刘丹丹¹， 郭持恒¹， 郭胜均²， 尚政国³， 颜鸽来²

1.黑龙江科技大学 电气与控制工程学院， 黑龙江 哈尔滨150022； 2.中煤科工集团重庆研究院有限公司， 重庆 400037；3.北京遥感设备研究所， 北京 100854

引用格式：

刘丹丹， 郭持恒， 郭胜均， 等. 基于深度学习的粉尘浓度等级检测算法［J］. 中国粉体技术， 2026， 32（4）： 1-10.

Liu Dandan， Guo Chiheng， Guo Shengjun， et al. Deep learning-based algorithm for dust concentration level detection［J］. China Powder Science and Technology， 2026， 32（4）： 1-10.

DOI：10.13732/j.issn.1008-5548.2026.04.012

收稿日期： 2026-01-10， 修回日期： 2026-05-11， 上线日期： 2026-06-04。

基金项目：国家自然科学基金项目，编号：62441306；黑龙江省省属本科高校基本科研业务费项目资助，编号：2024-KYYWF-1073。

第一作者：刘丹丹（1978—），女，教授，博士，博士生导师，研究方向为矿山安全监测与电气设备控制。E-mail：liudandan2003@163.com。

摘要：【目的】解决现有粉尘检测研究缺乏系统性浓度量化分析、难以满足粉尘浓度安全预警需求的问题，实现粉尘浓度等级的快速、准确检测。【方法】采用改进的粉尘浓度等级检测算法，对散射图像数据集进行检测；采用改进动态上采样、浓度感知特征增强、动态自适应标签分配3种核心方法，分别实现粉尘边缘重建、背景干扰抑制与多尺度匹配优化，进一步提升粉尘浓度等级检测精度。【结果】改进算法在目标测试集上取得的精确率为90.0%，召回率为86.7%，交并比的阈值为0.5时的平均精度均值为89.5%，有效解决粉尘浓度等级误检漏检问题。【结论】该方法可以实现从连续浓度值到离散等级的快速准确映射，避免传统方法逐像素反演计算量过大的弊端，证明将连续浓度测量转化为区间段浓度预警与判别的可行性。

关键词：粉尘检测；粉尘浓度等级识别；深度学习；特征增强

Abstract

Objective Dust is a typical hazardous byproduct in industrial production and fire scenarios， easily causing occupational health hazards， equipment wear， and even explosion risks. Accurate and efficient concentration detection is the core prerequisite for dust safety prevention and control. Traditional detection methods suffer from significant limitations in single-point measurement， high computational complexity of pixel-by-pixel concentration inversion， and poor real-time performance. Existing deep learning-based research mostly focuses on dust existence recognition and lacks systematic quantitative analysis of concentration. Meanwhile， restricted by the characteristics of small dust targets， low contrast， and blurred boundaries， they fail to meet the core requirements of on-site concentration safety early warning and risk quantitative assessment. Therefore， this paper carries out research on the optimization of dust concentration level detection algorithms.

Methods The YOLOv8.3.138 was adopted as the base detection model. First， a dust scattering image dataset was constructed using images captured by a charge-coupled device （CCD） industrial camera， and the images were divided into four risk levels according to relative concentration. After removing abnormal samples， the dataset was split into training， validation， and test sets at a ratio of 8：1：1. Meanwhile， a three-stage optimization strategy was proposed： a density-aware and edge-guided dynamic upsampling module was used to enhance fine-grained dust feature reconstruction； a concentration sensing feature enhancement （CAFE） module was employed to strengthen concentration-related feature responses； and a dynamic adaptive label assignment strategy （Dynamic ATSS） was adopted to optimize multi-scale target matching， thereby comprehensively improving the model detection performance.

Results and Discussion Validated by multiple sets of comparative experiments， the improved algorithm achieved a precision of 90.0%， a recall rate of 86.7%， and a mean average precision （mAP） of 89.5% at an intersection over union （IoU） threshold of 0.5 on the target test set. Ablation experiment results showed that the introduction of each of the three improved modules alone achieved stable performance improvement， and their joint use produced significant synergistic gains. Compared with the original YOLOv8 baseline model， the four core metrics were improved by 7.4%， 4.0%， 4.6%， and 7.9%， respectively. The comprehensive performance was significantly superior to mainstream object detection algorithms such as RetinaNet， EfficientDet， and YOLOv5. With only a small increase in the number of parameters and computational cost， the improved algorithm effectively solved the problems of false detection and missed detection of dust concentration levels， greatly improved the classification discrimination accuracy in critical concentration regions， and clarified two engineering application strategies for abnormal sample processing， balancing detection accuracy in normal scenes with response capability in abnormal scenes.

Conclusion The improved algorithm proposed in this study can realize real-time processing of dust scattering images and rapid classification of concentration levels. It accomplishes accurate mapping from continuous concentration values to discrete risk levels， avoids the drawback of excessive computational load of traditional inversion methods， and verifies the feasibility of converting continuous concentration measurement into interval-based concentration early warning. This work can provide an efficient technical solution for the quantitative assessment of industrial dust risks. Future work could further enhance its engineering applicability through model lightweighting and edge deployment.

Keywords： dust detection； dust concentration level detection； deep learning； feature enhancement

参考文献（References）

［1］Dong Yudie， Luo Na， Geng Jie， et al. Research progress on the technology of dust removal and individual protection for miners in coal mines： a review［J］. Journal of Environmental Engineering， 2026， 152（5）： 03126002.

［2］Nie Wen， Yang Bo， Du Tao， et al. Dynamic dispersion and high-rise release of coal dust in the working surface of a large-scale mine and application of a new wet dust reduction technology［J］. Journal of Cleaner Production， 2022， 351： 131356.

［3］Jiao Xiaoliang， Zhou Wei， Zhu Junpeng， et al. Study on dust hazard levels and dust suppression technologies in cabins of typical mining equipment in large open-pit coal mines in China［J］. Atmosphere， 2025， 16（4）： 461.

［4］Yuan A. Automatic dust detection， isolation， and mass measurement pipeline［J］. Research Notes of the AAS， 2026， 10（1）： 9.

［5］Wolfe C， Cauda E， Yekich M， et al. Real-time dust monitoring in occupational environments： a case study on using low-cost dust monitors for enhanced data collection and analysis［J］. Mining， Metallurgy & Exploration， 2024， 41（4）： 1709-1718.

［6］Chen Chao， Zheng Lina， Zhang Wenli， et al. Development and validation of an automatic filter-replacement coal dust concentration detector based on light extinction［J］. Measurement Science and Technology， 2025， 36（6）： 065801.

［7］柳小波， 范立鹏， 秦丽杰， 等. 机器视觉技术在矿山行业的应用现状与展望［J］. 有色金属（矿山部分）， 2025， 77（2）： 1-15.

Liu Xiaobo， Fan Lipeng， Qin Lijie， et al. Application status and prospects of machine vision technology in mining industry［J］. Nonferrous Metals （Mining Section）， 2025， 77（2）： 1-15.

［8］程学珍， 赵振国， 刘兴军， 等. 基于YOLOv4算法的煤矿井下粉尘检测方法［J］. 实验室研究与探索， 2022， 41（3）： 14-18.

Cheng Xuezhen， Zhao Zhenguo， Liu Xingjun， et al. Dust detection method for underground coal mines based on YOLOv4 algoritm［J］. Research and Exploration in Laboratory， 2022， 41（3）： 14-18

［9］邵安林， 刘畅， 岳星彤， 等. 新时期我国矿产资源行业高质量发展路径［J］. 金属矿山， 2024（1）： 2-6.

Shao Anlin， Liu Chang， Yue Xingtong， et al. Path of high-quality development of China’s mineral resources industry in the new era［J］. Metal Mine， 2024（1）： 2-6.

［10］曹萍萍. 矿井复杂环境下目标检测与识别关键技术研究［D］. 徐州： 中国矿业大学， 2024：3.

Cao Pingping. Study on key technologies object detection and recognition methods in complex mine environments［D］. Xuzhou： China University of Mining and Technology， 2024：3.

［11］王韦刚， 王炳蔚， 张云伟. TDFF：一种强鲁棒性的烟雾图像检测算法［J］. 激光与光电子学进展， 2021， 58（4）： 273-280.

Wang Weigang， Wang Bingwei， Zhang Yunwei. TDFF： strong robust algorithm for smoke image detection［J］. Laser & Optoelectronics Progress， 2021， 58（4）： 273-280.

［12］Li Xiuqing， Chen Zhenxue， Wu Qingming， et al. 3D parallel fully convolutional networks for real-time video wildfire smoke detection［J］. IEEE Transactions on Circuits and Systems for Video Technology， 2020， 30（1）： 89-103.

［13］Pan Jin， Ou Xiaoming， Xu Liang. A collaborative region detection and grading framework for forest fire smoke using weakly supervised fine segmentation and lightweight faster-RCNN［J］. Forests， 2021， 12（6）： 768.

［14］Khan S， Muhammad K， Hussain T， et al. DeepSmoke： deep learning model for smoke detection and segmentation in outdoor environments［J］. Expert Systems with Applications， 2021， 182： 115125.

［15］Wang Mingpu， Yao Gang， Yang Yang， et al. Deep learning-based object detection for visible dust and prevention measures on construction sites［J］. Developments in the Built Environment， 2023， 16： 100245.

［16］郑远攀， 许博阳， 王振宇. 改进的YOLOv5烟雾检测模型［J］. 计算机工程与应用， 2023， 59（7）： 214-221.

Zheng Yuanpan， Xu Boyang， Wang Zhenyu. Improved YOLOv5 smoke detection model［J］. Computer Engineering and Applications， 2023， 59（7）： 214-221.

［17］Wang Yue， Piao Yan， Wang Qi， et al. Early smoke recognition algorithm for forest fires［J］. Forests， 2024， 15（7）： 1082.

［18］刘丹丹， 侯展博， 胡丙阳， 等. 基于CCD图像的粉尘浓度面维度分布测量方法［J］.黑龙江科技大学学报， 2023， 33（1）： 99-102.

Liu Dandan， Hou Zhanbo， Hu Bingyang， et al. Surface dimension distribution measurement method of dust concentration based on CCD image［J］. Journal of Heilongjiang University of Science and Technology， 2023， 33（1）： 99-102.

［19］Pan Guanghao， Li Xuewei， Cheng Xiang， et al. MAS-YOLO： a fabric defect detection network based on YOLOv8［J］. Textile Research Journal， 2026， 96（5/6）： 562-580.

［20］Yan Guanghao， Li Xuewei， Cheng Xiang， et al. YOLOv8-ACU： improved YOLOv8-pose for facial acupoint detection［J］. Frontiers in Neurorobotics， 2024， 18： 1355857.

［21］Wang Xueqiu， Gao Huanbing， Jia Zemeng， et al. BL-YOLOv8： an improved road defect detection model based on YOLOv8［J］. Sensors， 2023， 23（20）： 8361.

［22］Liu Wenze， Lu Hao， Fu Hongtao， et al. Learning to upsample by learning to sample［C］//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2023： 6027-6037.

［23］Hu Jie， Shen Li， Albanie S， et al. Squeeze-and-excitation networks［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2020， 42（8）： 2011-2023.

［24］Ling Liuyi， Wei Guo， Liu Yuanwen， et al. RBPSD-YOLOv8： weld surface defect detection based on improved YOLOv8［J］. Journal of Real-time Image Processing， 2025， 22（4）：1-18.

［25］Huang Yuqin， Han Dezhi， Han Bing， et al. ADV-YOLO： improved SAR ship detection model based on YOLOv8［J］. The Journal of Supercomputing， 2025， 81（1）： 1-32.