Harnessing the Power of Machine Learning for Mapping of Iron Oxide Mineralization using Sentinel Satellite Spectral Imagery

In recent decades, satellite-based mineral exploration has been recognized as one of the most cost-effective and time-efficient tools for regional exploration. In this study, three advanced digital image processing techniques based on machine learning were used on multispectral Sentinel satellite data for mapping of iron oxides: Spectral Band Combination (SBC), Spectral Band Ratio (SBR), and Spectral Angle Mapper (SAM). The satellite data was preprocessed in parallel using image enhancement algorithms on high computational machines. The results demonstrated that the SAM technique was more promising and efficient for accurate mapping of iron oxide mineralization in South Africa’s Northern Cape region, with an overall accuracy of 81.62 percent. The cost-effective, robust, and efficient methodology developed in this study can explore iron oxide mineralization resources in other regions with similar geographical and geological characteristics. The findings of this study can be a starting point for developing a mine plan, along with other ground data sets from the mining regions.