摘要
车载监控系统各项功能及技术指标趋于成熟,实车试验因而得以顺利开展。与此同时,收集了丰富的车载监控数据。车载监控数据的类型与数量已经远达到了可以分析的技术要求,但是编码分析手段却并未随着数据量的增加而有所突破。为了有效利用车载监控数据,突破现有数据处理的瓶颈,便于后续挖掘工作的展开,提出了一种车载数据时空语义编码及分析方法。将时间序列符号化思想运用到交通工程中,充分考虑到驾驶数据特征,基于符号化聚合近似(SAX)的3个步骤,对选定的一段范例数据依次进行了正规处理、降维处理、离散及符号化处理。结果表明:经过语义编码后,先前维数很高、数据特征冗杂的驾驶时间序列数据合理地转换成了可读性强并且易于搜索定位的符号化序列,在实现大幅降低数据维度的同时又适时地保留了时间序列数据的主要特征。最后,通过案例分析演示了该方法在实际车辆驾驶安全性分析中的作用与优势。研究结果可为重点监控车辆高风险驾驶事件以及有针对性地开展驾驶安全培训等提供理论依据,同时也可为未来特定驾驶场景的快速提取进行技术储备。
Due to maturity of various functions and technical indexes related to automobile-mounted monitoring system, vehicle field test can be thus conducted smoothly. Meanwhile, abundant on-board monitoring data are collected. Although both the type and the quantity of such data have sufficiently satisfied technical requirement for analysis, no coding analysis approach breakthrough has been made with the increase in data size. To make effective use of on-board monitoring data, break through the current bottleneck of data processing, and facilitate follow-on data mining, we presented an on-board data space-time semantic encoding and analysis method. We applied the idea of time series symbolization to traffic engineering, and processed a piece of given sample data by normalization, dimensionality reduction, discretization and symbolization based on the 3 steps of SAX with consideration of driving data characteristics adequately. The result shows that the high-dimensionality miscellaneous driving time series data are rationally converted into highly readable, easy to search and locate symbolic series after semantic encoding, and the main characteristics of time series data are preserved after a substantial reduction of data dimensionality. Finally, we demonstrated the positive effects of this method on the analysis of actual vehicle driving safety based on case study. The analysis result provides theoretical foundations for key surveillance over high-risk vehicle driving events and targeted driving safety training, etc. Besides, this can be also seen as a technical reserve for rapid extraction of particular driving scenarios in the future.
作者
孙川
吴超仲
褚端峰
黄子超
李必军
SUN Chuan;WU Chao-zhong;CHU Duan-feng;HUANG Zi-chao;LI Bi-jun(School of Electromechanical and Automobile Engineering,Huanggang Normal University,Huanggang Hubei 438000,China;State Key Laboratory of Information Engineering in Surveying,Mapping and Remote Sensing,Wuhan University,Wuhan Hubei 430079,China;ITS Research Center,Wuhan University of Technology,Wuhan Hubei 430063,China;Research Institute of Highway,Ministry of Transport,Beijing 100088,China)
出处
《公路交通科技》
CAS
CSCD
北大核心
2019年第8期124-132,共9页
Journal of Highway and Transportation Research and Development
基金
湖北省教育厅科学研究计划项目(Q20182905)
大学生创新创业训练计划项目(20170514011)
关键词
交通工程
时空语义编码
数据挖掘
符号化聚合近似
车载数据
时间序列
traffic engineering
space-time semantic coding
data mining
symbolic aggregate approximation (SAX)
on-board data
time series
