Intelligent mineral identification using clustering and artificial neural networks techniques

Identifying of minerals in petrographic thin sections is done by mineralogist using polarized microscope rotation stage. Mineral identification will be a tedious work if the number of thin sections is large; this may cause some errors in final identification. Therefore, in this study, artificial neural networks (ANNs) are utilized for mineral identification. ANNs inspired by neural activities of humans have been widely being used in myriad fields of science, they are capable of estimating complex non-linear functions. Digital images are captured from every thin section, by plane-polarized and cross-polarized lights that yield twelve features (red, green, blue, hue, saturation and intensity in two states of lights) for identification of minerals. The first six features are related to plane-polarized light and the rest are related to cross-polarize light. Then, extracted features are fed into the ANN as inputs, which has been trained therefore minerals will be recognized. The high accuracy and precision of minerals identification in this study, have given the proposed intelligent system remarkable capabilities.