Mountain road winding and bumpy，traffic accidents caused by speeding frequently happened, mainly concentrated on curves. The present Curve warning system research are based on CCD, but the existing obstacles, weather , driving at night and road conditions directly affect the accuracy and applicability. The research is of predictability based on the geographic information and can told the driver road information and safety speed ahead of the road to reduce the incidence of accidents. In this paper, the main research contents include: to estimate forward bend curvature through the node classification method based on the digital map. Braking based on vehicle dynamics before entering the curve is used to identify the road. The critical safety speed which do not occur side-slip is calculated with the radius of curvature , side friction factor and so on using the vehicle lateral dynamics. The safe speed is also restricted by the safety evaluation of highway project and informed to the driver in advance. The pre-warning project can realize real-time dynamic information interaction between vehicle and road .Through the comparison and analysis under the same bend radius other model of safe speed, the results show that the model in terms of safe speed calculation provides a reasonable and accurate operation method, which helps drivers take an appropriate driving operation in time to insure safety.