面向驾驶员状态监测的人眼凝视方向判别

Eye gaze direction discrimination for driver status monitoring

基于眼球运动规律监测驾驶员状态是当前研究的热点,而获取人眼凝视持续时间、眼球扫视运动频率等眼球运动参数的基础和前提是实时判别人眼凝视方向。首先采用红外光源和红外滤光片的装置,通过CCD摄像头获取"暗瞳孔"图像;接着根据人脸特征粗定位人脸区域,并通过对人脸上半部分水平积分投影粗定位人眼区域;然后在定位出的人眼区域内,利用闪光点与周围瞳孔区域灰度值差异大的特征精确定位闪光点,并在闪光点周围提取暗瞳孔精确定位瞳孔质心;最后建立凝视方向判别模型,根据闪光点与瞳孔的相对位置判别出人眼的凝视方向。凝视区域判别实验结果表明,该算法对于获取眼球运动相关参数有较好的应用前景。

Driver status monitoring based on the law of eye movement is the current hot spot. While calculating the eye movement parameters such as eye fixation duration and saccadic eye movement frequency, real-time eye gaze direction distinguishing is the key technology. Firstly, infrared light, infrared filter and CCD camera are used to obtain ＂dark pupil＂ image. Secondiy, according to facial features, face region is positioned and then from that region, eye region is located by using horizontal integration projection. Thirdly, as the gray value of glint is much higher than that of pupil region in the eye area, glint and pupil centroid are precisely located respectively. Finally, according to the relative position of glint and pupil, a gaze direction model is established to determine gaze direction. Experiments of gaze regional discrimination demonstrate that this algorithm has a better application prospects in obtaining eye movement parameters.