With the increasing development of the vehicle population and the transportation，the traffic accident has become a serious societal problem day by day. Statistics show that nearly 35% to 45% of the traffic accidents are due to drivers’ fatigue. So, it is important to judge whether the driver fatigue driving and to remind the driver. Aimed at the problem, through the acquisition of the blink frequency and eye movement information, the use of statistical methods to assess whether the driver fatigue, and warning will be given timely. The main tasks are described as following. By fixing the camera fixed on the instrument panel, locating the edge of the human eye, obtaining the distance between the upper and lower eyelids, calculate the frequency of the driver's blink. The velocity and position of eye movement is calculated by detecting the movement of the pupil. By monitoring the blink frequency and eye movement of the driver during a certain period of time, the normal eye movement model is established and the corresponding database is updated constantly. The real-time data of the same time period and the model calculation data are compared, and the warning is given when the eye movement flat rate is consistent with the sleep characteristics. The study found that when the driver is in the early stage of fatigue driving, blink frequency, the speed of eye movement and the active area will be reduced. According to the speed of the eye movement and the activity area as the main judgment basis and blink frequency as an auxiliary, the accuracy of the judgment of the driver's driving condition can be accepted. This will play a positive role in promoting the development of real time detection technology of driver fatigue.