Developing AI-based Remote Sensing for Rapid Mineral Exploration

Thesis Abstract

Rapid and accurate mineral exploration remains a critical challenge in the resource extraction industry, hindered by the limitations of traditional survey methods that are often time-consuming, costly, and geographically constrained. Recent advancements in remote sensing technologies and artificial intelligence (AI) present opportunities to significantly enhance mineral detection efficiency, yet the integration of AI-driven analytics into remote sensing data interpretation for mineral exploration has not been comprehensively explored. This study aims to develop a robust AI-based framework for rapid mineral exploration using remote sensing data, focusing on optimizing mineral prospectivity mapping and reducing the time-to-discovery. The research objectives include designing an integrated AI model combining machine learning algorithms with remote sensing datasets, validating the model's effectiveness across diverse geological settings, and assessing its potential for operational deployment. The methodology adopts a mixed-methods research design, combining quantitative analysis of remote sensing spectral data with qualitative evaluation of model performance. The population of the study comprises multispectral and hyperspectral satellite imagery datasets from three mineral-rich regions the Mid-Continent region, the Pacific Northwest, and the Wyoming Basin, totaling 120 satellite images collected from public repositories such as Landsat 8, Sentinel-2, and Hyperion datasets. A stratified random sampling technique is employed to select 40 images per region, ensuring representative coverage of varied geological conditions and mineralization types. Data collection instruments include satellite sensors' spectral data, geographic information system (GIS) layers, and field-verified mineral occurrence records for ground-truthing. The AI model development involves training convolutional neural networks (CNNs) and support vector machines (SVMs) to classify mineral signatures, combined with feature extraction techniques such as principal component analysis (PCA) and normalized difference spectral indices (NDSI). Model validation employs cross-validation techniques and metrics including accuracy, precision, recall, F1-score, and receiver operating characteristic (ROC) curves. Analytical frameworks draw on the theory of pattern recognition and the spatial data mining paradigm, guided by the Technology Acceptance Model (TAM) to assess the model's feasibility for practical applications. Expected results include a high-accuracy mineral classification model with an average F1-score exceeding 85%, capable of identifying mineralized zones with a spatial resolution suitable for exploration scale. The AI framework is anticipated to outperform traditional spectral analysis methods by reducing processing time by approximately 40%, enabling near-real-time mineral prospects assessment. The models' robustness across different geological contexts will validate their adaptability and scalability. These findings will contribute to knowledge by providing a novel, AI-driven approach to mineral exploration, integrating advanced machine learning algorithms with remote sensing data for enhanced prospectivity mapping. The study will also develop operational guidelines for deploying AI-based remote sensing tools in mineral exploration projects, including considerations for data preprocessing, model training, validation, and interpretation. The main conclusion indicates that AI-enhanced remote sensing constitutes a transformative approach for rapid mineral exploration, offering substantial reductions in exploration costs and timeframes while improving detection accuracy. Based on the findings, it is recommended that mineral exploration companies adopt AI-integrated remote sensing systems, tailored to specific geological settings, to complement existing exploration methods. Further research should focus on expanding the training datasets to incorporate more mineral types, enhancing model generalization, and exploring the integration of emerging remote sensing sensors, such as LiDAR and thermal infrared, to improve detection capabilities. This study provides a foundational step toward operationalizing AI-driven remote sensing in mineral prospectivity assessment, contributing significantly to the evolution of geospatial analytics in mineral resource management.