基于低频出租车GPS轨迹数据的路网信息提取

发布时间：2018-01-04 12:38

本文关键词：基于低频出租车GPS轨迹数据的路网信息提取　出处：《武汉大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着全国各地城市经济蓬勃发展,我国的城市路网结构正处于快速变化之中,因而数字道路信息的快速、准确获得与及时更新变得尤为重要。诸如智能交通系统、城市规划系统等重要地理信息应用都依赖于准确及时的道路网络信息。目前,数字道路信息通常由专业设备采集制作,不仅费用昂贵,而且难以及时更新。运用遥感影像提取道路的方法发展较快,但由于遥感影像中混合了多种地物,道路提取容易受到干扰,提取难度较大,且同样存在成本高、信息滞后的问题。汽车的GPS轨迹能够反映道路网络的位置信息和几何结构,因此可以依据日常车辆的轨迹数据来获取道路信息,这种方法具有成本低廉、覆盖面大、数据易于获取、现势性强等优点。本文聚焦低频率低精度的出租车GPS轨迹数据,设计了一种路网信息提取的框架。本文依次提出了基于低频出租车GPS轨迹数据的道路交叉口、道路中心线、道路单双向信息及交叉口转向关系信息提取方法。首先,栅格化车辆轨迹数据得到轨迹图像,使用数学形态学滤波的方式处理该图像填补空洞、去除噪声、平滑表面,最后利用数学形态学细化的方法获取近似道路中心线和道路交叉口。为尽可能准确、全面的提取道路交叉口信息,在轨迹数据栅格化阶段生成多种分辨率图像,分别处理获取道路交叉口,并将提取结果融合,以达到提取采样数据稀疏区域道路交叉口的目的。其次,提出了两种道路中心线获取的方法:(1)使用Douglas-Peukcer算法化简由数学形态学细化获取的初步道路中心线,使其与真实道路形状更加接近;(2)以分布在相邻道路交叉口之间的轨迹点为数据源,利用多元自适应回归样条拟合路段。接着,提出Intersection-Link模型,作为多元自适应回归样条的数据源,并据此模型提取道路单双向信息和道路交叉口转向关系。最后,整合以上技术和方法,设计了一种面向大量低频低精度出租车GPS轨迹数据的道路网络信息提取框架。并以武汉市10天内出租车GPS轨迹数据作为数据源进行了实验,选取了武汉市两处核心区域,验证了本文方法的有效性。
[Abstract]:With the vigorous development of city economy throughout the country, city road network structure of our country is undergoing rapid change, so the digital road information fast, accurate and timely update has become particularly important. Such as intelligent traffic system, the information system of city planning and other important geographic use depends on road network information timely and accurately. At present, the digital road the information is usually produced by professional equipment acquisition, not only expensive, but also difficult to update. Using the method of remote sensing image road extraction has developed rapidly, but due to a variety of remote sensing images in the mixed material, road extraction is susceptible to interference, extraction difficult, and also has high cost, information lag can reflect the position information and problem. The geometric structure of the road network GPS track car, so you can according to the daily vehicle trajectory data to obtain road information, this method Has the advantage of low cost, large coverage, easy access to data, has the advantages of high practicality. The taxi GPS trajectory data focus on low frequency and low precision, designed a network information extraction framework. This paper puts forward the low-frequency taxi GPS trajectory data of Road intersection based on the central line of the road, Road and cross single two-way information export to relation information extraction method. Firstly, the grid data from vehicle trajectory trajectory images, using mathematical morphology methods to process the image to fill the holes, remove noise, smooth surface, finally uses mathematical morphology thinning method to obtain the approximate centre line of the road and road intersection. As accurately as possible, a comprehensive cross road extraction export information in the track data grid phase to generate multiple resolution image processing, respectively get the road intersection, and the extraction results of fusion, in order to achieve extraction Sample data sparse regions of the road intersection. Secondly, we propose two methods to obtain the central line of the road: (1) preliminary road center line using Douglas-Peukcer algorithm to simplify the mathematical morphology to get, to make it more close to the real road shape; (2) the distribution between adjacent road intersection track points as the data source, the regression spline fitting section using multivariate adaptive. Then, proposed Intersection-Link model, multivariate adaptive regression splines as data source, and based on the model of road extraction from single two-way information and road intersection to relationship. Finally, the integration of the above techniques and methods, design a number of low frequency and low precision for taxi GPS trajectory data of road network information extraction framework. Taking Wuhan city 10 days taxi GPS trajectory data as the data source selection experiments were carried out in Wuhan city at the core region, The validity of this method is verified.

【学位授予单位】：武汉大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U495;P228.4

【相似文献】