监督分类方法在遥感影像分类处理中的比较

Comparison of Supervised Classification Methods in Remote Sensing Image Classification

随着计算机解译技术的发展,遥感影像分类方法不断涌现,各种分类器分类精度不一的问题,对其应用产生了一定的影响。运用ENVI软件,采用同一地区的Landsat TM影像,通过目视解译选择相对应的训练样本和已有的验证样本进行试验,对监督分类的6种分类器(最大似然、神经元网络、支持向量机、最小距离、马氏距离、平行六面体分类方法)进行分类后的精度比较。通过对试验区的地物做分类结果的评判和比较研究,再经过分类后处理,得出分类结果的总体精度和Kappa系数。结果表明,最大似然分类方法的精度明显高于其他分类方法的精度,而对比分类影像的细部图,也优于其他分类法,即在监督分类中,最大似然分类法具有较好的分类效果。

With the development of computer interpretation technology, remote sensing image classifi- cation methods are emerging, and the classification accuracy of various classifiers is different, and their application has a certain influence. In this paper, the use of ENVI software, using the same area of Landsat TM images, through visual interpretation of the corresponding training samples and the ex- isting validation samples were tested, the classification of the six classifiers （maximum likelihood, neural network, Support vector machine, minimum distance, Mahalanobis distance, parallelepiped classification method） Through the classification and post-processing of the classification results of the test area, the overall accuracy and the Kappa coefficient of the classification result are obtained. The results show that the accuracy of the maximum likelihood classification method is significantly higher than that of other classification methods, and the detailed image of the classification image is superior to other taxonomy. In the supervised classification, the maximum likelihood classification method has better Classification effect.