多种融合方法支持下的Landsat7-ETM+影像自动解译精度评价

对ETM+影像进行多种算法融合实验,除应用ERDAS软件现有的融合方法(PCA、MLT)外,还利用IDL语言编程实现了SFIM、HPF、MB等融合算法,通过多次修改SFIM、HPF、MB融合算法中滤波器窗口的大小、滤波算子的实验,达到既不产生噪声又增强了图像纹理信息的融合效果。对融合后的影像进行了相同地物样本、不同分类方法的监督分类。以2002年内蒙古土地利用遥感调查数据为评价标准(内蒙古自治区遥感与地理信息系统重点实验室提供),用总体精度(overall accuracy)、kappa系数两种评价指数综合反映各种融合算法与各种分类方法结合的分类精度,并对各种分类方法及融合算法予以评价。

Accurate comment of Landsat7-ETM+image automatic interpreter supported by various methods

The experiment by merging various algorithm of ETM+image,besides applying the nowadays merging methods(PCA、MLT)of ERDAS software,It also make use of IDL program to reach the merging algorithm of SFIM,HPF、MB.Through revising the size of the filter window of SFIM、HPF、MB merging algorithm and the experiment of arithmetic operators,the ETM+Image can reach the merging effect of no noise and clear image veins.By using the same surface features sample and different way of supervised classification including the minimum distance classification,mahalanobis distance classification,maximum likelihood classification and spectral angle mapper classification.According to the standard data to the survey of Inner Mongolia remote sensing surface supplied by the Lab of Remote Sensing and Geographic Information System of Inner Mongolia,using the two exponents of overall accuracy and kap- pa coefficient to reflect and to comment objectively the accurate classification by compounding the various merging algorithm and the dif- ferent methods of classification.