This article presents a method to detect and segment mine waste deposits, specifically waste rock dumps and leaching wasted dumps, in Sentinel-2 satellite imagery using artificial intelligence. This challenging task has important implications for mining companies and regulators like the National Geology and Mining Service in Chile. Challenges include limited knowledge of mine waste deposit numbers, as well as logistical and technical difficulties in conducting inspections and surveying physical stability parameters. The proposed method combines YOLOv7 object detection with a vision transformer classifier to locate mine waste deposits, as well as a deep generative model for data augmentation to enhance detection and segmentation accuracy. The ViT classifier achieved 98% accuracy in differentiating five satellite imagery scene types, while the YOLOv7 model achieved an average precision of 81% for detection and 79% for segmentation of mine waste deposits. Finally, the model was used to calculate mine waste deposit areas, with an absolute error of 6.6% compared to Google Earth API results.

In this paper, we introduce a cutting-edge system that leverages state-of-the-art deep learning methodologies to generate high-quality synthetic thermal face images. Our unique approach integrates a thermally fine-tuned Stable Diffusion Model with a Vision Transformer (ViT) classifier, augmented by a Prompt Designer and Prompt Database for precise image generation control. Through rigorous testing across various scenarios, the system demonstrates its capability in producing accurate and superior-quality thermal images. A key contribution of our work is the development of a synthetic thermal face image database, offering practical utility for training thermal detection models. The efficacy of our synthetic images was validated using a facial detection model, achieving results comparable to real thermal face images. Specifically, a detector fine-tuned with real thermal images achieved a 97% accuracy rate when tested with our synthetic images, while a detector trained exclusively on our synthetic data achieved an accuracy of 98%. This research marks a significant advancement in thermal image synthesis, paving the way for its broader application in diverse real-world scenarios.

Chile’s mining industry, a global leader in copper production, faces challenges due to increasing volumes of mining waste, particularly Waste Rock Dumps (WRD) and LeachingWaste Dumps (LWD). The National Service of Geology and Mining (SERNAGEOMIN) requires assessment of the physical stability (PS) of these facilities, but current methods are hindered by data scarcity and resource constraints. This study proposes a simplified evaluation methodology using first-order parameters from open-access data. By integrating Geographic Information Systems (GIS) and Artificial Intelligence (AI)—utilizing models like YOLOv11 and convolutional neural networks—we automate the detection and characterization of WRD and LWD from satellite imagery, extracting critical parameters for PS assessment. This approach reduces analysis time and minimizes human error. Validated in the Antofagasta Region, Chile’s primary mining area, we identified and evaluated 70 WRD and 54 LWD. The results demonstrate the effectiveness of prioritizing deposits based on potential risk, enhancing SERNAGEOMIN’s capacity for supervision. The successful application suggests scalability to other mining regions and adaptability to different facility types, including tailings storage facilities. This work offers a practical tool to improve safety and risk management in the mining industry, addressing critical challenges in PS evaluation under current regulatory constraints.

MineWaste Storage Facilities (MWSFs) in Chile present substantial environmental and safety risks due to their extensive scale and the hazardous nature of their contents. This study proposes an automated detection approach that integrates Sentinel-2 satellite imagery with advanced deep learning models to address these critical issues. A central contribution of this research is the development of MineWasteCL_DB, a comprehensive public dataset comprising over 30,000 annotated images and 320,093 labels for diverse MWSF types, including Tailings Storage Facilities (TSFs), Waste Rock Dumps (WRDs), and Leaching Waste Dumps (LWDs). The study employs the YOLOv8x-seg model, selected for its high precision, to validate the presence of 96.15% of officially registered TSFs. Furthermore, it identified 141 WRDs and 112 LWDs in the Antofagasta Refgion, facilities absent from any official national registry. These findings underscore the methodology’s potential for widespread application and the necessity for routine monitoring across additional regions. The results provide a robust framework for advancing the understanding and management of MWSFs, thereby improving regulatory oversight and promoting environmental safety. The methodology supports not only the efficient monitoring of registered facilities but also the preliminary identification and prospective registration of unregistered sites. This capability enhances the oversight capacities of regulatory authorities while fostering the protection of environmental and public safety.

Periodic assessment of tailings storage facility stability is essential for mining safety and environmental protection, yet relies on slow, expert-driven manual searches for critical parameters within unstructured documents. In regulatory contexts like Chile, document heterogeneity, terminological variability, and high manual review costs create major operational bottlenecks. While recent advances in large language models and Retrieval-Augmented Generation (RAG) show potential for querying technical documentation, their performance is constrained by the information retrieval stage, particularly in non-English languages and specialized technical domains. This paper proposes AsyncMultiQuery TSF-RAG, a novel RAG system designed to mitigate embedding-based retrieval sensitivity to query formulation variations in Spanish technical documents. The system generates multiple alternative query formulations and processes them asynchronously through complementary embedding models, consolidating results through re-ranking and deduplication. Evaluation uses real-world technical reports from Chilean mining companies covering the five official tailings facility categories. For each case, expert-validated ground truth is constructed using a hybrid evaluation framework combining automatic scoring by large language models with human validation. Results demonstrate that widely used open-source embedding models exhibit limited performance on Spanish technical-mining documentation. In contrast, the proposed system automates over 70% of critical parameter searches, reducing search time by 99% compared to manual procedures while maintaining accuracy above 70%. These findings validate that query reformulation through multiple embedding models significantly improves retrieval effectiveness in specialized non-English domains, positioning AsyncMultiQuery TSF-RAG as a viable support tool for expert systems in regulatory auditing workflows.