滤波检测与主成分分析技术及其应用

Application of filter-detection and principal component analysis to the study of the multivariate data in lithological-mineralized section

将多元样本滤波检测与主成分分析结合用于研究岩石 矿化剖面 .本方法采用Sobel水平检测滤波和主成分分析 (PCA)技术 ,其原理是 :由于地球化学数据存在“野值”或地质体单元之间差异较小 ,很难有效地用单变量或原始多元样本划分岩石 矿化剖面中的地质体单元 ,为此 ,以广西大厂矿田钻孔岩芯样品化学分析数据为例 ,首先将 12种化学元素的样本数据设想构成一个“虚拟”的垂直面 ,用Sobel水平检测滤波器进行检测 ;根据滤波检测后的单元素 (变量 )波形特征和多元素 (多变量 )组合特点 ,划分岩石 矿化剖面中的地质体单元 ;然后 ,对指定的地质体单元进行主成分分析 ,提取原始数据的地质信息 .在PCA图解中可以发现原始变量的相互关系、数据点之间的空间特征 .

To partition the geological cells and summarize the information from a multivariate data set, the Sobel edge filter and Principal Component Analysis (PCA) are used to study the lithological-mineralizational section. Due to the outliers of geochemical data and the geochemical difference between the geological cells, it is difficult that the geological cells are effectively distinguished by single variable or a group of variables. Using the distribution of chemical elements in section of boreholes in the Dachang Tin-polymetallic Ore Field, Guangxi, as an example, a vertical 'suppositional plane' is set by a data set of 12 chemical elements, and it is detected by the Sobel edge filter. According to the wave type of single variable and the compounding of wave types for a group of variables, the geological cells in the lithological-mineralizational section are distinguished. Then, Principal Component Analysis is used to summarize the geological information for the appointed geological cells. The summarized information shown in the PCA plots is used for finding relationships among original variables and spatial variety among samples.