关键词: 交通运输工程; 草原公路；视觉信息加工；马尔可夫链；信息熵；加权秩和比法

Abstract: In order to explore the difference in drivers visual information processing modes in each typical driving environment on prairie highway and analyze the driving safety on prairie highway from the visual level, the real vehicle experiment on prairie highway was conducted based on the eye tracker. The eye movement data of drivers in six real traffic environments were extracted, including free driving, vehicle-following, oncoming vehicles, rear vehicles cut-in, roadside risks and intersections. Firstly, the evaluation indicator system was constructed from two dimensions: visual allocation and visual search. The dynamic clustering method was used to divide the drivers visual area of interest into five categories, and the information entropy was combined with the Markov chain stationary distribution theory to construct the information entropy of fixation point distribution to quantify the complexity of the drivers visual attention allocation. Then the saccade amplitude and saccade velocity were introduced to analyze the drivers search process for visual information. Finally, the weighted rank-sum ratio evaluation model based on the CRITIC method was constructed to comprehensively evaluate the visual information processing intensity of drivers in different driving environments. The results show that the distribution of drivers fixation points varies in different driving environments, but the current lane is always the main viewing area for drivers to obtain visual information. The visual evaluation indicators show significant statistical differences with the change in the driving environment. When drivers encounter roadside risks, the visual information processing intensity is the largest, and the driving safety is relatively lower.

Key words: traffic and transportation engineering; prairie highway; visual information processing; Markov chain; information entropy; weighted rank-sum ratio method