Research on downhole drilling target detection based on improved Yolov8n

Area of Science:

• Mining Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Underground coal mine drilling is crucial for managing gas, water, and geological hazards.
• Effective real-time monitoring and target detection are vital for enhancing coal mine safety and operational efficiency.

Purpose of the Study:

• To develop an improved algorithm for real-time target detection at underground coal mine drill sites.
• To enhance the efficiency and accuracy of identifying key drilling site components and personnel.

Main Methods:

• An improved algorithm based on Yolov8n was proposed, incorporating a C2f_PKI module in the Backbone and FDPN/DASI fusion modules in the Head.
• A Focaler_MDPIoU loss function and MLCA attention mechanism were integrated to improve feature fusion and focus.
• The model was trained to detect five key categories: grippers, drill chucks, coal miners, mine helmets, and drill pipes.

Main Results:

• The improved model achieved a 14% reduction in storage size and a 17% decrease in parameters compared to the original.
• Precision was enhanced by 1.4%, and mean average precision (mAP) increased by 0.9%.
• The algorithm demonstrated improved recognition accuracy and generalization ability in complex underground environments.

Conclusions:

• The proposed algorithm offers a significant technical advancement for underground target recognition in challenging mining conditions.
• The enhancements contribute to strengthening disaster prevention and control capabilities in coal mines.
• This research provides a more efficient and accurate solution for real-time monitoring of underground coal mine drilling operations.