影像融合和面向对象技术在湿地分类中的应用  

Application of Image Fusion and Object-oriented Technology in the Wetland Classification

作　　者：王乐[1,2] 王超[1,2] 

机构地区：[1]首都师范大学城市环境过程与数字模拟国家重点实验室培育基地,北京100048 [2]首都师范大学资源环境与地理信息系统北京市重点实验室,北京100048

出　　处：《地理空间信息》2017年第8期22-26,29,共6页Geospatial Information

摘　　要：利用SPOT全色波段影像和Landsat ETM+多光谱影像,采用波段特征分析法和改进的最佳波段指数法确定最佳融合波段。选择具有代表性的HSV变换、Brovey变换、PCA变换、Gram-schmidt变换和小波变换方法进行影像融合。针对融合后的影像采用定量评价法进行质量评价,通过不同地物的光谱、空间和纹理信息等进行比较分析构建适当的分类特征和规则,采用面向对象的遥感分类方法进行分类。结果表明,波段特征分析法和改进的最佳波段指数法结合,可以获得最佳融合波段。各融合方法均有效提升了影像效果,其中HSV和GS变换融合方法更好地保持了影像的多光谱和高分辨率特性,融合后各地物特征分类明显,可以有效应用于湿地分类;采用主成分分析法来设置面向对象分类中的波段权重,可以利用各波段信息量的差异进行影像的分割。基于各种遥感指数的面向对象分类方法用于湿地分类获得了93.62%的分类精度,与传统的分类方法相比有了很大进步,在湿地分类中具有很大的应用潜力。Based on SPOT5 panchromatic images and Landsat ETM+multi-spectral images,this paper used the band characteristics analysis method and improved optimum index factor method to determine the best fusion band,and selected five kinds of image fusion methods including HSV transform,Brovey transform,principal component analysis transform,Gram-schmidt transform and wavelet transform to fuse images.The paper used the quantitative evaluation method to evaluate the fused image quality,and then built the classification characters and rules through different object's spectral,spatial and texture information.Finally,this paper used object-oriented method to classify remote sensing images.The results show that the band characteristic analysis method and improved optimum index factor are very good at getting the best fusion bands.All the fusion methods can effectively enhance the effect of image,the HSV and GS transform methods can maintain the characteristics of multispectral and high resolution preferably.After fused,the features of classification are more evident.The classification accuracy of object-oriented classification method based on remote sensing indexes for wetland classification is 93.62%,compared with the traditional classification method has made great progress,so it has great application potential for wetland classification.

关 键 词：湿地分类 影像融合 最佳波段选择 评价指标 遥感指数 面向对象分类 

分 类 号：P237[天文地球—摄影测量与遥感]