关于驾驶员眼睛检测与跟踪的研究

Study on the eye location method and track algorithm of driver

本文采用红外LEDs和CMOS图像传感器获取人脸图像和眼睛候选区域,再用支撑向量机(SVM)眼睛分类器验证并确定眼睛的位置,完成对驾驶员眼睛的准确定位;在眼睛的跟踪上,针对Kalman滤波和Mean Shift理论本身的缺陷,提出Kalman滤波和Mean Shift相结合的跟踪算法,不仅提高了跟踪的效率和跟踪的鲁棒性,还实现了模板的自动更新。

We use IR（ infra- red ） LEDs and CMOS sensors to get the face image and the eye candidates region in the article, then use SVM eye classifier to identify the real eye regions accurate. To improve theoretic limitation of Kalman filter and Mean Shift, an algorithm for tracking of eyes, which combines Mean Shift and Kalman filter, is proposed. The proposed algorithm was found to be reasonably robust and accurate in tracking the eyes. And it update the model.