基于驾驶行为的疲劳驾驶检测方法研究

发布时间：2018-10-18 10:34

【摘要】：随着我国机动车数量的持续增长,道路交通安全问题也日益严峻,道路交通事故逐渐成为造成人类伤亡的主要原因之一,其中57%的灾难性事故与驾驶员疲劳驾驶有关。因此,加强疲劳驾驶检测技术的研究,防止疲劳驾驶行为的发生,对提高道路交通安全具有十分重要的意义。本文主要研究基于信息融合的疲劳驾驶检测方法,通过分析驾驶行为数据的变化特征来判断驾驶员的驾驶状态。首先,本文对国内外的研究现状进行了广泛调研,在总结前人研究的基础上,介绍了驾驶行为与疲劳驾驶的关系以及疲劳驾驶的形成机理和表现特征。并利用模拟驾驶平台开展驾驶实验,设计并完成了疲劳驾驶和正常驾驶两组实验,采集了25名驾驶员在不同驾驶状态下的驾驶行为数据,并对数据进行了整理与筛选,建立了疲劳驾驶样本数据库。其次,分析了驾驶员在不同驾驶状态下的驾驶行为特征。运用统计分析法对驾驶行为参量的时间序列变化趋势、均值和标准差进行了对比分析。并提出采用样本熵对驾驶行为数据的复杂度进行分析。通过研究,明确了驾驶员疲劳驾驶时的操作行为和车辆运行状态的变化特征,最终提取了速度、方向盘转角和车辆横向位置作为区分驾驶状态的特征参量。再次,依据模式分类的基本原理,采用KNN方法建立了基于单参数的疲劳驾驶检测算法,并引入DTW距离对算法进行了改进。研究表明,基于单参数的检测算法对疲劳驾驶的识别准确率总体不高,但采用DTW距离改进算法的识别性能更好。最后,建立了基于信息融合的疲劳驾驶检测算法。提出了一种改进的加权投票法对基于单参数的疲劳驾驶检测算法进行了决策层融合。为与决策层融合方法进行对比,采用BP和GA_BP神经网络对多个驾驶行为特征进行了特征层融合。通过对比分析各疲劳驾驶检测算法的识别准确率与运行效率,发现基于加权投票法的融合算法和基于GA_BP神经网络的融合算法的识别效果均较好,但前者的识别效果更优。
[Abstract]:With the continuous growth of the number of motor vehicles in China, road traffic safety problems are becoming increasingly serious. Road traffic accidents have gradually become one of the main causes of human casualties, 57% of which are related to driver fatigue driving. Therefore, it is of great significance to strengthen the research of fatigue driving detection technology and prevent the occurrence of fatigue driving behavior to improve road traffic safety. In this paper, the fatigue driving detection method based on information fusion is studied, and the driver's driving state is judged by analyzing the changing characteristics of driving behavior data. First of all, this paper has carried on the extensive investigation to the domestic and foreign research present situation, has summarized the predecessor research foundation, has introduced the relationship between driving behavior and fatigue driving, as well as the fatigue driving formation mechanism and the performance characteristic. Using the simulated driving platform to carry out driving experiments, two groups of experiments of fatigue driving and normal driving are designed and completed. The driving behavior data of 25 drivers under different driving conditions are collected, and the data are sorted out and screened. A database of fatigue driving samples was established. Secondly, the characteristics of driver's driving behavior under different driving conditions are analyzed. The trend of time series, mean value and standard deviation of driving behavior parameters are analyzed by statistical analysis. Sample entropy is used to analyze the complexity of driving behavior data. Through the research, the operating behavior of the driver during fatigue driving and the changing characteristics of the vehicle running state are defined. Finally, the speed, steering wheel angle and the lateral position of the vehicle are extracted as the characteristic parameters to distinguish the driving state. Thirdly, according to the basic principle of pattern classification, the fatigue driving detection algorithm based on single parameter is established by using KNN method, and the DTW distance is introduced to improve the algorithm. The results show that the detection accuracy of fatigue driving based on single parameter detection algorithm is not high, but the performance of the improved algorithm based on DTW distance is better. Finally, a fatigue driving detection algorithm based on information fusion is established. An improved weighted voting method is proposed for decision level fusion of fatigue driving detection algorithm based on single parameter. In order to compare with the method of decision level fusion, BP and GA_BP neural networks are used to fuse the characteristics of multiple driving behaviors. By comparing and analyzing the recognition accuracy and running efficiency of each fatigue driving detection algorithm, it is found that the fusion algorithm based on weighted voting method and the fusion algorithm based on GA_BP neural network have better recognition effect, but the former has better recognition effect.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.254;TP18

【参考文献】