利用国产CBERS卫星影像进行矿业城市土地覆盖分类

Land Cover Classification in Mining Areas Using CBERS Remotely Sensed Images

中巴地球资源卫星(CBERS)作为目前最主要的国产资源卫星,在土地覆盖监测中具有广泛的应用前景。结合土地覆盖遥感分类新技术的发展和应用,以典型煤矿城市为例,对CBERS遥感影像应用于矿业城市土地覆盖分类的若干新方法进行了试验与分析。在这些新方法中,面向对象分类方法克服了传统基于像素分类中存在的问题,支持向量机在分类精度方面比传统分类器具有一定的优势,混合像元分解方法则实现了亚像元级地物成分比例的估算。这些新方法的有效应用将促进国产卫星数据源在典型区域的应用,同时服务于遥感专题应用信息处理精度的改进和提高。

As a major satellite remote sensing system,CBERS（China-Brazil Earth Resources Satellite） has wide applications in land cover monitoring.Oriented to the new developments of land cover classification,some novel remote sensing classification approaches are experimented and applied to land cover classification in two mining industrial cities.In those novel methods,support vector machine（SVM）outperforms traditional classifiers in terms of classification accuracy,and object-oriented classification methods can overcome the problems of traditional pixel-based classification to a great extent.In contrast with those hard classification schemes,mixed pixel decomposition（or unmixing） can estimate the abundance of each endmember in a pixel,which is helpful to target identification and sub-pixel classification.Those new approaches are able to push the applications of CBERS remotely sensed data to land cover classification in mining areas,and further improve the accuracy and performance of remote sensing applications.