Abstract: At present, the fatigue problem of long time driving for the drivers of Urban Transit was solved by rules and alert button. These methods increase the labor intensity of the driver, the driver fatigue driving detection effect is limited. So this article put forward an active online real-time detection method based on the site work environment of drivers, the method was used to analyze the face information through a video sequence, identify and locate the human eye by using active appearance model, implement the detection of driver fatigue by PERCLOS Algorithm. Experimental results showed that the model and the Algorithm were able to implement the detection of driver fatigue.