From 3D Perception to Safety Reasoning: A Graph-Based Framework for Real-Time Underground Mine Monitoring

2026-06-02

AI Summary

Underground coal mining operations face dynamic safety hazards such as equipment proximity violations, structural instabilities, and visual blind spots, which conventional monitoring systems struggle to detect effectively due to their lack of 3D scene understanding and contextual memory. This work proposes the first real-time monitoring framework that integrates graph-structured knowledge representation with multi-level safety reasoning, combining 3D semantic perception, uncertainty-aware anomaly detection, a rule engine, an edge-deployed large language model (LLM), and a GraphRAG memory-augmented mechanism to enable both immediate risk identification and long-term safety pattern discovery. Evaluated at 30 FPS, the system achieves 92.7% accuracy; across 115 hazardous scenarios, rule-based checks alone cover 57% of cases, which improves to 76% with LLM integration and further rises to 93% when augmented with GraphRAG-enabled memory reasoning, substantially surpassing the limitations of predefined rules and enhancing detection of complex safety hazards.

Abstract

Underground coal mining requires personnel and heavy equipment to operate within shared, confined, and poorly illuminated spaces where hazards such as equipment proximity violations, structural instabilities, and occluded blind spots are difficult to anticipate. Conventional monitoring systems, including fixed cameras and rule-based proximity alerts, can detect predefined events but lack the 3D scene understanding and contextual memory needed to identify complex or evolving hazards. This paper presents a continuous monitoring framework that converts colourised 3D point clouds into structured and traceable safety reasoning outputs. The framework combines 3D semantic perception, uncertainty-based anomaly detection, rule-based hazard checks, on-device LLM reasoning, and GraphRAG -based memory analysis to identify immediate hazards and interpret longer-term safety patterns. Scene and temporal graphs serve as the explicit knowledge structure, linking perception outputs across reasoning stages. To overcome the scarcity of labeled underground data, real roadway scans, controlled object placement, and high-fidelity longwall simulation were combined to generate diverse hazard scenarios, while self-supervised pretraining improved segmentation from limited annotations. The perception model achieved 92.7% accuracy at 30 FPS with low memory usage. Across 115 hazard scenarios, rule-based checks achieved 57% coverage, increasing to 76% with contextual LLM reasoning and 93% with memory-based reasoning using historical records. Qualitative results show uncertainty-derived anomaly signals support the interpretation of out-of-distribution hazards beyond predefined classes. Overall, graph-based knowledge representation combined with 3D perception and layered safety reasoning provides a practical foundation for intelligent decision support in underground mine monitoring.

Problem

• underground mine safety
• 3D perception
• hazard detection
• contextual reasoning
• real-time monitoring

Innovation

• graph-based reasoning
• 3D semantic perception
• GraphRAG
• uncertainty-based anomaly detection
• on-device LLM