地球信息科学学报 ›› 2012, Vol. 14 ›› Issue (6): 719-727.doi: 10.3724/SP.J.1047.2012.00719

• 地球信息科学理论与方法 • 上一篇    下一篇

基于层次随机图的道路选取方法

李木梓, 徐柱, 李志林, 张红, 遆鹏   

  1. 西南交通大学地球科学与环境工程学院, 成都 610031
  • 收稿日期:2012-11-10 修回日期:2012-12-01 出版日期:2012-12-25 发布日期:2012-12-25
  • 通讯作者: 徐柱(1972-),男,博士,教授,博士生导师,研究方向为时空数据分析与挖掘、空间数据综合、空间数据共享等。E-mail:xuzhucn@gmail.com E-mail:xuzhucn@gmail.com
  • 作者简介:李木梓(1985-),女,博士,研究方向为空间数据库、空间数据综合。E-mail:green.sunny@163.com
  • 基金资助:

    国土资源公益性行业科研专项经费(201111013);国家自然科学基金项目(40971209);中央高校基本科研业务费专项(SWJTU11CX059,SWJTU11CX063,SWJTU10ZT02)资助。

A Hierarchical Random Graph Based Selection Method for Road Network Generalization

LI Muzi, XU Zhu, LI Zhilin, ZHANG Hong, TI Peng   

  1. Department of Remote Sensing and Geographic Information Engineering, Faculty of Geosciences and Environmental Engineering, Southwest Jiaotong University, Chengdu 610031, China
  • Received:2012-11-10 Revised:2012-12-01 Online:2012-12-25 Published:2012-12-25

摘要:

道路选取是道路网自动综合的关键问题之一,这方面已有多年研究,虽已取得很大进展,但尚不能自动完成一定比例尺下道路网的选取。本文提出一种基于层次随机图的道路选取方法,通过构建道路网的层次聚类结构以辅助道路选取。层次随机图是一种复杂网络模型,表现为一个二叉树,它不仅可以将复杂的道路网进行层次聚类,而且在可视化的同时提供了不同粒度的聚类信息。在构建道路网的层次随机图的基础上,本文采用累计权重数来衡量每条道路在整体层次结构中的重要性,并据此进行道路选择。我们将该方法应用到不同模式的实际道路网中进行道路选取试验,包括方格形、方格放射状、环形放射状、自由式路网等,以对应的谷歌地图作为参考进行道路选取符合数量、符合长度的定量评价和观察对比定性评价。试验表明本方法的选取结果与谷歌地图符合度很高。此外,与典型的基于路划长度和基于度中心度(degree centrality)的选取方法相比,本文方法更优。最后给出了本文方法优缺点的讨论和进一步研究的展望。

关键词: 道路网, 制图综合, 层次随机图, 复杂网络, 道路网络

Abstract:

Generalization of road network is one of the focuses in map generalization. Road network generalization can be considered as the combination of two processes. One is selective omission, and the other is the simplification of selected roads. Selective omission is the key process, in which it is hard to maintain the overall and key local structures of original networks. Many solutions have been proposed for road selective omission. But previous solutions cannot maintain these structures in the process of selective omission. It will solve the problem if we can build the hierarchical structure of road networks and make selection based on the structure. This paper presents a novel method for selective omission. The method first builds the hierarchical structure of road networks. It is based on Hierarchical Random Graph (HRG) which transforms a graph into a dendrogram, which is widely used in complex networks. HRG goes beyond simple clustering and provides clustering information at all levels of granularity for visualization. But HRG is over detailed for multi-scale representation as its dendrogram usually contains tens or even more layers. So, after building HRG of road networks, we propose a measure named Accumulated Probability Number (APN) to simply HRG hierarchy. APN reflects the importance of each road in the whole network. It should be noted that we use road ‘strokes' as vertices and the connections between them as edges when transforming a road network into a graph. The proposed approach is validated with case studies of road network generalization. Different patterns of road networks are considered including grid, ring-star-hybrid, grid-star-hybrid, irregular patterns. The corresponding Google Map is used as the reference for evaluation of road selection. The results of APN-based selection match well with the reference.

Key words: road networks, map generalization, complex networks, hierarchical random graph, road networks