An Improved YOLOv11s Object Detection Algorithm for Slope and Road Hazard Monitoring in Open-Pit Mines

Automated visual monitoring of open-pit mine disasters requires detecting multiple hazard types with extreme scale variation—from large landslides spanning hundreds of pixels to small rockfalls under 32 × 32 pixels—against complex textured backgrounds, while maintaining real-time performance on edge devices. Existing deep learning methods address individual hazard types in isolation and struggle to balance multi-scale detection accuracy with computational efficiency. We present CEAHL-YOLOv11s, an enhanced YOLOv11s architecture integrating five specialized modules: ContextGuidedBlockDown for large-scale context modeling, C3k2_EMSC for efficient multi-scale feature extraction, AIFI_RepBN for global attention with re-parameterized batch normalization, HAFB for hierarchical attention fusion, and Detect_LADH for lightweight detection heads. Using a dataset of 7285 images with 16,368 annotated instances across four disaster categories (landslides, rockfalls, road collapses, fallen trees), CEAHL-YOLOv11s achieves 79.35% mean average precision at 154.18 frames per second on NVIDIA RTX 4090, representing a 4.26 percentage point improvement over baseline YOLOv11s. The model outperforms Faster R-CNN by 16.14%, RT-DETR by 10.21%, and YOLOv11m by 4.92% while using 2% fewer parameters than YOLOv11m. Ablation studies confirm synergistic module interactions, with per-class performance ranging from 67.85% (fallen trees) to 89.65% (landslides). The unified framework enables practical deployment for real-time multi-hazard monitoring in resource-constrained mining environments, eliminating the complexity of maintaining separate specialized systems for individual disaster types.

Authors

Institutions

Publication Details

Journal
Applied Sciences
Published
2026-11-19
DOI
https://doi.org/10.3390/app162211457
Primary Topic
Advanced Neural Network Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

An Improved YOLOv11s Object Detection Algorithm for Slope and Road Hazard Monitoring in Open-Pit Mines

Bibo Dai, Caimou Qiu, Chunyang Zhang, Xing Wang et al.
Applied Sciences
Advanced Neural Network Applications
article

An Improved YOLOv11s Object Detection Algorithm for Slope and Road Hazard Monitoring in Open-Pit Mines

Bibo Dai, Caimou Qiu, Chunyang Zhang, Xing Wang, Shuying Hu
article en

Abstract

Automated visual monitoring of open-pit mine disasters requires detecting multiple hazard types with extreme scale variation—from large landslides spanning hundreds of pixels to small rockfalls under 32 × 32 pixels—against complex textured backgrounds, while maintaining real-time performance on edge devices. Existing deep learning methods address individual hazard types in isolation and struggle to balance multi-scale detection accuracy with computational efficiency. We present CEAHL-YOLOv11s, an enhanced YOLOv11s architecture integrating five specialized modules: ContextGuidedBlockDown for large-scale context modeling, C3k2_EMSC for efficient multi-scale feature extraction, AIFI_RepBN for global attention with re-parameterized batch normalization, HAFB for hierarchical attention fusion, and Detect_LADH for lightweight detection heads. Using a dataset of 7285 images with 16,368 annotated instances across four disaster categories (landslides, rockfalls, road collapses, fallen trees), CEAHL-YOLOv11s achieves 79.35% mean average precision at 154.18 frames per second on NVIDIA RTX 4090, representing a 4.26 percentage point improvement over baseline YOLOv11s. The model outperforms Faster R-CNN by 16.14%, RT-DETR by 10.21%, and YOLOv11m by 4.92% while using 2% fewer parameters than YOLOv11m. Ablation studies confirm synergistic module interactions, with per-class performance ranging from 67.85% (fallen trees) to 89.65% (landslides). The unified framework enables practical deployment for real-time multi-hazard monitoring in resource-constrained mining environments, eliminating the complexity of maintaining separate specialized systems for individual disaster types.

Applied SciencesVol. 16(22)
Central South University (CN), Wuhan University of Technology (CN), Sinosteel (China) (CN)
Advanced Neural Network Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon