摘要
本文基于Landsat 8开源遥感影像,运用监督分类的方法对研究区土地利用覆盖进行自动分类,并对分类结果进行后处理。以目视解译的分类结果为精度验证数据,通过计算混淆矩阵进行分类精度评价。得到如下结论:支持向量机Majority/Minority后处理的效果最佳,总体精度为71.47%,Kappa系数为0.6322。未经过后处理的神经网络的分类效果最差,总体精度为68.55%,Kappa系数为0.5973。支持向量机Majority/Minority后处理效果最好,可用于大面积快速提取土地利用覆盖分类。
Based on the open source remote sensing image of Landsat 8,this paper uses the supervised classification method to automatically classify the land cover in study area,and post-process the classification results.The classification accuracy was evaluated by calculating confusion matrix,using the visual interpretation results as the accuracy verification data.The conclusion is as follows:the result of Majority/Minority post-processing is the best,with an overall accuracy of 71.47%and a Kappa coefficient of 0.6322.The neural network without post-processing had the worst classification effect,with an overall accuracy of 68.55%and a Kappa coefficient of 0.5973.Majority/Minority support vector machines have the best postprocessing effect and can be used to quickly extract land use cover classification in a large area.
作者
王家福
WANG Jia-fu(Yunnan Chihong Zinc Germanium Co.,Ltd.Qujing 655000,Yunnan)
出处
《世界有色金属》
2022年第16期125-128,共4页
World Nonferrous Metals
关键词
土地利用覆盖分类
Landsat
8
监督分类
分类后处理
Land use and cover classification
Landsat 8
supervision classification
classification post processing
