基于颜色特征的人脸检测方法

Face detection methods based on the color feature

针对RobCup家庭机器人对人脸检测的要求,研究了基于颜色特征的人脸检测方法。该方法首先将人脸图像进行非线性分段色彩变换光线补偿处理,减少光线对肤色的影响,然后在YCbCr颜色空间中建立肤色模型,分割出肤色区域。在颜色空间YCbCr中,嘴巴区域包含的红色分量要高于蓝色分量,利用这个特征分割出嘴巴区域。在YIQ颜色空间中,通过I分量来区分眼睛与皮肤,分割出人眼,最后根据嘴巴、眼睛的几何中心特征映射人脸,人脸检测的仿真结果验证了该方法的有效性。

The skin face detection method based on color image is studied according to the desired ability of face detection in Home Leagues of RobCup.In this method,human face image processed by the non-linear subsection color transformation is used to compensate light impact,it which reduces the impact of light on skin color,and then establish skin model is then established in the YCbCr color space,skin color segmented regions segmented.In the YCbCr color space,the mouth area contains more red components higher than the blue components,which is used to separate the mouth area is separated by using this feature.In the YIQ color space,the I component is used to distinguish the eyes and skin.Finally according to the mouth and the eyes geometry center characteristics,human face is mapped.Face detection simulation results verify the effectiveness of the proposed method.