Open Pit Mine Integrating YOLO Optimization Model and Triangulation Network Stope Data Extraction Method

• Published in: Archives of mining sciences = Archiwum górnictwa Vol. 70; no. 2; p. 205
• Main Author: Li, Tianwen
• Format: Journal Article
• Language: English
• Published: Warsaw Polish Academy of Sciences 2025
• Subjects: Accuracy; Anniversaries; Deep learning; Mineral resources; Mines; Open pit mining; Optimization models; Remote sensing; Semantics; Triangulation; Unmanned aerial vehicles
• ISSN: 0860-7001; 1689-0469
• DOI: 10.24425/ams.2025.154659

To achieve the automatic, rapid, and precise extraction of stope data from open-pit mines, this paper introduces a novel stope data extraction method based on an enhanced Mine-YOLO model integrated with a triangulated network. An attention mechanism is incorporated to improve the capture of channel, spatial, and global multi-scale features, enabling the model to effectively consider both global context and boundary details of open-pit stopes while enhancing its ability to distinguish positive samples through an optimized loss function. Following dataset training and validation, the average accuracy for stope identification and segmentation using the Mine-YOLO model has improved by 0.15 and 0.079 respectively compared to the baseline model. The Mine-YOLO framework is employed to extract stope areas from DEM data; subsequently, indices such as stope area, volume, and mining depth are automatically calculated via a constructed triangulation network. The average errors in extracted stope area, volume, and mining depth are found to be 0.058, 0.047, and 0.002 respectively – demonstrating that the proposed methodology possesses high accuracy and significant practical application value.