摘要
统计分析了湖北保康磷矿区有关目标地物(道路、建筑物、坡耕地、植被、水体及阴影等)的SPOT 5影像特征,认为它们具有一定相似性和差异性,依赖单一的遥感分类方法难以实现对矿区开采面及固体废弃物信息的准确提取。采用决策树分类方法,设置一定的分类规则,结合数字高程模型和含矿地层等相关辅助数据,逐一对矿区相关地物进行分类,经分类结果后处理,分类精度达83.4%。
Based on a statistic analysis of spectral characteristics of such objectives as road,building,sloping farmland,vegetation,water and shade on the SPOT 5 remote sensing image in the Baokang phosphorite ore district of Hubei Province,this paper holds that spectral properties of these objectives have certain similarity and difference,and it is difficult to extract the mining area and the solid waste accurately based only on a single classification method.Making use of the decision tree classification and setting up some classification rules in combination with the related auxiliary data from the digital elevation model and the ore-bearing strata,the authors successfully classified the objects in the ore district into various categories.Subsequent processing of the classification results shows that the classification precision can reach 83.4%.
出处
《国土资源遥感》
CSCD
2009年第2期87-90,共4页
Remote Sensing for Land & Resources
基金
中国地质调查局"重点成矿带及矿集区矿产资源开发多目标遥感调查与监测"项目(1212010611208)
关键词
决策树分类
正射纠正
形状指数
Decision tree classification
Orthophoto rectification
Shape index
