基于决策树的漓江上游土地覆盖分类

发布时间：2018-03-19 19:12

本文选题：漓江上游　切入点：决策树　出处：《测绘科学》2016年03期 　论文类型：期刊论文

【摘要】：针对山区植被分类受地形复杂、植被类型多样、验证数据获取困难等因素限制基于多光谱数据的亚热带山区土地利用/覆盖分类存在困难,探究利用物候信息对亚热带山区植被实施分类的效果。综合运用归一化植被指数(NDVI)、比值植被指数(RVI)、归一化水指数(NDWI),同时考虑到海拔高度对植被类型的影响,建立决策树模型。该模型基于多时相Landsat TM影像,利用了不同地物类型的物候特征和光谱差异,将漓江上游地区分为8种土地覆盖类型。实验结果表明,分类结果总体精度达到86.40%,Kappa系数为0.83。
[Abstract]:In view of the difficulty of land use / cover classification in subtropical mountainous areas based on multi-spectral data, the classification of vegetation in mountainous areas is restricted by the complex terrain, diverse vegetation types and difficulty in obtaining data. To explore the effect of applying phenological information to classification of vegetation in subtropical mountain areas. Comprehensive use of normalized vegetation index (NDVI), ratio vegetation index (RVI), normalized water index (NDWI), and taking into account the influence of altitude on vegetation type, A decision tree model was established, which was based on multitemporal Landsat TM images, and was divided into 8 land cover types in the upper reaches of the Lijiang River by using the phenological characteristics and spectral differences of different ground objects. The experimental results show that the land cover of the upper reaches of the Lijiang River can be divided into 8 types. The overall accuracy of the classification is 86.40 and the Kappa coefficient is 0.83.
【作者单位】：山东科技大学测绘科学与工程学院;中国科学院遥感与数字地球研究所数字地球重点实验室;环境保护部卫星环境应用中心;
【基金】：国家自然科学基金项目(41471369) 国家科技支撑计划(2012BAC16B01)
【分类号】：P237

【参考文献】