自然资源学报 ›› 2014, Vol. 29 ›› Issue (9): 1589-1597.doi: 10.11849/zrzyxb.2014.09.013

• 资源研究方法 • 上一篇    下一篇

基于面向对象的海岸带土地利用信息提取研究

王彩艳1,2, 王瑷玲1,2, 王介勇3, 王瑞燕1,2, 姜峥嵘4   

  1. 1. 山东农业大学资源与环境学院, 山东泰安271018;
    2. 土肥资源高效利用国家工程实验室, 山东泰安271018;
    3. 中国科学院地理科学与资源研究所, 北京100101;
    4. 乳山市国土资源局, 山东乳山264500
  • 收稿日期:2013-06-02 修回日期:2014-01-01 出版日期:2014-09-20 发布日期:2014-09-20
  • 通讯作者: 王瑷玲(1972-),女,山东临沂人,博士,教授,主要从事土地可持续利用和评价研究。E-mail:ailingwangdf@163.com E-mail:ailingwangdf@163.com
  • 作者简介:王彩艳(1988-),女,山东乳山人,硕士研究生,主要从事土地规划与利用研究。E-mail:jhckenan888@163.com
  • 基金资助:

    国家自然科学基金项目(41001109);山东省自然科学基金(ZR2013DM006);山东省高等学校科技计划项目(J11LC11)。

Coastal Zone Land Use Information Extraction Based on Objectoriented Classification Method

WANG Cai-yan1,2, WANG Ai-ling1,2, WANG Jie-yong3, WANG Rui-yan1,2, JIANG Zheng-rong4   

  1. 1. College of Resources and Environment, Shandong Agricultural University, Taian 271018, China;
    2. National Engineering Laboratory for Efficient Utilization of Soil and Fertilizer Resources, Shandong Agricultural University, Taian 271018, China;
    3. Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;
    4. Bureau of Land and Resource of Rushan City, Shandong Province, Rushan 264500, China
  • Received:2013-06-02 Revised:2014-01-01 Online:2014-09-20 Published:2014-09-20

摘要:

针对海岸带复杂多样的土地利用信息,选取山东省乳山市海岸带为研究区域,运用 Landsat TM遥感影像数据,基于面向对象分类方法,利用不同地物的光谱、形状、纹理和空间关系等特征,通过多尺度分割、隶属度函数法和标准最邻近分类法提取研究区土地利用信息,并对分类结果进行精度评估。结果表明:①隶属度函数法和标准最邻近分类法结合,提取出乳山市海岸带12 种土地利用类型信息,很好地区分了盐田和养殖水面、林地和园地,可提取出主要的道路和河流等细长线状地物;②将提取结果与最大似然法对比,面向对象分类方法提取精度达到82.50%,Kappa系数为0.809 1,分别比最大似然法提高了11.44%和0.105 5,很好地避免“同物异谱”和“异物同谱”对分类精度造成的影响,有效地避免了“椒盐”现象。面向对象分类方法提取中分辨率遥感影像精度较高,为海岸带土地利用信息的快速、准确提取提供了有效的技术手段。

关键词: 海岸带, 面向对象分类, 乳山市, 土地利用信息, Landsat TM

Abstract:

Since land use patterns in coastal zones are of ten characterized by complexity and diversity,it is difficult to acquire accurate land-use information from remote sensing image usingtraditional methods. Thus, this paper, taking the Rushan coastal zone, located in the east of Shandong Peninsula as a case, tries to fill this gap and develop a new method based on objectoriented classification to extract more accurate coastal zone land-use information. We collect Landsat TM image as the basic information sources for further data processing and analysis.Based on the method of object-oriented classification, this paper makes use of the characteristicsof different objects, such as spectrum, shape, texture and spatial relationships, and extractsl and-use information in the study area through multiresolution segmentation, Membership Functionand Standard Nearest Neighbor, and then evaluates classification accuracy. The results include:1) Twelve kinds of land-use information are classified in the Rushan coastal zone byboth Membership Function and Standard Nearest Neighbor methods. The salt and aquaculturewater, the woodland and orchard are well identified, and the elongated linear features wereclearly extracted. 2) Through comparing the extraction results using maximum likelihood method,the object-oriented classification accuracy was 82.50%, Kappa was 0.8091, 11.44% and 0.1055 higher than that using the maximum likelihood method. The results showed that the object-oriented classification method is an effective technical method for extraction of land-use informationquickly and accurately in coastal zones, because it can improve the classification precision of high resolution remote sensing image through attenuating effectively the effects onthe classification accuracy due to "same object with different spectrums" and "different objectswith same spectrums", and "salt-pepper".

Key words: Rushan County, land-use information, Landsat TM, coastal zone, object-oriented classification

中图分类号: 

  • TP751