Integrating image processing and deep learning for effective analysis and classification of dust pollution in mining processes

Abstract

A comprehensive evaluation method is proposed to analyze dust pollution generated in the production process of mines. The method employs an optimized image-processing and deep learning framework to characterize the gray and fractal features in dust images. The research reveals both linear and logarithmic correlations between the gray features, fractal dimension, and dust mass, while employing Chauvenel criteria and arithmetic averaging to minimize data discreteness. An integrated hazardous index is developed, including a logarithmic correlation between the index and dust mass, and a four-category dataset is subsequently prepared for the deep learning framework. Based on the range of the hazardous index, the dust images are divided into four categories. Subsequently, a dust risk classification system is established using the deep learning model, which exhibits a high degree of performance after the training process. Notably, the model achieves a testing accuracy of 95.3%, indicating its effectiveness in classifying different levels of dust pollution, and the precision, recall, and F1-score of the system confirm its reliability in analyzing dust pollution. Overall, the proposed method provides a reliable and efficient way to monitor and analyze dust pollution in mines.