Abstract: Drowsy driving can be detected by the eye states. Eye location is the key problem for eye states detection. In order to improve real-time performance and solve uncertainty of head gesture, an eye lactation detection method based on dual spaces is proposed. Characteristics of skin-colors are used to detect face and saturation and lightness are used to detect the iris. Custom convolution template is used to locate no-close eye. Experimental results show that the correction rate of the algorithm is 94.76% and it is robust to faces with glasses, illumination changing, head gesture with good real-time performance.

Key words: skin-color detection, iris detection, convolution template, eye location

摘要： 驾驶疲劳可以从眼睛的状态中反映出来，而眼睛定位是判断眼睛状态的关键。为解决眼睛定位实时性和头部旋转不确定性问题，该文提出一种基于双空间的眼睛定位方法：在彩色空间中采用肤色特征快速检测人脸，在此基础上使用饱和度和亮度特征快速检测虹膜。针对非开眼问题，在灰度空间中使用自定义圆卷积模板法定位眼睛。实验结果表明，该方法定位正确率为94.67%，能够满足实时性要求，对戴眼镜、头部旋转、光照变化具有一定的鲁棒性。

关键词: 肤色检测, 虹膜检测, 圆卷积模板, 人眼定位

CLC Number: 

• TP391