感兴趣区域遥感图像分类与支持向量机应用研究

Research on SVM and its application of remote sense image classification for regions of interest

提出了基于SVM的遥感图像分类方法并构建了分类模型,该方法以唐山1∶50000TM局部图为分类数据来源,由用户选择感兴趣的区域,分别提取该区域绿地、公共用地和房屋的图像特征,并以此为训练样本进行训练,采取交叉校验的方法获得SVM的最优惩罚因子C和间隔γ参数进行图像分类。实验结果表明,此分类方法准确率高、稳定快捷,是SVM在遥感图像分类中的一个很好的应用。

A classification method and model is proposed based on SVM for remote sense image.By selecting Regions Of Interest （ROI） from the 1：50 000TM image of Tangshan city area,extract the feature of greenbelt,public lands,building and so on,the parameters of C and T are achieved by cross validation method,with these textures to train and parameters to classify the RS image,the fact shows that the classification method based on SVM has a high accuracy and a fast,stably efficiency.