利用决策树模型的湿地提取与分类

Wetland extraction and classification using a decision tree model

为解决因湿地环境复杂且类型多样导致光谱混淆而难以对其自动遥感提取的问题,采用决策树模型的湿地信息提取方法,以Landsat OLI影像光谱特征和经缨帽变换后的数据为基础,结合不同类型湿地的环境特征和空间特征信息,提出先分区再分类的思想,构建决策树分类模型.对原始影像进行缨帽变换,利用变换后的湿度分量及地物的光谱特征规律,将研究区划分为水体区域、植被区域和非植被区域,然后分别对各个区域进行再分类,逐层分级,最终实现不同类型湿地的分级提取.研究结果表明:采用分区分类思想构建决策树模型,可以有效提取湿地信息,精度达87.50%.

In order to solve the problem of automatic remote sensing extraction, because of unclear spectrum caused by complex environment and diverse types of wetland, this paper used a decision tree model to extract wetland information, and by combining with the data characteristics of tasseled cap transformation and geometric feature information of wetland, proposed this method which is based on the Landsat OLI image spectral characteristics to construct the decision tree classification model. Firstly, this study handled the original image to do tasseled cap transformation, the study area is divided into water area, vegetation and non vegetation area by the transformed humidity component and spectral features law, then reclassify every area and classify layer and layer, finally realize the classification extraction of different types of wetland. The research results show that the method of using partition classification idea to construct a decision tree model can effectively extract the wetland information, and the accuracy can be up to 87.50%.