A DECADE OF ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING IN REMOTE SENSING FOR MINERAL IDENTIFICATION: A SYSTEMATIC REVIEW

Abstract

Artificial Intelligence (AI) and Machine Learning (ML) have become increasingly integrated with Remote Sensing (RS) to support mineral identification in both terrestrial and planetary contexts. Recent advances in hyperspectral sensors (e.g., the GaoFen-5 [GF-5] satellite and HySpex airborne imaging systems) and analytical algorithms have enabled more precise mineral mapping. Despite these advances, several challenges remain, including inconsistent preprocessing, limited standardized datasets, and issues of interpretability that restrict the operational use of AI-based mineral identification models. This study applied a PRISMA 2020-guided systematic review of literature published between 2015 and 2025. A total of 137 records were screened; after applying inclusion and exclusion criteria, 40 studies were selected for eligibility, and 14 were analyzed in detail. The classification of studies was organized into three methodological categories: classical machine learning (e.g., Support Vector Machines, Random Forest), deep learning (e.g., Convolutional Neural Networks, Vision Transformers), and statistical approaches. Classical algorithms remain competitive when training data are limited, while deep learning methods provide stronger performance in capturing spectral–spatial features, particularly when large datasets are available. Multimodal fusion of VNIR, SWIR, and TIR data improved classification accuracy by up to 13% in selected studies. However, overfitting risks and limited interpretability were consistently reported as critical drawbacks. This review highlights the need for explainable AI frameworks, open benchmark datasets with harmonized metrics, and lightweight edge computing architectures to enable operational deployment in mineral exploration. Future research should prioritize cross platform generalization and reproducibility to ensure scalability across diverse terrestrial and planetary environments.