基于马尔可夫链的驾驶人视觉转移特征

Driver's visual transition characteristics based on the Markov chain

为了不割裂视觉转移过程的连续性,运用动态聚类方法,对驾驶人注视点的视野平面解析坐标进行聚类,实现驾驶人注视区域划分。在考虑驾驶人注视行为在时间轴上关联性的基础上,分析驾驶人注视行为在各注视区域间转移模式的齐次性,运用马尔可夫链理论,探讨各注视区域间视觉转移概率的求解方法,并对5名不同驾驶经验驾驶人的眼动数据进行统计,求得其一步转移概率矩阵和平稳分布,分析驾驶人视觉转移特征。分析结果表明:驾驶人对任一注视目标都需要多次重复注视才能获取足够的信息,且注视点主要集中在前方车道远处、前方车道近处、右侧车道和左侧车道4个区域。

In order to maintain continuity of visual transfer process, the dynamic cluster method was used to process the analytic coordinates of driver＇s fixation points in visual fields, and the driver＇s fixation areas were distinguished. Taking the relevance of fixation behaviors along time scales into consideration, the homogeneity of driver＇s visual transition pattern among different fixation areas was analyzed. The solutions of visual transition probabilities were exploited with the Markov chain theory, and the one-step transition probability matrix and the stationary distri- bution of five drivers wi, th different driving experiences were evaluated from the eye movement data. The results indicate that many repeated fixations are necessary for driver to get sufficient information and the most fixation points are grouped in the far side of front lane, the near side of front lane, the right lane and the left lane. 2 tabs, 5 figs,11 refs.