基于KPCA和SAM的城市植被遥感分类研究

Research on Classification for City's Vegetation Type Based on KPCA and SAM

研究高分辨率遥感城市绿地信息自动提取技术是城市遥感技术应用亟待解决的问题之一。城市绿地分布破碎,林种多样,林相不齐,具有极强的非线性特征。核主成分分析(KPCA)可以表达图像像素间的高阶关系,因而可以提取图像的非线性特征,同时提供一组相互独立的主成分。通过实验分析核函数的参数,比较变换前后的平均可分性,进行波段选择。将KPCA与SAM分类方法结合,构建基于KPCA的SAM城市植被分类方案。实验结果表明,该方案比传统的分类方法精度高。城市6种绿地类型的分类总精度为80.6%;合并为草地、园地与林地绿地类型时分类总精度达91.7%,可以满足城市植被分类与生态评价的需求。

It is an important problem for city＇s remote sensing application to research extracting green - land from high resolution IKONOS image automatic. City＇s green- land is fragmentation, multiplicity of type, and appearance out of order. It is strong nonlinear characteristic. The KPCA can express this relationship with pixels and provide an independent KCP. The selection of kernel - func- tion, including the selection of the function＇s parameter and the suitable number of training samples in the KPCA transform method is discussed in this paper. That is useful scheme for developing a city＇s vegetation classification combined KPCA and SAM. The result indicates that total classification precision is 80.6% for 6 kind types and 91.7% for incorporative type of grass,garden plot and woodland.And it can meet needs of city＇s vegetation classification and ecological appraisement.