GIS-based multifractal/inversion methods for feature extraction and applications in anomaly identification for mineral exploration.

摘要

Mineralization is often intertwined with other processes spatially. This makes it difficult to extract features for mineral exploration. However, the existing techniques are far from adequate in support of this purpose.; A series of multifractal feature extraction techniques in spatial, Walsh, eigenspace domains and other methods were developed in a GIS environment for mineral prospecting in this thesis.; Techniques in spatial domain including spatial moments, gradient parameters, and local singularity are reviewed and implemented with the emphases on singularity analysis which extracts features on the basis of local self-similarity and spatial association property.; A new multifractal method (W-A) was developed in the Walsh domain. W-A model is demonstrated to be advantageous for extracting abruptly change features. This advantage comes from its square wave functions of Walsh transformation (WT).; A new multifractal singular-value decomposition (MSVD) model is developed on the basis of scale invariance in eigen-space for features extraction. The eigenimage and power spectrum structure of the studied area are investigated. The extracted feature using MSVD method characterizes rich textures, particularly capable of extracting weak and subtle features from data with strong influence of background and (fault) sharp change values.; New Gauss inversion (GI) and hierarchical decomposition methods have been developed for distinguishing probability density function (PDF) from mixing populations. The forward modeling, the least square (LS) segmentation, and the GI are compared. These methods were used in estimating the spatial and entropy distributions. These features have rich textures portraying underground intrusions that are related with the hydrothermal mineral alteration in the study area.; The data from southwestern Nova Scotia, Canada, were processed. The results have shown that the features extracted using the techniques developed are associated with a prior mineral deposits knowledge well.