摘要
使用摄像头采集视频图像,对输入图像做预处理(图像灰度化、中值滤波等);通过学习训练的方法构造基于类Haar特征的层叠式分类器,利用基于类Haar特征的层叠式分类器从输入图像中直接定位人眼;把人眼部分的图像截取出来,二值化人眼图像;然后计算二值化图像中垂直方向上瞳孔的宽度大小,从而判断眼睛的状态;最后通过多次的捕捉,计算眼睛闭合的频率来得出其疲劳状态。试验验证了上述算法的有效性。
With help of Haar-like cascaded classifier designed by Adaboost algorithm and constructed by training method, driver's eyes were located by the pretreated input video images. Fatigue states of driver's eyes were classified by the frequency of blinking. In order to calculate the frequency, the open degrees of the driver' s eyes were calculated first, and the eyes' images were transferred into binary images, then the pupil widths on vertical direction of the binary images were calculated to judge the fatigue states of eyes. Experiments proved its validity.
出处
《公路交通科技》
CAS
CSCD
北大核心
2008年第7期128-131,共4页
Journal of Highway and Transportation Research and Development
基金
国家自然科学基金项目(60472006)
关键词
交通工程
疲劳检测
类Haar层叠式分类器
眼睛定位
图像处理
traffic engineering
fatigue state detection
Haar-like cascaded classifier
eyes location
imageprocessing