基于影像特征多尺度拓扑的城市土地覆盖信息提取  

Urban Land Cover Extraction Based on Multi-scale Topology

作　　者：王俊海 阮仁宗[1] 林鹏 许玲丽 罗宁 WANG Jun-hai;RUAN Ren-zong;LIN Peng;XU Ling-li;LUO Ning(School of Earth Sciences and Engineering,Hohai University,Nanjing 210098;Northern Information Control Research Institute Group Corporation Limited,Nanjing 211153,China)

机构地区：[1]河海大学地球科学与工程学院,江苏南京210098 [2]中国兵器北方信息控制研究院集团有限公司,江苏南京211153

出　　处：《地理与地理信息科学》2018年第3期74-79,共6页Geography and Geo-Information Science

基　　金：江苏省研究生科研与实践创新计划项目(KYCX17_0505);中央高校基本科研业务费(学生项目)(2017B669X14);中国科学院战略性先导科技专项(XDA05050106)

摘　　要：城市的土地覆盖信息对于城市的合理规划与建设、城市的生态评估等具有重大的参考价值,因此,如何快速且准确地提取城市土地覆盖信息是当前亟须解决的问题之一。该文选取南京市作为研究区,利用QuickBird影像进行城市土地覆盖信息提取研究。首先对影像进行多尺度分割,在典型样本分析的基础上创建多尺度的影像分类特征集;然后利用模拟退火算法优化分类特征集并进行影像的初始分类;在此基础上,构建初始分类对象之间的拓扑关系,在同尺度上进行拓扑补充、在不同尺度上进行空间叠加,从而对初始分类结果进行再分类。最终分类结果的总体精度达87.651%,Kappa系数为0.8587,比初始分类结果的精度提高了近9%。研究结果表明,基于多尺度拓扑的分类方法可有效提高遥感影像的分类精度。Urban land cover information plays an important role in urban planning and construction and the ecological assessment of cities.How to obtain urban land cover information quickly and accurately is a challenging problem.In this paper,a core area of Nanjing City was taken as the study area.Firstly,the QuickBird remote sensing image was segmented at multiple scales and an appropriate scale set was established.And based on the analysis of typical samples,the image classification feature sets were created at different scales.Then,the simulated annealing algorithm was used to optimize the classification feature set and the initial classification of images at different scales was conducted.Finally,objects at different scales were overlaid and topological relationship between objects was used for further improvement of classification results.Based on topological relationship,objects were further reclassified and classification accuracy was improved.In this paper,the total classification accuracy reaches 87.651%,Kappa coefficient is 0.8587,compared with the accuracy of the initial classification results,the precision was improved by nearly 9%.It is found that the classification accuracy of high spatial resolution remote sensing images can be improved effectively based on the multi-scale object topology.

关 键 词：面向对象影像分析 多尺度拓扑 QUICKBIRD影像 南京市 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置] P237[自动化与计算机技术—控制科学与工程]