城市交通轨迹大数据的语义分析和可视化

发布时间：2017-12-27 10:03

本文关键词：城市交通轨迹大数据的语义分析和可视化　出处：《浙江大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,随着"数字城市","智慧城市"等概念的兴起,基于GPS定位采样的交通轨迹数据研究正成为越来越热门的课题。交通轨迹数据具有很强的时空属性,同时兼有大数据的特征。对轨迹数据的高效组织和合理展示是城市交通研究课题中的一大难点。本文开发了一套架构于Hadoop生态之上的分布式文件系统,用来存储海量交通数据。并使用Hive数据仓库管理海量交通轨迹数据。针对交通轨迹数据规模大的特点,本文采用了视图,索引等技术,优化数据库检索语句,降低检索开销。同时,按照"天"和"小时"这两个时间节点对交通轨迹数据分区,优化存储结构,减少全盘扫描操作的频率。为了减小海量数据的存储规模,本文还尝试了基于轨迹数据的压缩技术。此外,本文另辟蹊径,提出了一种基于语义分析处理轨迹数据信息的方法,通过对城市热点区块间联系的研究,分析交通轨迹数据背后隐含的语义信息。该方法基于词向量模型,由深度学习框架TensorFlow编程实现。在可视化环节,通过研究城市热点区块的交通流量情况,分析整个城市的交通模式,并根据车辆细轨迹数据信息,对城市交通管理提出一些建议。最后,根据本文研究内容,实现了 B/S架构的交通轨迹语义分析和可视化系统原型。该原型系统集成了从轨迹数据的采集、预处理、轨迹数据分析、可视化展示等一系列操作的自动化批处理流程。
[Abstract]:In recent years, with the emergence of "digital city", "intelligent city" and other concepts, the research of traffic trajectory data based on GPS positioning sampling is becoming more and more popular. The traffic trajectory data has a strong spatial and temporal attribute, and also has the characteristics of large data. The efficient organization and rational display of track data is a difficult problem in urban traffic research. In this paper, a distributed file system based on Hadoop ecology is developed to store mass traffic data. And use Hive data warehouse to manage mass traffic trajectory data. In view of the large scale of traffic trajectory data, this paper adopts the technology of view and index to optimize the database retrieval sentence and reduce the retrieval cost. At the same time, according to the two time nodes of "day" and "hour", the storage structure is optimized to reduce the frequency of full scanning operation. In order to reduce the storage size of massive data, this paper also attempts a compression technology based on trajectory data. In addition, this paper proposes a new way to process trajectory data based on semantic analysis. By analyzing the links between urban hot spots, we can analyze the implicit semantic information behind traffic trajectory data. The method is based on the word vector model and is programmed by the depth learning framework (TensorFlow). In the visualization link, through studying the traffic flow situation of the urban hot spots, the traffic mode of the whole city is analyzed, and some suggestions for urban traffic management are put forward according to the information of vehicle fine track. Finally, according to the research content of this paper, the traffic trajectory semantic analysis and visual system prototype of B/S architecture are realized. The prototype system integrates a series of automated batch processing processes, such as track data collection, preprocessing, trajectory data analysis, visual display and a series of operations.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U12;TP311.13

【参考文献】