基于光谱和纹理的SVM矿化蚀变信息提取研究

针对传统矿化信息提取方法单一,利用光谱或纹理、信息量相对较少、需要大量样本的缺陷,利用基于光谱和纹理的支持向量机(SVM)原理,建立矿化信息提取模型.选择青海泽库县析界日地区作为典型研究区.首先提取研究区光谱和纹理信息,选取训练样本;然后求解最优超平面,进而确定决策函数;最后泛化推广识别其他待识别的样本.通过所提取的遥感蚀变异常信息与重砂异常点叠加分析,叠加基本吻合;从野外实地验证来看,均发现了不同程度的矿化现象,并指出了5个重点异常区.

THE SPECTRUM AND TEXTURE-BASED SVM MINERALIZATION-ALTERATION INFORMATION EXTRACTION

Due to the singleness of the spectrum-texture-mineralization information extraction and less information extracted the principle of the spectrun-texture-based SVM is applied to set up the mineralization information extraction model.Xijieri area,Zeku county,Qinghai province is chosen as the test area.Firstly the spectrum and texture information was extracted and training sample pickod up,then the optimum hyperplane calculated and decision-making function determined,lastly it is used to distinguish other samples to be distinguished.The extracted remote sensing anomly is basically coincided with the heavy mineral anomly and the more or less mineralization is recognized by field check.Five key anomalies are outlined.