地球信息科学学报 ›› 2013, Vol. 15 ›› Issue (5): 625-634.doi: 10.3724/SP.J.1047.2013.00625

• 地球信息科学理论与方法 •    下一篇

面向网页文本的地理要素变化检测

王曙1, 吉雷静2, 张雪英2, 赵仁亮3, 陈晓丹2, 余浩4   

  1. 1. 英国利兹大学地理学院, 利兹 LS2 9JT;
    2. 南京师范大学虚拟地理环境教育部重点实验室, 南京 210046;
    3. 中国国家基础地理信息中心, 北京 100830;
    4. 南京邮电大学计算机学院, 南京 210003
  • 收稿日期:2013-06-07 修回日期:2013-07-05 出版日期:2013-09-29 发布日期:2013-09-29
  • 通讯作者: 张雪英(1970-),女,四川人,博士,教授,主要从事地理信息系统、地理信息智能处理和服务等方面研究。E-mail:zhangsnowy@163.com E-mail:zhangsnowy@163.com
  • 作者简介:王曙(1989-),男,山西人,硕士生,主要从事地理信息智能数据处理、地理信息系统等方面研究。E-mail:shuwang8951@hotmail.com
  • 基金资助:

    国家测绘科技项目“网络地理信息变化检测技术研究”;国家自然科学基金项目(40971231);“863”计划项目(2007AA12Z221)

Change Detection of Geographic Features Based on Web Pages

WANG Shu1, JI Leijing2, ZHANG Xueying2, ZHAO Renliang3, CHEN Xiaodan2, YU Hao4   

  1. 1. School of Geography, The University of Leeds, LS2 9JT, United Kingdom;
    2. Key Laboratory of Virtual Geographical Environment, Ministry of Education, Nanjing Normal University, Nanjing 210046, China;
    3. National Geomatics Center of China, Beijing 100830, China;
    4. School of Computer Science, Nanjing University of Posts and Telecommunication, Nanjing 210003, China
  • Received:2013-06-07 Revised:2013-07-05 Online:2013-09-29 Published:2013-09-29

摘要:

地理要素变化检测已成为国家地理信息“十二五”规划和全国地理国情普查的重要组成部分。网页文本中蕴含海量的地理要素信息,尤其是新闻、政府、社交平台等网站的网页文本更新频繁,可为地理要素变化检测提供现势性的数据源。本文针对网页文本中地理要素变化的语言描述特点,构建了表达地理要素变化的语义知识库,设计了搜索引擎和通用主题相结合的网页爬虫,实现了相关网页文本的高效获取;采用规则模型和条件随机场模型,分别进行网页文本中地理要素变化信息抽取,包括地理要素名称、位置(地名)、时间和属性等。实验结果显示,本文设计的网页爬虫具有较高的相关网页文本获取能力,地理要素变化信息抽取的准确率能够达到70%以上,但是,语义知识库的完备程度对于信息抽取性能具有较大影响。研究成果表明,以网页文本为数据源的地理要素变化信息获取方法,能提供一种快速检测地理要素变化的新途径,与实地调绘和遥感影像检测等方法结合应用具有较好的优势互补性,可作为有力的辅助手段解决地理要素的持续更新和实时更新问题。

关键词: 信息抽取, 文本解析, 网页文本, 网页爬虫, 地理要素变化

Abstract:

Geographic features change detection has became a vital component of the national geographical information 12th Five-Year-Plan and the national geographic general survey. In web pages, billions of geographic feature changes were contained, especially in government official websites, news homepages, social portals and etc. The web pages of these websites update frequently, which could provide the latest data for geographic infor-mation change detection. Considering the complex characteristics of the web geographic information description, this paper did some valuable achievements. First of all, the geographic information knowledge base was established by summarizing the geographic information words and phrases, which could give the great supports to geographic information semantics change detection. Then, the web geographic information was obtained using two kinds of web crawler technologies. Combining the Google Custom Search crawler and general topic crawler, the web geographic information obtainment could be more complete in both scope and depth. Thirdly, the geographic information was parsed and extracted from the web text, which showed users the related features, place names, times and attributes. Last but not least, the prototype system was finally developed and the results were analyzed. The experiments indicated that the accuracy of related web pages obtainment and features change detection were over 74% and 70% respectively. In addition, the results of geographic information change detection highly relied on the integrity of knowledge base, which need to be completed further. Moreover, the uncertainty and fuzziness of web geographic information also limited the change detection results. Therefore, the web page based geographic information change detection could be a supplementary method of geographic information change detection. Combining the traditional surveying detection and remote-sensing imagery detection methods, it could solve the problems of continuous updating and timely updating of geographic information efficiently.

Key words: web crawler, text parsing, geographic feature changes, web text, information extraction