自然资源学报 ›› 2015, Vol. 30 ›› Issue (2): 350-360.doi: 10.11849/zrzyxb.2015.02.017

• 资源研究方法 • 上一篇    

基于高分1号杭州湾南岸滨海陆地土地覆盖信息提取方法研究

程乾, 陈金凤   

  1. 浙江工商大学区域生态环境与空间信息技术研究所, 杭州310018
  • 收稿日期:2014-02-22 修回日期:2014-07-19 出版日期:2015-02-20 发布日期:2015-02-10
  • 作者简介:程乾(1968-),男,河北易县人,教授,博士,主要从事水环境遥感方面的工作。E-mail:qiancheng525@163.com
  • 基金资助:
    国家高分辨率对地观测重大专项(E05-Y30B02-9001-13/15-4);国家自然科学基金项目(41271417)。

Research on the Extraction Method of Landcover Information in Southern Coastal Land of Hangzhou Bay Based on GF-1 Image

CHENG Qian, CHEN Jin-feng   

  1. Institute of Regional Eco-environment and Spatial Information Technology, Zhejiang Gongshang University, Hangzhou 310018, China
  • Received:2014-02-22 Revised:2014-07-19 Online:2015-02-20 Published:2015-02-10
  • About author:10.11849/zrzyxb.2015.02.017

摘要: 针对滨海陆地复杂地物环境下如何提高土地覆盖信息遥感提取精度的关键问题,基于最新发射的国产高分(高分辨率)1 号和资源3 号影像,采用面向对象分类法和最大似然法,开展杭州湾南岸滨海陆地土地覆盖信息的遥感提取。结果表明,相比最大似然法,面向对象法结合高分1 号既考虑对象的光谱、空间和纹理等多种属性特征,又充分利用高分影像所提供的丰富纹理和空间信息,对于土地类型多样、边界模糊等混合像元具有较好的识别能力,获得较高分类总精度(90.4%)和Kappa 系数(0.876 7);分割尺度对高分影像分类精度具有重要影响,高分1 号2 m和8 m最优分割尺度分别为63%和65%,资源3 号影像最优分割尺度为66%;高分1号比资源3 号更能体现在植被和水体等地物信息提取方面的优势。

Abstract: Under the complex coastal land environment, how to improve the extraction accuracy of land cover information by remote sensing is a key problem. This paper, taking southern coastal land of Hangzhou Bay as the study area, using most newly launched GF-1 satellite and Resources satellite No.3 remote sensing images extract the land cover information by the object-oriented classification method compared with maximum likelihood method. The results show that comprising with the maximum likelihood method, the object oriented method with GF-1 image are more suitable for extraction of coastal land information. It not only considers the object spectral, spatial and texture features, but also makes full use of the rich texture and spatial information of GF-1 image, which has better recognition ability for various land types, distribution of fuzzy coastal land boundary of mixed pixels, thus getting higher classification accuracy of 90.4% and Kappa coefficient of 0.8767. The choice of segmentation scale is an important influence on high precision image classification, and the results show that image segmentation optimal scales of GF-1 2 m and 8 m images are 63% and 65%, and the optimal segmentation scale of Resources No.3 images is 66%. Comprising with resources satellite No.3, GF-1 image can reflect more advantages in land cover information extraction in the aspects of ground vegetation and water.

中图分类号: 

  • P237