Analyzing point cloud of coal mining process in much dust environment based on dynamic graph convolution neural...

Area of Science:

• Geosciences
• Computer Science
• Mining Engineering

Background:

• Underground coal mines generate significant dust, impacting environmental perception and worker safety.
• Accurate point cloud data is crucial for converting mining coordinates to geodetic coordinates, essential for large-scale operations.

Purpose of the Study:

• To develop a robust method for identifying marker balls within coal mine point clouds.
• To enable accurate coordinate conversion for large-scale, fully mechanized mining faces.
• To enhance the intelligence and safety of underground mining operations.

Main Methods:

• Generated multi-density geometry point clouds from complete and incomplete data, addressing uneven distribution.
• Increased the weight of point cloud normal vectors in network training to enhance sensitivity to geometric features.
• Utilized advanced deep neural networks, including dynamic graph convolution neural network (DGCNN), for direct point cloud analysis.

Main Results:

• The proposed method, combined with DGCNN, accurately identifies marker balls in massive coal mining point clouds.
• The approach effectively handles uneven point cloud distribution and enhances geometric feature extraction.
• Demonstrated improved sensitivity of the network model to subtle geometric features.

Conclusions:

• The developed method significantly improves the accuracy of marker ball identification in coal mine point clouds.
• This research enhances production efficiency and safety in fully mechanized mining.
• It lays the groundwork for intelligent mining and mitigation of hazards like dust explosions.