A comprehensive evaluation method for dust pollution: Digital image processing and deep learning approach.

Area of Science:

• Environmental Science
• Computer Science
• Engineering

Background:

• Dust pollution is a significant environmental and health hazard, particularly in mining.
• Existing dust monitoring methods may lack accuracy or automation.
• Effective dust management is crucial for industrial safety and sustainability.

Purpose of the Study:

• To develop and evaluate a novel method for dust pollution assessment.
• To integrate grayscale average (GA) analysis and deep learning (DL) for image-based dust classification.
• To enhance dust monitoring accuracy and applicability across various industrial environments.

Main Methods:

• A dust diffusion simulation system was used to generate 300 sample images.
• Grayscale average (GA) analysis was employed to correlate image data with dust mass.
• Fractal dimension (FD) was incorporated to refine classification criteria.
• Deep learning (DL) models were trained and validated for dust classification.

Main Results:

• The combined GA and DL method achieved a testing accuracy of 92.2%.
• High precision, recall, and F1-score values were obtained, indicating robust performance.
• The method demonstrated effectiveness in classifying dust pollution levels.
• The approach showed versatility for application beyond mining operations.

Conclusions:

• The novel image-based method offers an automated and reliable solution for dust pollution monitoring.
• This approach significantly advances environmental monitoring, enhancing safety and health outcomes.
• The integrated GA and DL technique contributes to mitigating dust pollution and promoting sustainable industrial practices.