GIS支持的Eigenface法——适于只有一个模型单元时的矿产定量预测方法

在研究和开发程度较低地区开展矿产预测，往往只能有很少矿产地用以建立定量预测模型，此时大多数统计分析方法不适用。一种GIS的基于“单元簇”概念的多源地学信息综合分析方法即Eigenface法，比较适合于只有一个模型单元时的建模。单元簇是相邻若干网格单元的空间定量组合Eigenface法先求出这种高维组合变量空间的某个低维特征子空间，然后计算未知单元簇和已知单元簇在特征子空间上投影点的距离来评价未知单元的的找矿有利性。Eigenface法与GIS集成是将单元作为区图元，利用GIS空间－属性分析功能提取预测信息并对单元和单元簇进行操作。以新疆乌斯拉台－亦格尔达坂地区铜铅锌银多金属矿产预测为例说明了其应用效果。

A MINERAL RESOURCE PREDICTION METHOD USING IN AREAS WITH ONLY ONE MODEL CELL: GIS-SUPPORTED EIGENFACES

In lowly developed regions, there are often few known mineral occurrences for modeling for statistical mineral resources prediction. A GIS-supported modeling method, the cell-cluster based eigenfaces, can be used in areas with only one model cell. A cell cluster is a spatial association of adjacent raster cells, each described by p geovariables. The model is a subset of the eigenvectors of the covariance matrix of the known cluster. An unknown cell is taken as the center of a cluster and then evaluated by the Euclid distance from the projection of the cluster on the eigenvector subset to the projection of the known cluster on the same subset. Cells are taken as polygon elements thus cells and cell clusters can be manipulated by GIS. A case study of Cu, Pb-Zn and Ag resource prediction in Uslatai-Ygerdaban region, Xinjiang, is presented.