采用随机森林法的天绘数据干旱区城市土地覆盖分类

Random forest classification of land cover information of urban areas in arid regions based on TH- 1 data

基于天绘一号(TH-1,或称MS-1)卫星多光谱数据,采用随机森林分类方法(random forests classification,RFC)对位于中亚干旱区的我国新疆维吾尔自治区阿勒泰地区北屯市及周边区域的土地覆盖进行了分类研究。针对北屯市不透水层与裸土混杂的情况,将纹理特征与植被信息构建最优组合,建立有效的RFC分类器,提高对易混淆土地覆盖类型的分类识别精度。结果表明,采用RFC的分类精度高于最大似然法分类结果,总体分类精度提高了近10%。经过优化选择的特征组合在对干旱区中小城市土地覆盖进行分类时表现良好,能得到较高精度的分类结果,可满足新疆中小城市发展规划对土地覆盖信息的需求。

Random- forest classification（ RFC） method was used to extract the land cover information from the TH-1 satellite remotely sensed multispectral data in Beitun Town and its adjacent areas within the arid region of Altay,Xinjiang. Owing to the mixture of the impervious covers and the exposed soils inside the city,the textural and vegetation features were derived from the TH- 1 panchromatic image and multispectral bands and subsequently applied to creating optimal feature set so as to implement the RFC classification. The optimized classifier can achieve better identification of some confused land cover classes. The results show that the RFC possesses higher accuracy than the conventional maximum likelihood classification（ MLC） with the same TH- 1 image,with their total accuracy being 82. 26% and 72. 61%,respectively. In addition,favorable applicability is observed in the land cover classification in the arid urban region using optimized combined multi- feature methods,which can provide land cover information for the urban development and planning in the medium and small cities of Xinjiang.