地球信息科学学报 ›› 2014, Vol. 16 ›› Issue (6): 997-1004.doi: 10.3724/SP.J.1047.2014.00997

• • 上一篇    

面向对象的南海珊瑚礁地貌单元提取

龚剑明1(), 朱国强2,3,*(), 杨娟3, 左秀玲3, 石伟3   

  1. 1. 中国科学院学部工作局,北京 100190
    2. 兰州交通大学测绘与地理信息学院,兰州 730070
    3. 中国科学院地理科学与资源研究所 资源与环境信息系统国家重点实验室,北京 100101
  • 收稿日期:2014-09-21 修回日期:2014-10-23 出版日期:2014-11-10 发布日期:2014-11-01
  • 通讯作者: 朱国强 E-mail:jmgong@126.com;zhugq@lreis.ac.cn
  • 作者简介:

    作者简介:龚剑明(1975-),博士,湖北黄冈人,研究方向为遥感与地理信息系统应用。E-mail:jmgong@126.com

  • 基金资助:
    国家“863”重大项目课题(2012AA12A406)

A Study on the Object-oriented Model for Geomorphic Unit Extraction of Coral Reefs in the South China Sea

GONG Jianming1(), ZHU Guoqiang2,3,*(), YANG Juan3, ZUO Xiuling3, SHI Wei3   

  1. 1. Bureau of the Academic Divisions, Chinese Academy of Sciences 100190, China
    2. Faculty of Geoinfomatics, Lanzhou Jiaotong University, Lanzhou 730070, China
    3. State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
  • Received:2014-09-21 Revised:2014-10-23 Online:2014-11-10 Published:2014-11-01
  • Contact: ZHU Guoqiang E-mail:jmgong@126.com;zhugq@lreis.ac.cn
  • About author:

    *The author: CHEN Nan, E-mail:fjcn99@163.com

摘要:

南海珊瑚礁地貌信息的提取是珊瑚礁资源利用、生态保护与管理及可持续发展的重要地学基础。本文提出了面向对象的珊瑚礁地貌单元提取模型,针对不同的地貌单元,以不同地貌单元的最优分割尺度、光谱参数、形状参数来分割影像并合并成不同对象,从而获得相应的地貌单元。通过大量实验得出自然地貌的最优分割尺度区间为[140,600],其中附礁生物稀疏带及丛生带、礁坑发育带的光谱参数和形状参数分别为0.9和0.1,其他自然地貌单元的光谱参数和形状参数分别为0.8和0.2;人工地貌的最优分割尺度区间为[25,170],其光谱参数和形状参数分别为0.8和0.2。进一步以南沙群岛簸箕礁WorldView-2高分辨率遥感影像为例提取地貌单元,并结合混淆矩阵和Kappa系数对分类结果进行了精度评价,地貌单元提取总体精度达到了85.75%,Kappa系数为0.8349。结果表明,该方法可有效运用南海珊瑚礁遥感影像的光谱特征、纹理特征,以及影像数据不同波段的组合特性,综合了影像和珊瑚礁地貌的关联特性,充分利用了珊瑚礁不同地貌相带的异质性,获得了理想的南海珊瑚礁地貌的整体信息,满足了我国南海珊瑚礁地貌信息提取和地貌数字产品生成的需求。

关键词: 南海, 珊瑚礁, 地貌, WorldView-2遥感影像, 面向对象

Abstract:

Geomorphology of coral reefs is the foundation for understanding the reef-building process and mechanism. It is worthwhile to acquire the geomorphological information of coral reefs in the South China Sea because this information can provide valuable insights for the resources utilization, ecological protection and management as well as sustainable development of coral reefs in China. Based on the development of current remote sensing techniques, an object-oriented geomorphic unit extraction method for coral reefs is proposed in this paper. The satellite images are segmented by the optimal scales, spectral parameters and shape parameters and the segmented objects are combined to different objects to obtain the geomorphic units. Through a large number of experiments, the optimal scales for natural geomorphic units range from 140 to 600. Spectral parameters and shape parameters for sparse reef associations zone, numerous reef associations zone, and reef pits development zone are all 0.9 and 0.1, respectively. Spectral parameters and shape parameters for other natural geomorphic units are 0.8 and 0.2, respectively. On the other hand, optimal scales for man-made geomorphic units range from 25 to 170, while their spectral parameters and shape parameters are all 0.8 and 0.2, respectively. The WorldView-2 high resolution remote sensing image of Boji Reef in Nansha Islands is used as experimental data in this paper and the accuracy of classification results is assessed with the confusion matrix and Kappa coefficient, the overall accuracy of geomorphic unit extraction reaches 85.75% and the Kappa coefficient is 0.8349. The results show that this method can effectively take advantage of the features of spectra, texture and the stacking of different wavebands of satellite images. In addition, this method can grasp the correlation of satellite image and reef geomorphic zonation and make full use of the heterogeneity of reef geomorphology. Therefore, an ideal geomorphic zonation that has a higher accuracy can be obtained to acquire geomorphological information and produce geomorphological digital products of coral reefs in the South China Sea.

Key words: The South China Sea, coral reef, geomorphology, World View-2 remote sensing, object oriented