测绘通报 ›› 2018, Vol. 0 ›› Issue (1): 138-142.doi: 10.13474/j.cnki.11-2246.2018.0027

• 行业观察 • 上一篇    下一篇

GF-2影像面向对象典型城区地物提取方法

王蕾1, 杨武年1, 任金铜1,2, 邓晓宇1   

  1. 1. 成都理工大学国土资源部地学空间信息技术重点实验室, 四川 成都 610059;
    2. 贵州工程应用技术学院贵州省教育厅生物资源开发与生态修复特色重点实验室, 贵州 毕节 551700
  • 收稿日期:2017-04-20 出版日期:2018-01-25 发布日期:2018-02-05
  • 通讯作者: 杨武年。E-mail:ywn@cdut.edu.cn E-mail:ywn@cdut.edu.cn
  • 作者简介:王蕾(1993-),女,硕士生,研究方向为资源与环境遥感。E-mail:wangleitl@qq.com
  • 基金资助:

    国家自然科学基金(41372340;41671432);四川省国土资源厅应用基础研究项目(KJ-2016-12);四川省教育厅科研项目重点项目(172A0027);贵州省教育厅自然科学研究项目(黔教合KY字(2015)448号)

Object-oriented Extraction Method of Typical Urban Features Based on GF-2 Images

WANG Lei1, YANG Wunian1, REN Jintong1,2, DENG Xiaoyu1   

  1. 1. Key Laboratory of Geoscience Spatial Information Technology of Ministry of Land and Resources, Chengdu University of Technology, Chengdu 610059, China;
    2. Key Laboratory of Bioresource Development and Ecological Restoration of Guizhou Provincial Department of Education, Guizhou University of Engineering Science, Bijie 551700, China
  • Received:2017-04-20 Online:2018-01-25 Published:2018-02-05

摘要:

国产高分遥感影像信息丰富,提供了精准的地物空间细节,深入研究高分数据处理及其提取城区地类目标信息的方法具有重要意义。本文以国产高分二号(GF-2)遥感影像为数据源,利用规则集的面向对象分类方法,通过ESP尺度分析工具选取得出最优分割尺度,建立各类地物的特征体系及分类规则,最终提取出研究区典型城区地物信息,并将之与传统基于像元的SVM监督分类结果作比较。结果表明:规则集的面向对象分类总体精度为92.23%,Kappa系数为0.9,比SVM监督分类有大幅度提高。对高分二号等高分辨率影像,面向对象的分类方法精度更高,图示效果更好,是城区地物提取的有效方法。

关键词: 高分二号, 面向对象, 多尺度分割, 典型城区地物

Abstract:

Domestic Gaofen remote sensing images have abundant information,which can provide accurate details of spatial objects.It is of great significance to study Gaofen data processing and the method of extracting urban objects in depth.Taking the remote sensing image of domestic GF-2 as the data source,the optimal segmentation scale is obtained through the ESP scale analysis tool which is based on the object-oriented classification method of rule set.And then the feature system and the classification rules of various features are established to extract the information of typical urban features. The results are compared with the traditional pixel-based SVM supervised classification.The results show that the overall accuracy of object-oriented classification of rule set is 92.23%,and the Kappa coefficient is 0.9,which is significantly improved compared with SVM supervised classification.For high-resolution images such as GF-2,the object-oriented classification method is more accurate and has a better graphical effect,which is an effective method for urban object extraction.

Key words: GF-2, object-oriented, multi-scale segmentation, typical urban feature

中图分类号: