基于YCbCr颜色空间相邻帧差建模的眼睛定位算法研究

目的：研究一种基于YCbCr相邻帧差法检测头部运动及眼睛位移、定位眼睛位置的算法,使之能在基于眼睛状态的疲劳驾驶监测中真正发挥作用。方法：利用YCbCr颜色空间改进相邻帧差法,并依据相邻帧差法对运动物体检测的原理,建立眼睛位移检测的模型。运用该模型检测出后一帧相对于前一帧的眼睛位移量,参照前一帧眼睛所在的矩形区域,框选出后一帧图像中眼睛的候选区域,再利用形态学方法处理,最终得到后一帧图像眼睛的精确矩形区域。结果：本算法的眼睛定位处理速度为14帧/s,较传统眼睛定位算法的9帧/s提升了55%,能确保实时检测到人眨眼瞬间（0.2～0.4 s）的时间间隙,且正确率与传统眼睛定位算法相当。结论：本研究的眼睛定位算法大幅提高了眼睛检测算法的速度,能确保实时检测到人眨眼瞬间（0.20.4 s）的时间间隙,提高疲劳监测指标PERCLOS的检测准确性。

Research on algorithm of positioning of the eyes based on adjacent frame YCbCr domain difference model

Objective： To research an eye precise localization method based on eye movement detection model, which is used to improve the speed of the localization method in traditional eye precise localization methods. Methods： Firstly, this paper improves temporal differencing with the skin color segmentation model and builds a model to detect the movement of eyes meanwhile. It＇s easy to get the movement range of eyes in adjacency photographs using the eye movement detection model in this paper and then we can get the rectangular region consisted by the eyes in the photograph which is to be detected. Finally, using the morphology method can help to get the accuracy region of eyes. Results： The experimental result shows that the speed of localization method in this paper increase to 14fps and the accuracy is almost the same as the traditional eye localization methods. Conclusion： The experiment demonstrates the eye precise localization in this paper realizes a higher speed and ensure the accuracy at the same time.