一种有效的用于疲劳驾驶检测的人眼定位算法  

An Effective Eye Location Method for Drowsy Driving Detection

作　　者：胡德轩[1] 冯玉田[1] 宁萍强[1] 

机构地区：[1]上海大学通信与信息工程学院,上海200072

出　　处：《计算机技术与发展》2009年第12期9-12,共4页Computer Technology and Development

基　　金：上海市重点学科建设项目(T0102)

摘　　要：对疲劳驾驶检测系统来说,人眼定位对眼睛状态的正确判别起着关键作用。文中提出一种灰度图像中改进型眼睛定位算法。首先采用水平和垂直积分投影法大致定位出双眼窗口,然后综合运用动态阈值的模板卷积法和形态学操作来定位眼部黑块。随后通过黑斑验证,精确定位出眼球中心。实验图片全部取自CAS-PEAL和ORL双人脸数据库,结果表明该算法速度快,准确率高,稳定性强,对一定程度的头部旋转和戴眼镜等情况有较强的适应性,能应用于疲劳驾驶检测等需要快速定位人眼的场合。For a driver fatigue monitoring system, one of the most important problems to wive is eye location. Present an improved appreach to achieve fast and effective eye location. Firstly, horizontal and vertical integral projection is used to detect eye area. Then, dynamic tlreshold of binary- conversion of convolution template and morphological operations is used comprehensively to locate black area of eyes. After checking its validity, eye center can be located accurately. Experiment pictures conducted on CAS- PEAL and ORL faces show that this algorithm is time- saving and robust to a lot of complicated situation such as head rotation with a certain degree and face with glasses. This method can be used in occasions like driver fatigue monitoring system or any other occasions that need fast eye location.

关 键 词：眼睛定位 双向积分投影 动态阈值 模板卷积 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]