基于数据挖掘技术的驾驶行为分析

发布时间：2019-01-17 11:24

【摘要】：汽车已经成为最普及的交通工具,同时中国每年发生的交通事故数量达数百万起,导致了巨大的人员伤亡与财产损失,因此交通事故的预防研究具有重大意义。在车辆设计,道路建设已经难以对事故预防有突破性进展的环境下,驾驶员的驾驶行为研究得到了越来越多的关注。然而目前还缺乏统一的对驾驶行为习惯描述的量化指标,通过量化指标来评判驾驶安全水平的研究成果也屈指可数。因此本文提出基于OBD(On-Board Diagnostics:车载诊断系统)技术,从驾驶行程、操作类型、行驶速度、行驶加速度、发动机转速五个维度分析驾驶行为习惯。利用长期采集的普通驾驶员的车辆运行数据,结合驾驶行为习惯具有长期稳定不变的特点,提取出驾驶行为习惯的量化指标。本研究的数据来源于“优驾智能盒子”采集的普通驾驶员长达2年以上的行驶数据,并且这些驾驶员的主要行驶区域为重庆地区。“优驾智能盒子”通过车辆普遍带有的OBD接口获取车辆的行驶数据,然后与智能手机通过蓝牙相连接,将重要的数据显示在智能手机上并将驾驶操作数据传入服务器中,使每位驾驶员的车辆运行数据可以不间断的记录保存。目前市场上的汽车大多数统一使用国际通用的OBDⅡ标准,使得“优驾智能盒子”可以通过OBD接口实时获取不同厂商,不同车型的汽车运行数据。随着OBD技术的发展,其能获取的车辆运行数据也在不断增加。本文从五个维度提取了与驾驶行为相关的车辆运行数据,并经过变换得到了57项指标,引入了金融领域的时间序列稳定性分析方法对这57项指标进行稳定性检验,得到了17项满足稳定性的指标,作为驾驶行为量化指标。并在提取的驾驶行为量化指标的基础上,还进行了驾驶员分类研究,研究结果显示:利用驾驶行为量化指标进行分类的驾驶员能够区分具有不同驾驶行为特征的驾驶员。本研究为驾驶行为习惯量化研究提供了理论依据,为事故倾向性研究提供了新的思路与方法。通过OBD方法获取的分析数据比传统实验方法获取的数据更容易,数据量更多,分析结果更具真实性与应用型。
[Abstract]:Automobile has become the most popular means of transportation, and the number of traffic accidents in China reaches to millions every year, resulting in huge casualties and property losses, so it is of great significance to study the prevention of traffic accidents. In the environment where vehicle design and road construction have been difficult to make a breakthrough in accident prevention, more and more attention has been paid to the study of driver's driving behavior. However, there is still a lack of unified quantitative indicators to describe driving behavior habits, and the research results of evaluating driving safety level by quantitative indicators are few and far between. Therefore, based on OBD (On-Board Diagnostics: vehicle Diagnostic system) technology, the driving behavior habits are analyzed from five dimensions: driving stroke, operation type, driving speed, driving acceleration and engine speed. The quantitative index of driving behavior is extracted by using the vehicle running data collected for a long time and the driving behavior habit has the characteristics of long-term stability and invariance. The data in this study are derived from the driving data of ordinary drivers collected by "excellent driving Smart Box" for more than 2 years, and the main driving regions of these drivers are in Chongqing area. The "superior driving smart box" obtains the driving data of the vehicle through the OBD interface that the vehicle generally has, and then connects with the smartphone through Bluetooth, displays the important data on the smartphone and passes the driving operation data to the server. Enables each driver's vehicle operation data to be kept without interruption. At present, most of the cars in the market use the international standard of OBD II, which makes the "superior driving intelligent box" can obtain the data of different manufacturers and different models of vehicles in real time through the OBD interface. With the development of OBD technology, the number of vehicle running data is increasing. In this paper, the vehicle running data related to driving behavior are extracted from five dimensions, and 57 indexes are obtained by transformation. The time series stability analysis method in financial field is introduced to test the stability of these 57 indexes. A total of 17 indexes satisfying stability were obtained, which were used as quantitative indicators of driving behavior. On the basis of the extracted quantitative index of driving behavior, the driver classification research is also carried out. The results show that the driver with different driving behavior characteristics can be distinguished by using the driving behavior quantification index to classify driver. This study provides a theoretical basis for the quantitative study of driving behavior and a new way of thinking and method for the study of accident tendency. The analytical data obtained by the OBD method is easier to obtain than the traditional experimental data, and the results of the analysis are more realistic and practical.
【学位授予单位】：第三军医大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.25

【参考文献】