Mobile Phone Detection For Mining Teleoperation Safety

Although remote operation improves operator safety by re- moving personnel from hazardous mining environments, it introduces critical challenges in maintaining situational awareness across multiple distraction vectors. This work presents an integrated monitoring frame- work that combines four safety validations: posture analysis, hand posi- tion tracking, operator presence detection, and mobile phone detection. Initial deployment revealed a critical performance gap: while three be- havioral validations operated effectively, standard object detectors failed dramatically for phone detection, achieving only 0.03 F1-score. This deficit motivated a systematic investigation that led to a comprehensive comparative analysis of all five YOLO11 variants (n, s, m, l, x) retrained on 2,879 domain-specific images using 5-fold cross-validation to ensure rigorous validation. Our key finding is that all variants achieved statisti- cally equivalent performance ([email protected] > 95%, p = 0.958), establishing YOLO11-Nano as the optimal choice due to its 415 FPS throughput and minimal 4.2 MB footprint. The integrated four-component frame- work achieves an overall accuracy of 87% while implementing a risk- based alert hierarchy that prioritizes mobile phone use (the most cogni- tively engaging distraction) above posture deviation, hands-off controls, and operator absence. This work offers mining operations a scalable, efficiency-optimized solution for comprehensive distraction monitoring.