基于决策树方法的淡水沼泽湿地信息提取——以扎龙湿地为例

The Information Extraction of Freshwater Marsh Wetland Based on the Decision Tree Method:Taking Zhalong Wetland as An Example

为了及时准确地获取湿地基础信息,对湿地进行动态监测和保护。以扎龙湿地为研究区,以区域湿地遥感信息提取为目标,采用TM影像数据、DEM数据、归一化植被指数、纹理信息等复合识别指标构建决策树模型,对研究区不同地类进行分类。然后与传统的最大监督分类法所得到的结果进行对比。结果表明,采用基于指数的决策树分类方法对扎龙湿地类型进行分类,较传统的最大似然监督分类精度提高了14.6%;总体Kappa系数提高了0.1751,分类精度较监督分类有明显的提高,证明基于多源数据决策树分类方法是内陆淡水沼泽湿地信息提取的有效手段。

In order to achieve timely and accurately basic information about wetland,which can be applied to the dynamic monitoring and protection of the wetland.The author chose zhalong wetland as the research area,during the process of extracting regional remote sensing information by using the TM image data,DEM data,normalized vegetation index,texture information compound identification index,finally the author classified the types of zhanglong wetland through constructing a decision tree model.For checking the feasible degree of method of classification based on decision tree model,the author made a comparison between the traditional maximum classification of supervision and the decision tree model.The results showed that：the decision tree method based on the index was used to classify,classification accuracy increased by 14.6%,overall Kappa coefficient increased by 0.1751,supervised classification accuracy was improved noticeably.Building decision tree classification which adopted multi-source data for extracting information of inland freshwater mire wetland was a very effective approach.