摘要
提出一种基于支持向量机(SVM)遥感数据矿化蚀变信息提取的新方法。该方法首先根据蚀变岩及矿体围岩的实测光谱数据,利用光谱角度制图法(SAM)提取训练样本,应用交叉比对(cross-validation)算法确定最优SVM模型参数,选择径向基(RBF)核函数,训练SVM分类器模型;然后,用训练好的SVM模型进行遥感矿化蚀变信息提取;最后,选择青海芒崖地区的ETM数据进行遥感矿化蚀变信息提取试验。试验结果经野外检查和验证,效果良好。
A new method for extracting mineralization information from remote sensing image based on Support Vector Machines (SVM) is presented in this paper. According to the field measured spectral data of mineralized alteration rocks and wall rocks, the authors first extracted the training examples by Spectral Angle Mapper (SAM), and then selected the RBF as the kernel function. After that, cross -validation algorithm was applied to seek superior SVM model parameters. This model was used to extract mineralization information from remote sensing image in Mangya area, Qinghai province. Practice has proved that this method is effective in extracting mineralization information.
出处
《国土资源遥感》
CSCD
2006年第2期16-19,i0002,共5页
Remote Sensing for Land & Resources
基金
国家"十五"科技攻关计划项目(2003BA612A-04)
中的"SVM遥感数据矿化信息提取技术研究"资助
关键词
SAM
SVM
矿化蚀变信息
提取
遥感数据
SAM
SVM
Mineralied and altered rock information
Extraction
Remote sensing data