Information extraction and integration in mineral exploration.

Abstract

Geologic information extraction and integration are the main goals of this study. Tools are designed to aid in exploration for common mineral deposits by intelligently and efficiently processing spatial geological data. Gabor filters, comprising Gaussian-attenuated sinusoidal weight vectors, are used for textural discrimination. A highly non-linear logic operator was designed for "valley", "ridge", edge, and intersection extraction from multispectral images to cover most of the possible local lineament types. A zonation detector (a non-linear logic operator) indicates the presence or absence of lithologic zonation, the number and the types of zones using a series of automatically expanding moving windows. The ultimate window size represents the zonation size. Two different types of raster-based expert systems help optimize pixel-by-pixel knowledge extraction and representation over the spatial information and throughout the different raster feature layers. First a 2-D expert system is used for classification, ranking, recognition and searching for important pattern associations in the feature space. Second, a multilayer adaptive raster-based expert system allows the processing of multiple geologic features, and operates over each pattern in the feature layers. The fuzzy integral method of evidence fusion is used to integrate information from a variety of mineral exploration sources. This nonlinearly combines objective mineral occurrence evidence, in the form of a fuzzy membership function, with subjective evaluation of the worth of the sources with respect to the decision. An application of these methods to the Tombstone mineral district in southern Arizona demonstrates its ability to pick out circular features from TM imagery, Gabor transforms and lineament patterns, as well as identify favorable zonation for new mineral occurrence. The final product at this time is a probability map to guide the exploration geologist.