,利用最小外接法优化了嘴部状态参数。结果表明,提出的嘴部特征定位方法具有可行性、实时性。

This paper studied human＇s mouth location and status analysis in order to research pattern recognition of fatigue status combined with other fusion feature on the face. Firstly,the paper collected frame image of face based on YCbCr complexion and AdaBoost cascade connection method so as to locate the face position. Secondly,after locating face position,it roughly located mouth feature based on preexistent knowledge of three chambers and five eyes. And then it analyzed eyes status based on iterative and adaptive threshold segmentation method. Finally,it optimized the parameters of mouth status based on minimum external method and eyes status. The results show that the proposed method is feasible and real-time.