使用vision.cascadeobjectdetector检测视频帧中的人眼的位置。检测器使用提琴琼斯检测算法和检测训练分类模型。人脸跟踪使用KLT算法.能够减少人头倾斜造成的影响,只有第一次检测人眼，以后都是检测特征点，运算速度快。-Use vision.cascadeobjectdetector detects video frames eye position. The detector uses a detection algorithm Tiqinqiongsi classification model training and testing. Face tracking using KLT algorithm can reduce the impact caused by the tilt of the head, only the first detection of the human eye, after all detected feature points, computing speed.