摘要
针对传统矿化信息提取方法单一利用光谱或纹理,信息量相对较少且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。选择青海黄南州吉地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它待识别的样本。通过所提取的遥感蚀变异常信息与原有矿区叠加分析,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4个重点异常区。
The traditional mineralization which was extracted from remote sensing images only by spectrum or texture has many shortcomings. It may lack information or has a great demand for samples. A new method for extracting mineralization from remote sensing images by SVM based on spectrum and texture is presented in this paper. Jidi was selected as the typical research area. Firstly, selecting the training sample from the research area on the basis of spectrum and texture; secondly, finding out the most optimal hyperplane and decision-making functions finally, extracting the alteration information. After analyzing the spot investigation and comparing the data of the known alteration areas in the mineral geology mapping of field, it is proved that the alteration information is nearly in accordance with the known alteration areas, and four important alteration ahnormity districts have also been plotted out.
出处
《地球与环境》
CAS
CSCD
北大核心
2008年第1期81-86,共6页
Earth and Environment
基金
中国地质调查项目(1212010660601)
