基于SVM决策支持树的城市植被类型遥感分类研究

Research on Remote Sensing Classification of Urban Vegetation Species Based on SVM Decision-making Tree

城市植被类型不同,生物量不同,其生态功能与绿化效应也不同。在目前难直接获取城市“绿量”实测数据的情况下,可以绿地面积和植被类型间接反映绿地的生物量和绿化效应。本文利用高分辨率卫星影像IKONOS,以实验区与验证区城市植被类型信息为对象,在对常用的参数和非参数分类方法进行对比实验的基础上,对SVM的核函数进行了分析,构建了基于SVM决策树的城市植被类型分类模型。分类实验结果表明:与其他传统方法分类结果比较,SVM的决策树分类方法对植被类型的分类精度达到83.5%,绿化面积总精度接近95%,取得了良好的效果。

Different vegetation species have different biological quality and produce different ecological functions and greenery effect. Considering the ＂Vegetation Quality＂ is difficult to be obtained, the biological quality and ecological effect of urban greenery can be indirectly reflected using urban green-land area and vegetation species. Based on the comparison of the traditional statistic parameter and non-parameter classification methods and the analysis of kernel-function of SVM, SVM decision-making tree model for urban vegetation classification is designed in this paper using the high resolution imagery data of IKONOS. The classification results are compared to other traditional methods and have an average vegetation classification accuracy of about 83.5% and green-land area accuracy nearly 95%.