Abstract: This study is conducted to explore the influence of road visibility on vehicle collision probability in haze weather. Firstly, relevant data of some road networks in Beijing were collected when haze weather occurs, including road visibility, average velocity of vehicles, driver's visibility and deceleration data, etc. Secondly, through analyzed the collected data, a deceleration characteristic model was constructed to reflect the deceleration characteristics of drivers with different visibility in each time periods when haze weather occurs. Further, a driver's vehicle speed model was constructed to calculate that the range of the vehicle speed exceeds the average road speed with different visibility in each time period when haze weather occurs.Finally, by comparing the average value of velocity difference features obtained from collected data with the actual collision probability, the high incidence regions of vehicle collisions in each time period when haze weather occurs are counted out. The results show that the road visibility is positively correlated with the average velocity of vehicles, and the influence of visibility should be fully considered when adjusting deceleration intensity and deceleration characteristic value. By comparing the values of the range beyond the speed limit, the time period when the vehicle speed needs to be focused on is pointed out.The real-time average value of vehicle speed difference features is positively correlated with the actual collision frequency, which can be used to estimate the vehicle collision probability. The visibility regions corresponding to the highest vehicle collision probability in each time period are different, and these visibility regions are the high incidence regions of vehicle collisions.

Key words: haze, visibility, vehicle collision probability, vehicle velocity model, road network data