测绘学报 ›› 2018, Vol. 47 ›› Issue (3): 403-412.doi: 10.11947/j.AGCS.2018.20170373

• 地图学与地理信息 • 上一篇    下一篇

跨比例尺新旧居民地目标变化分析与决策树识别

陈利燕1,3, 张新长2,1, 林鸿3, 杨敏4   

  1. 1. 中山大学地理科学与规划学院, 广东 广州 510275;
    2. 广州大学地理科学学院, 广东 广州 510006;
    3. 广州市城市规划勘测设计研究院, 广东 广州 510060;
    4. 武汉大学资源与环境科学学院, 湖北 武汉 430072
  • 收稿日期:2017-06-30 修回日期:2017-11-16 出版日期:2018-03-20 发布日期:2018-03-29
  • 通讯作者: 张新长 E-mail:eeszxc@mail.sysu.edu.cn
  • 作者简介:陈利燕(1981-),女,博士后,高级工程师,研究方向空间数据更新与融合.E-mail:jimigao@163.com
  • 基金资助:
    广州市博士后启动基金(201517040001);国土资源部城市土地资源监测与仿真重点实验室开放基金(KF-2016-02-020);国家自然科学基金重点项目(41431178);广东省自然科学基金重点项目(2016A030311016);广州市工信委信息化项目(GZIT2016-A5-147)

Change Analysis and Decision Tree Based Detection Model for Residential Objects across Multiple Scales

CHEN Liyan1,3, ZHANG Xinchang2,1, LIN Hong3, YANG Min4   

  1. 1. School of Geography and Planning, Sun Yat-Sen University, Guangzhou 510275, China;
    2. School of Geographical Sciences of Guangzhou University, Guangzhou 510006, China;
    3. Guangzhou Urban Planning and Design Survey Research Institute, Guangzhou 510060, China;
    4. School of Resources and Environment Science, Wuhan University, Wuhan 430072, China
  • Received:2017-06-30 Revised:2017-11-16 Online:2018-03-20 Published:2018-03-29
  • Supported by:
    Guangzhou Postdoctoral Science Foundation (No. 201517040001);The Open Fund of Key Laboratory of Urban Land Resources Monitoring and Simulation,Ministry of Land and Resources (No. KF-2016-02-020);The National Natural Science Foundation of China (No. 41431178);The Natural Science Foundation of Guangdong Province of China (No. 2016A030311016);GZIT (No. 2016-A5-147)

摘要: 变化分析与探测是跨比例尺地图数据更新的核心问题之一。以往研究主要关注时间维上地理实体时空演化引起的地图目标变化,甚至将地图目标变化等同于地理实体真实变化,忽略了尺度维上由地图综合导致的表达变化。本文以居民地数据为例,从表层形式和深层缘由对跨比例尺新旧地图数据间的目标变化进行深入分析。在此基础上,引入机器学习领域的决策树方法构建变化信息识别模型。该模型的目标是判别时态变化和表达变化两种类型,从而提取用于更新小比例尺地图数据的真正变化信息。结合广州市多比例尺地图数据库更新任务及实际数据进行验证,结果显示设计的变化探测模型可以达到90%以上的整体精度。

关键词: 数据更新, 多比例尺, 变化识别, 决策树

Abstract: Change analysis and detection plays important role in the updating of multi-scale databases.When overlap an updated larger-scale dataset and a to-be-updated smaller-scale dataset,people usually focus on temporal changes caused by the evolution of spatial entities.Little attention is paid to the representation changes influenced by map generalization.Using polygonal building data as an example,this study examines the changes from different perspectives,such as the reasons for their occurrence,their performance format.Based on this knowledge,we employ decision tree in field of machine learning to establish a change detection model.The aim of the proposed model is to distinguish temporal changes that need to be applied as updates to the smaller-scale dataset from representation changes.The proposed method is validated through tests using real-world building data from Guangzhou city.The experimental results show the overall precision of change detection is more than 90%,which indicates our method is effective to identify changed objects.

Key words: spatial data updating, multiple scales, change detection, decision tree

中图分类号: