Abstract
Accurate Ore-Waste discrimination is essential for optimizing mining operations, reducing ore dilution, and ensuring efficient resource exploitation. Conventional approaches, relying on subjective ore control interpretation by geologists, are time-consuming and cause safety hazards. This study investigates the significance of accurate Ore-Waste discriminations by employing hyperspectral remote sensing as an alternative approach. Ore and waste rock samples were collected from mock piles of an open pit gold mine. Using the Fire Assay analysis, the gold grade for each sample was measured, providing ground truth information about the ore and waste content. The hyperspectral image analysis pipeline incorporated various stages, including image pre-processing, spectra continuum removal, an innovative algorithm for absorption peak extraction utilizing the second derivative method, filtering out irrelevant features, and selecting crucial absorption peaks based on their frequencies. Utilizing a decision tree classification technique on the selected absorption peaks resulted in a robust algorithm discriminating between ore and waste rock samples. The classification outcome demonstrated an impressive F1-score of 93%, stating the reliability and efficiency of the developed algorithm. This study emphasizes the practical significance of hyperspectral image analysis in mining resource management. The high classification accuracy attained underscores the algorithm's potential for real-world applications in Ore-Waste discrimination, offering a more objective and time-efficient alternative to conventional methods.