基于Landsat 8影像的不同监督分类方法精度对比分析
Comparison of Accuracy of Different Supervised Classification Methods Based on Landsat 8 Images

监督分类是遥感图像分类的常用方法。为了探究不同监督分类方法的精度，分别使用神经网络、波谱角、最大似然、最小距离、光谱信息散度、马氏距离、平行六面体、支持向量机法、二进制编码等9种算法监督分类方法对云南省禄丰县Landsat 8影像进行分类及精度分析，结果表明：从总体精度来看，神经网络分类法、支持向量机法、最大似然法的精度较高；从各地类分类精度来看，最大似然法对草地以及耕地的分类精度最高，神经网络分类法对林地和未利用地的分类精度最高，支持向量机法对城乡工矿居民用地和水域的分类精度最高。
Supervised classification is a common method of remote sensing image classification. In order to explore the accuracy of different supervised classification methods, we used neural network, spec-tral angle, maximum likelihood, minimum distance, spectral information divergence, Mahalanobis distance, parallelepiped, support vector machine, binary coding methods to classify and precision analysis with Landsat 8 images in Lufeng county, Yunnan Province. The results show that: from the perspective of overall accuracy, the accuracy of the neural network, support vector machine, and maximum likelihood method is high; from the accuracy of different land types, the maximum like-lihood method has the highest classification accuracy for grassland and cultivated land, the neural network classification method has the highest classification accuracy for forest land and unused land, and the support vector machine method has the highest classification accuracy for urban and rural industrial and mining residential land and water.