时空拓扑特性下浮动车数据的多维扩展修复方法研究

Floating car has been widely used because of its price advantage, but the repair methods are limited due to its few acquisition parameters. The subject intend to break through the disadvantages of simple use the mathematical methods for low-dimensional traffic data repair, introduce road network topology to data repair and make linear topology extended to the surface topology. The method try to mining the traffic data associated with geographic features by combine the traffic data with geography information, deeply analyze the periodic time-varying characteristics from both micro and macro views, and then the repairing models of multiple dimension such as week, day, time, space and so on will be constructed. So the modifying method will also be changed from linear correlation to facial correlation and multi-dimension correlation.The method try to parameterize the influence of multiple topologies to data correlation based on the traffic data space-time relativity. Analyze the delay and leading effect caused by the distance of the surrounding roads in order to better characterization of its multi-mode correlation characteristics, provide theoretical basis for the identifying and modifying of low dimension data. This subject focuses on the intersectant research area of traffic and geography, its aim is to probe solving the common problems of identifying and modifying the low dimension traffic and geographical data, provide theoretical basis for it and provide reference for data modifying in other fields as well.

浮动车因为采集方式的价格优势而得到广泛应用，但由于采集的数据参数单一，不能用交通流机理辨识错误数据，导致修复方法受到制约。课题拟突破单纯利用数学方法对低维度交通数据的修复弊端，探索挖掘交通数据与地理数据的多模式关联特征，利用地理信息理论和张量理论修复低维度交通数据。课题以交通地理数据的时空相关特性为研究对象，把路段线性拓扑扩展到面拓扑，参数化表达路网多次拓扑对数据相似性的影响，分析目标路段与周边路段的距离导致的数据迟滞及超前效应，更好的表征交通地理数据的关联特性；从宏观和微观角度深化扩展分析数据的周期时变特性，构建周、天、时刻、空间等多维度修复模型，将低维度数据的线性相关修复扩展为面相关性修复和多维相关性修复，为低维交通地理数据的辨识与修复提供理论依据。研究旨在探索解决低维度交通地理数据辨识和修复中的共性问题，为低维数据的修复提供理论支撑，同时为其它领域数据的修复提供借鉴和参考。

浮动车作为交通状态采集的重要工具得到了广泛使用，但浮动车数据只有单一的车速参数，没有占有率及流量等参数，因此不能按照交通流数据修复的方法对数据进行修复；目前对浮动车数据的处理多用线性插值或历史平均方法进行修复，修复精度低。本项目突破单纯利用数学方法对低维度交通数据的修复弊端，通过对低维度交通地理数据的周期时变特性及拓扑路网时空相关性的深入扩展研究，提出了一种构建低维度交通地理数据错误辨识和数据修复模型的新方法，揭示了路网多次拓扑下路段距离及拓扑层次所引起的数据迟滞及超前效应。本项目以低维度交通地理数据的时空特性作为研究对象，参数化表达了路网单次拓扑和多次拓扑特性下道路的空间相关性以及数据的时间迟滞效应与上下游拓扑路段的距离关系，阐明了目标路段拓扑范围的选择对修复结果的影响，为相关低维交通地理数据的修复提供了理论基础，研究成果可应用于提高浮动车数据的质量。