随机森林在城市不透水面提取中的应用研究

Application of Random Forest Model in the Urban Impervious Surface Extraction

基于高分一号数据采用随机森林模型,引入归一化水体指数NDWI、归一化植被指数NDVI、土壤调节植被指数SAVI、建筑指数BAI、Brightness指数、ICA独立分量、纹理信息及原始影像波段等共17个特征,在matlab环境下提取了深圳市某区域不透水面,并与最大似然分类和支持向量机方法进行对比.结果表明:随机森林的多特征组合方法能有效提升分类精度,在不透水面信息提取中比传统参数分类方法(MLC)总体精度提高了7.681%,Kappa系数提高了0.119 4.

Impervious surface is an important index of urbanization and environmental quality assessment.This article is based on GF1 data using random forest model, extracted 17 Features, including normalized difference water index（NDWI）, normalized difference vegetation index（NDVI）, soil adjust vegetation index（SAVI）, building index （BAI）, brightness index, independent component, texture in- formation and image bands.Extract impervious surface of Shenzhen region in the matlab,and with the maximum likelihood classification method and support vector machine were compared. Results show that random forest method can effectively improve classification accuracy, the overall accuracy increased by 7.681% than traditional parameter classification method（MI.C）, the Kappa coefficient increased 0.119 4.