Real-time and reliable detection of dust concentration is critical for ensuring safety in coal mine operations. However, existing image-based dust concentration detection methods often suffer from limited interpretability and insufficient accuracy. To address these challenges, a dust concentration image recognition model that integrates prior interpretable features with a multi-kernel residual attention network was proposed in this study. First, an image feature library was constructed to capture multidimensional characteristics, including color, texture, and geometric features. The Pearson correlation coefficient and mutual information were then employed to select a subset of linear and nonlinear features that are highly correlated with dust concentration, thereby enhancing model interpretability. Subsequently, a multi-kernel learning framework was developed, in which radial basis function kernels were used to separately map the selected linear and nonlinear features into high-dimensional feature spaces. A weighted fusion strategy was introduced to balance representation capability and interpretability across different feature types. To further improve model performance, residual connections were incorporated to facilitate gradient propagation during training, and a channel attention mechanism was introduced to dynamically emphasize critical image features. These architectural designs enhanced the model's robustness and adaptability to complex underground mining environments. Experimental results demonstrate that the proposed model achieves superior performance on the benchmark dataset, with a mean squared error (EMSE) of 0.90 mg²/m⁶, a mean absolute error (EMAE) of 0.74 mg/m³, and a mean relative error (EMRE) of 1.76%. The coefficient of determination (R²) reaches 0.916 0, significantly outperforming comparative models such as Support Vector Regression and Random Forest. Ablation experiments further confirm the effectiveness of the multi-kernel fusion strategy, residual architecture, and attention mechanism in improving prediction accuracy. Overall, this study achieves a unified framework that balances model interpretability and recognition accuracy, providing a reliable, transparent, and high-precision solution for intelligent dust concentration in coal mines. The proposed approach offers important practical value for enhancing mine safety monitoring and risk prevention.
NIEBaisheng, ZHANGHao, GONGJie, et al. Macro and meso four-stage reaction mechanism of pulverized coal explosion[J]. Journal of China University of Mining & Technology, 2023, 52(6): 1129-1145.
[3]
WATSONJ G, TROPPR J, KOHLS D, et al. Filter processing and gravimetric analysis for suspended particulate matter samples [J]. Aerosol Science and Engineering, 2017, 1(2): 93-105.
[4]
LIUX, HUANGY, WANGZ, et al. On-line measurement of ultralow mass concentration particulate based on light scattering coupled with beta ray attenuation method[J]. Fuel, 2022, 329: 125461.
ZHOUFubao, ZHENGLina, FENGZikang, et al. Development of a miner personal dust continuous monitor based on the principle of oscillation balance [J]. Journal of China Coal Society, 2024, 49(2): 876-884.
[7]
YANGX H, YUH M, WANGY H, et al. Study on the modularized airflow-diverging system and dust-control technology in the fully mechanized heading face [J]. Powder Technology, 2022, 402: 117325.
CAICunliang, WANGYang, FENGJinliang, et al. Optical fiber sensor for the coal mine dust mass concentration in well testing[J]. Optical Instruments, 2017, 39(2): 10-15.
[10]
CHENJ, NIEZ, ZHAOF, et al. Improving the stability of electrostatic induction dust concentration detection using kalman filtering algorithm aided by machine learning [J]. Process Safety and Environmental Protection, 2023, 174: 882-890.
CHENChangjiang, YANGXiaoming, LITianqian, et al. Detection of concentration micro-SF6 and FPGA implementation by relative time‑difference ultrasonic method[J]. Journal of Xihua University (Natural Science Edition), 2014, 33(3): 54-56.
SUNJiping, LIXiaowei. Research on identification and anti‑interference method of mine external fire based on image smoothness [J]. Journal of China University of Mining & Technology, 2025, 54(1): 215-226.
[15]
XANGZ, LID Y, ZHENGX, et al. A novel coal dust characteristic extraction to enable particle size analysis [J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 1-12.
[16]
YANGB Q, CHENQ. PM2.5 concentration estimation based on image quality assessment[C]//4th IAPR Asian Conference on Pattern Recognition (ACPR), Nanjing, China. New York: IEEE, 2017.
TULuyang, CHENQinghua, CHENGYingsong, et al. Research on dust indentification and concentration detection method based on machine vision [J]. Safety in Coal Mines, 2025, 56(8): 48-58.
[19]
YINJ J, LEIJ Y, FANK X, et al. Integrating image processing and deep learning for effective analysis and classification of dust pollution in mining processes[J]. International Journal of Coal Science & Technology, 2023, 10(4): 17-33.
[20]
LUOZ Y, HUANGF F, LIUH. PM2.5 concentration estimation using convolutional neural network and gradient boosting machine[J]. Journal of Environmental Sciences, 2020, 31(1): 19-28.
[21]
OBREGÓNL, VELÁZQUEZC. Discrimination limit between mean gray values for the prediction of powder concentrations[J]. Powder Technology, 2007, 175(1): 8-13.
[22]
WANGS F, YINJ J, LIANGY T, et al. Dust pollution evaluation based on grayscale average and fractal dimension of digital image[J]. Journal of Cleaner Production, 2022, 379: 134691.
[23]
YANGP, CHENL, TIANJ, et al. Dust particle detection in surveillance video using salient visual descriptors[J]. Computers & Electrical Engineering, 2017, 62: 224-231.
[24]
LIG H, WUJ P, LUOZ W, et al. Vision-based measurement of dust concentration by image transmission[J]. IEEE Transactions on Instrumentation and Measurement, 2019, 68(10): 3942-3949.
[25]
BAIX S, ZHANGN, CAOX Y, et al. Prediction of PM2.5 concentration based on a CNN-LSTM neural network algorithm[J]. PeerJ, 2024, 12: e17811.
[26]
LIL F, GIRGUISM, LURMANNF, et al. Ensemble-based deep learning for estimating PM2.5 over California with multisource big data including wildfire smoke[J]. Environment International, 2020, 145: 106143.
ZHANGFan, ZHANGJiarong, CHENGHaixing. Research review on intelligent object detection technology for coal mines based on deep learning [J]. Coal Science and Technology, 2025, 53(Sup 1): 284-296.
[29]
ARJOMANDIH, KHEIRALIPOURK, AMARLOEIA. Estimation of dust concentration by a novel machine vision system[J]. Scientific Reports, 2022, 12(1): 1-10.
GONGXiaoyan, FENGHao, FUHaoran, et al. A method for detecting dust particles in excavation working face based on image analysis[J]. Journal of Mine Automation, 2024, 50(4): 55-62.
DINGZhen, LIHaodang, ZHANGQinghua. Research on intelligent hazard early warning architecture and key technologies for coal mine[J]. Journal of Mine Automation, 2023, 49(4): 15-22.
[34]
HOUDOUA, EIB I, KHOMSIK, et al. Interpretable machine learning approaches for forecasting and predicting air pollution: A systematic review[J]. Aerosol and Air Quality Research, 2024, 24(1): 1-25.
ZHAOMeng, WEIYuzhong, LIZheng, et al. Defogging algorithm for coal mining face dust and fog images based on a semi-supervised network[J]. Coal Science and Technology, 2025, 53(Sup 1): 346-353.
GAIYonggang. A defogging algorithm for coal mine underground images based on dark channel guided filtering and lighting correction [J]. Journal of Mine Automation, 2024, 50(6): 89-95.
[39]
WANGZ, YANGS, ZHANGJ, et al. Estimation of coal dust parameters via an effective image-based deep learning model[J]. Computers in Industry, 2025, 164: 104200.
[40]
GRAVETTERF J, WALLNAUL B. Statistics for the behavioral sciences [M]. Boston: Cengage Learning, 2016: 768.
YAOJuncai. Image quality assessment method based on contrast sensitivity visual characteristics of human vision system[J]. Chinese Journal of Liquid Crystals and Displays, 2011, 26(3): 390-396.
[43]
丰明坤. 基于视觉特性的图像质量综合评价方法研究[D]. 南京:南京邮电大学,2016.
[44]
FENGMingkun. Research on methods of image quality comprehensive assessment method based on visual characteristics [D]. Nanjing: Nanjing University of Posts and Telecommunications, 2016.
LUQingchang, ZHANGTu, WANGQin, et al. Comparison of accessibility indices for extreme weather events from spatiotemporal perspectives[J]. Journal of Zhejiang University (Engineering Science), 2024, 58(7): 1387-1396.
[47]
ABUS A H, ZAINALA, Al-RIMYB A S, et al. Intrusion detection in IoT using Gaussian fuzzy mutual information-based feature selection[J]. Engineering, Technology & Applied Science Research, 2024, 14(6): 17564-17571.
[48]
HARALICKR M, SHANMUGAMK, DINSTEINI H. Textural features for image classification[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1973, 3(6): 610-621.
[49]
GULATIA, FELAHATPISHEHA, VALDERRAMAC E. Feature engineering through two-level genetic algorithm[J]. Machine Learning with Applications, 2025, 21: 100696.