基于车载信息融合的驾驶行为分析

发布时间：2018-04-10 06:38

本文选题：驾驶行为分析　切入点：OBD-II　出处：《哈尔滨工业大学》2014年硕士论文

【摘要】：目前随着我国经济的快速持续发展,大众私家车拥有量与日俱增,大量汽车驾驶初学者也随之加入到驾驶员行列中。由于大部分新手的驾驶技术不熟练,对车况不熟悉,并且交通安全意识薄弱,驾驶员个人因素与车辆因素已成为现阶段交通事故发生的主要原因。针对此种情况,急需驾驶辅助设备来及时提醒驾驶员车况信息并及时纠正驾驶员当前危险驾驶行为。为了解决此问题,本文主要研究了基于车载信息融合的驾驶行为分析方法。该方法分别从驾驶员角度以及车辆角度出发,再结合两部分的信息,在特定驾驶环境下,综合分析来判定当前驾驶状态。本文研究过程主要分为以下三个阶段。第一阶段利用驾驶训练模拟器以及车载信号模拟仿真平台分别采集驾驶员操作信息和车辆诊断系统(OBD-II)状况信息。第二阶段对已采集的驾驶操作与车况数据进行数据预处理和特征提取。第三阶段利用信息融合技术,本文提出了基于人-车-环境三者融合的驾驶行为分析方法。在该阶段中,分别利用特征层和决策层车载信息融合技术。在特征层上,首先从正面来分析驾驶状态。提出了一种基于人的驾驶行为分析判定方法。该方法了利用支持向量机(Support Vector Machine,SVM)结合方向盘转角变化速率与刹车油门踩踏速率两个特征分析当前驾驶行为是安全还是危险。其次,从侧面来分析驾驶状态。提出了一种基于车的驾驶行为分析判定方法。该方法利用强分类器(Ada Boost)结合车速转速相对比,节气阀转速相对比以及发动机负荷三个特征(OBD-II车载信息)分析当前驾驶行为是安全还是危险。在决策层上,根据基于人的驾驶行为分析结果和基于车的驾驶行为分析结果进行车载信息融合,利用BP(Back Propagation)神经网络算法对当前驾驶状态进行最终判定。经过实验可得,本文提出的方法的正确率为99%以上,说明了本文方法的有效性。
[Abstract]:At present, with the rapid and sustainable development of economy in China, the number of private cars is increasing, and a large number of beginners have joined the ranks of drivers.As most novice drivers are not proficient in driving skills, are not familiar with vehicle conditions, and the awareness of traffic safety is weak, drivers' personal factors and vehicle factors have become the main causes of traffic accidents at the present stage.In view of this situation, driving aids are urgently needed to remind drivers of vehicle condition information and correct driver's current dangerous driving behavior.In order to solve this problem, this paper mainly studies the driving behavior analysis method based on vehicle information fusion.The method is based on the driver's angle and the vehicle's angle, and then combines the information of two parts, and synthetically analyzes the current driving state in a specific driving environment.The research process of this paper is divided into the following three stages.In the first stage, driver operation information and OBD-IIs status information are collected by driving training simulator and vehicle signal simulation platform respectively.In the second stage, the collected data of driving operation and vehicle condition are preprocessed and feature extraction.In the third stage, using information fusion technology, a driving behavior analysis method based on the integration of man, car and environment is proposed.In this stage, the feature layer and decision-level information fusion technology are used respectively.At the feature level, first analyze driving status from the front.In this paper, a method of driving behavior analysis based on human is proposed.In this method, support Vector Machine (SVM) is used to analyze whether the current driving behavior is safe or dangerous, combined with two characteristics of steering wheel angle change rate and brake throttle trampling rate.Second, analyze the driving state from the side.A method of analyzing and judging driving behavior based on vehicle is proposed.This method uses the strong classifier Ada boost to analyze whether the current driving behavior is safe or dangerous by comparing the speed and speed of the vehicle, the speed of the valve and the load of the engine.At the decision level, according to the results of driving behavior analysis based on human beings and the analysis results of driving behavior based on vehicle, the information fusion is carried out on the vehicle, and the current driving status is finally determined by the BP(Back propagation neural network algorithm.The experimental results show that the accuracy of the proposed method is over 99%, which shows the effectiveness of the proposed method.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.25;TP202

【相似文献】