基于SAR数据后向散射及光学图像纹理特征的橡胶林提取

摘要

Based on the all-polarimetric SAR data of TerraSAR-X satellite and Google high-defini-tion satellite image,this paper extracted and analyzed the backscattering coefficient and texture features of rubber plantation in Mengla County,identified the rubber plantation by the classification method of de-cision tree.The results showed that the backscattering of HH,HV,VH and VV in the fully polarimetric TerraSAR data and the calculation between them were the important foundations of the decision tree algo-rithm for the classification of forest land, agricultural land, water and construction land.As a kind of special plantation,the rubber plantation had a fixed row spacing and unique texture characteristics which could be identified from natural plantation.The rubber plantation was extracted by the decision tree clas-sification method combined with backscattering and optical image texture features of SAR data, which showed that the overall accuracy was 87.6%, the Kappa coefficient was 0.81 and the accuracy of the producer and the user were over 90 %.%以勐腊县为研究区,基于TerraSAR-X卫星的全极化SAR数据以及谷歌高清卫星影像图,对橡胶林的后向散射系数以及其纹理特征进行提取分析,采用决策树分类方法对勐腊县的橡胶林进行分类识别.研究结果表明:全极化TerraSAR数据的HH、HV、VH、VV的后向散射及其之间的计算是构建决策树算法,实现森林、农用地、水体和城市建设用地分类的重要基础;橡胶林作为一种特殊的人工林,其具有固定的行株距以及特有的纹理特征,能把橡胶林和天然林更好地区分开;利用结合SAR数据后向散射及光学图像纹理特征的决策树分类方法对橡胶林进行提取,总体精度为87.6%,Kappa系数为0.81,橡胶林的生产者精度和用户精度均达到90%以上.