基于ETM影像数据融合的土地覆盖分类研究

在ERDAS软件支持下，对ETM遥感影像数据的TM1-TM5，TM7与其全色波段TM8进行融合，采用主成分、乘积法、Brovey转换三种融合方法，重采样方法分别为邻域法、立方卷积法及双线性内插法。采用相同的训练样本区及最大似然法分类方法，对融合产生的9幅影像及未融合影像进行土地覆盖分类，通过对分类影像的Producers Accuracy，Users Accuracy，Kappa三者的精度数据和地物波谱信息的对比分析，在总体上，上述的影像融合方法对提高土地覆盖分类的精度不明显，但就某些地物类型来说，还是值得采用的；三种融合方法和三种重采样方式它们之问相比较而言，乘积法融合法和立方卷积重采样法相对较为可取。

Study on Land Cover Classification Based on the Landsat7 Image Data Fusion

Supported by the ERDAS8.5 software, the TM1-TM5 and TM7 of landsat7 remote sensing image data are fused with the panchromatic band of landsat7. The methods of fusing include the principal component, the multiplicative, the brovey transform. The methods of resample include the nearest neighbor, the cubic convolution and the bilinear interpolation. The identical training sample area and the maximum likelihood classification method, the land cover of the nine sheets of image fused and one sheet of image non-fused are adopted. Through contrasting and analysing the accuracy data of the producers accuracy, the users accuracy and the kappa and the ground object spectrum information, it is not obvious that the mentioned image fusion methods can increase the accuracy of the land cover classification, but for some ground object types, the fusion methods are value to be used. Compared the three fusion methods and the three resampling techniques with themselves, the multiplicative fusion method and the cubic convolution resampling technique are relative better.