Related Experiment Video
Updated: Oct 3, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
668
Identifying Balls Feature in a Large-Scale Laser Point Cloud of a Coal Mining Environment by a Multiscale Dynamic
Zhizhong Xing1, Shuanfeng Zhao1, Wei Guo1
1College of Mechanical Engineering, Xi'an University of Science and Technology, Xi'an 710054, China.
ACS Omega
|February 21, 2022
Summary
This study introduces a multiscale dynamic graph convolution neural network (MS-DGCNN) for identifying geometric features in coal mine point clouds. The advanced model enhances environmental perception for safer, intelligent mining operations.
Area of Science:
- Geospatial AI
- Mining Engineering
- Computer Vision
Background:
- Coal mining operations generate gas, necessitating advanced environmental perception for safety and intelligent automation.
- Identifying geometric features of coal mine working faces is crucial for effective environmental perception and disaster prevention.
- Existing methods require improvement for processing large-scale 3D point cloud data from dynamic mining environments.
Purpose of the Study:
- To develop and evaluate a novel deep learning model for geometric feature identification in large-scale coal mine point clouds.
- To enhance the environmental perception capabilities for intelligent and unmanned coal mine operations.
- To provide a foundational method for real-time geometric feature recognition in complex mining settings.
Main Methods:
- An upgraded dynamic graph convolution neural network (DGCNN), termed multiscale dynamic graph convolution neural network (MS-DGCNN), was developed.
- MS-DGCNN was designed to extract both local and global features directly from point cloud data.
- The model utilized a combination of max-pooling and average-pooling as a symmetry function for feature extraction.
Main Results:
- The proposed MS-DGCNN demonstrated superior feature expression ability and improved model generalization compared to other point cloud processing deep neural networks.
- The model successfully identified geometric features, exemplified by detecting spherical objects, within large-scale coal mine working face point clouds.
- Performance benchmarks confirmed the effectiveness of MS-DGCNN in learning complex geometric patterns from custom point cloud datasets.
Conclusions:
- MS-DGCNN offers a powerful tool for geometric feature identification, significantly advancing environmental perception in coal mining.
- This research establishes a precedent for applying MS-DGCNN in the energy sector, particularly for creating transparent coal mine working faces.
- The study provides a robust foundation for developing safer, more intelligent, and automated coal mining systems.
