Fire and smoke real-time detection algorithm for coal mines based on improved YOLOv8s

Area of Science:

• Computer Vision
• Artificial Intelligence
• Mining Safety Engineering

Background:

• Effective fire and smoke detection is critical for coal mine safety.
• Existing models struggle to balance accuracy and computational complexity for deployment in resource-constrained mining environments.

Purpose of the Study:

• To develop an efficient and accurate fire and smoke detection model for coal mines.
• To address the limitations of current models in terms of computational resources and detection performance.

Main Methods:

• Optimized the YOLOv8s object detection model by incorporating faster convolutions in the backbone and neck.
• Integrated attention mechanisms into the backbone and head components to enhance detection accuracy.
• Utilized data augmentation to improve the model's generalization capacity.

Main Results:

• Reduced model complexity by 23.0% fewer parameters and 26.4% fewer Floating-Point Operations (FLOPs) compared to YOLOv8s.
• Achieved a mean Average Precision (mAP) of 91.0%, a 2.5% improvement over the baseline model.
• Demonstrated suitability for real-time detection in environments with limited computational resources.

Conclusions:

• The proposed method effectively reduces computational complexity while improving detection accuracy for fire and smoke.
• The optimized model is well-suited for real-time fire and smoke detection in coal mines.
• This advancement contributes to enhanced safety in mining operations.