几何和图论特征对高分辨率遥感影像土地利用分类的影响

基金项目: 国家科技支撑计划资助项目(2006BAJ09B01)；国家教育部博士点专项基金资助项目(20060486041)

详细信息

Influence of Geometric and Graph Theoretical Measures on Land Classification Using High-Resolution Remote Sensing Images

¹ School of Remote Sensing and Information Engineering,Wuhan University,129 Luoyu Road,Wuhan 430079,China

Funds: 国家科技支撑计划资助项目(2006BAJ09B01)；国家教育部博士点专项基金资助项目(20060486041)

摘要: 利用高空间分辨率影像中地物的几何特征和空间邻域信息,结合我国土地利用状况和空间分布的特点,在支持向量机分类方法的基础上,利用几何和图论特征实现高分辨率遥感影像土地利用状况分类,并分析讨论了几何和图论特征对分类结果的影响。实验结果表明,该方法可以有效地识别未利用地、农业用地、工业用地和商住用地4种典型地类,其分类结果也可为土地利用变化监测提供依据。
- 高分辨率遥感影像 /
- 土地利用 /
- 直线提取 /
- 图论特征
Abstract: We focus on geometric structure and neighborhood information in high spatial resolution images,combining with local situation and spatial distribution of land use,to accomplish land use classification using geometric and graph theoretical measures based on a SVM,and discuss how these measurements affect the classification results.The experiments use QuickBird images.The result shows that this method can classify the sampled images into four typical classes as rural,agricultural,industrial and commercial-residential regions.The final results can also be used for land use change detection and monitoring.
- high-resolution remote sensing images /
- land use /
- line extraction /
- graph theoretical measures