实时人眼定位算法研究与设计

Research and Design of Real-time Eye Location Algorithm

为了实现准确和实时的人眼定位,提出了一种有效的人眼定位方法。采用基于统计肤色模型和人脸几何特征约束定位人脸,根据在YCbCr色度空间中肤色的聚类性独立于亮度,建立肤色模型,然后采用自适应阈值进行肤色分割。在人眼定位中,通过获取一个属于眼睛区域像素的随机样本,利用期望最大化算法(EM)调整椭圆型眼参数模型。在包含两个眼睛的矩形框内精确定位人眼,计算量小,速度快。实验结果表明,该方法能够准确、快速地定位人眼,可以满足非接触式、不同光照条件、实时的人眼定位系统的要求.

In order to achieve accurate and real-time eye location system,an efficient eye location method was proposed.First,a face location method based on a statistical skin-color model and face geometry constraints were adopted.Skin-color model was established according to skin-color clustering which is independent of the brightness in YCbCr chromatic space.Then,adaptive threshold for skin color segmentation was used.In the eyes location,by getting a random sample from the pixels that belong to the eye area and using the expectation maximization algorithm（EM）,an ellipic parameter model has been adjusted.Locating the eyes accurately in a rectangular box which contains two eyes,the calculation is smaller and the speed is higher.The experiment result shows that the method can accurately and quickly locate the eyes,meeting requirements of eye location system about non-contact,different light conditions and real-time.