基于无侵入测量指标的个体差异化驾驶疲劳检测

发布时间：2018-04-24 02:27

本文选题：交通工程 + 疲劳检测　；参考：《中国公路学报》2016年10期

【摘要】：针对现有一些疲劳驾驶检测方法因未考虑个体差异而导致检测精度较低的问题,在考虑个体差异的基础上建立疲劳驾驶检测模型,以适应驾驶人在驾驶行为和生理特征上存在的个体差异。通过高仿真驾驶模拟器对值夜班8h后的受试人员进行疲劳驾驶模拟试验,采集每位驾驶人的数据并从中提取出19个驾驶行为指标和4个眼动指标。试验过程中通过主观问询方式获取驾驶人疲劳程度信息,在此基础上建立考虑个体差异的分层有序Logit(MOL)模型,以及未考虑个体差异的有序Logit(OL)模型和人工神经网络(ANN)模型来评估疲劳程度。结果表明:23个变量中,闭眼时间比例、平均瞳孔直径、车道偏移标准差、方向盘反转次数是疲劳检测中最重要的4个变量;3个模型中分层有序Logit模型的检测精度最高,表明考虑个体差异确实可以提高模型检测精度。
[Abstract]:Aiming at the problem that some existing fatigue driving detection methods do not take individual differences into account, a fatigue driving detection model is established based on individual differences. In order to adapt to drivers in driving behavior and physiological characteristics of individual differences. The fatigue driving simulation test was carried out on the subjects after 8 hours of night shift by using a high simulation driving simulator. The data of each driver were collected and 19 driving behavior indexes and 4 eye movement indexes were extracted from them. In the course of the experiment, the fatigue degree information of the driver is obtained by subjective inquiry, and then a hierarchical and ordered logit MOL model considering individual differences is established. The fatigue degree is evaluated by an ordered Logitelian (OLL) model and an artificial neural network (Ann) model which does not take into account individual differences. The results show that the ratio of eye closure time, the average pupil diameter, the standard deviation of lane deviation, the frequency of steering wheel reversal are the most important variables in fatigue detection, and the hierarchical and ordered Logit model has the highest detection accuracy among the 23 variables. It shows that considering individual differences can improve the accuracy of model detection.
【作者单位】：西南交通大学交通运输与物流学院;西南交通大学综合交通运输智能化国家地方联合工程实验室;同济大学道路与交通工程教育部重点实验室;
【基金】：“十二五”国家科技支撑计划项目(2014BAG01B03) 中央高校青年教师百人计划项目(2682015BR006)
【分类号】：U491.254

【相似文献】