Driver's risk perception characteristics are uncertain factors in the road traffic environment. To improve driving safety, the quantification of risk perception of drivers with different styles is studied. First, typical dangerous scenarios with different risk states are designed. The data of drivers' subjective cognition and objective behavior are obtained through questionnaires and driving simulation. The Pearson correlation analysis is used to determine the evaluation indicators of driver's risk perception. The fuzzy C-means clustering model is used to classify drivers into four styles: radical, mature, ordinary and conservative. A multiple linear regression model is constructed to analyze the change rule of risk perception of drivers with different styles in different risk scenarios. The results show that the risk perception of aggressive drivers is smaller than that of other types of drivers, and the driver's risk perception increases with the increase of scene risk level, in addition, the driver's style changes with the change of risk scenarios. This study can provide a theoretical basis for the driver's personalized strategy of safe driving.