基于密度感知引导的煤泥浮选泡沫分割方法

陈贵震; 王红艳; 刘鑫; 熊峰; 邹国锋; 代伟

doi:10.16383/j.aas.c250607

基于密度感知引导的煤泥浮选泡沫分割方法

doi: 10.16383/j.aas.c250607 cstr: 32138.14.j.aas.c250607

陈贵震^1,,
王红艳^1,,
刘鑫^1,,
熊峰^1,,
邹国锋^2,,
代伟^1,

1.
中国矿业大学信息与控制工程学院徐州 221116
2.
山东理工大学电气与电子工程学院北京 100190

基金项目: 国家自然科学基金 (62373361, 52504313), 江苏省研究生科研与实践创新计划 (KYCX25_2842), 江苏省杰出青年基金(BK20240102), 中国矿业大学研究生创新计划 (2025WLKXJ106) 资助

详细信息

作者简介:
陈贵震：中国矿业大学信息与控制工程学院博士研究生. 主要研究方向为复杂工业过程图像信息处理. E-mail: chen_cumt@cumt.edu.cn

王红艳：中国矿业大学信息与控制工程学院讲师. 主要研究方向为复杂工业过程智能建模, 智能检测与智能控制. E-mail: wanghongyan@cumt.edu.cn

刘鑫：中国矿业大学信息与控制工程学院教授. 主要研究方向为系统辨识, 数据驱动的工业建模和软测量. E-mail: liuxin_hit@cumt.edu.cn

熊峰：中国矿业大学信息与控制工程学院硕士研究生. 主要研究方向为基于深度学习的工业视觉系统. E-mail: ts25060090a31ld@cumt.edu.cn

邹国锋：山东理工大学电气与电子工程学院副教授. 主要研究方向为低压电气故障智能检测与电力视觉技术. E-mail: gfzou@sdut.edu.cn

代伟：中国矿业大学信息与控制工程学院教授. 主要研究方向为复杂工业过程建模、运行优化与控制. 本文通信作者. E-mail: weidai@cumt.edu.cn

计量
- 文章访问数: 18
- HTML全文浏览量: 11
- 被引次数: 0
出版历程
- 收稿日期: 2025-11-07
- 录用日期: 2026-05-13
- 网络出版日期: 2026-06-09

Coal Flotation Foam Segmentation Method With Density-aware Guidance

1.
School of Information and Control Engineering, China Uni-versity of Mining and Technology, Xuzhou 221116
2.
School of Electrical and Electronic Engineering, Shandong University of Technology, Zibo 255049

Funds: Supported by National Natural Science Foundation of China (62373361, 52504313), the Postgraduate Research & Practice Innovation Program of Jiangsu Province (KYCX25_2842), Distinguished Young Scholars of Jiangsu Province (BK20240102), and the Graduate Innovation Program of China University of Mining and Technology (2025WLKXJ106)

More Information

Author Bio:
CHEN Gui-Zhen　Ph.D. candidate at the School of Information and Control Engineering, China University of Mining and Technology. His main research interest is image information processing for complex industrial processes

WANG Hong-Yan　Lecturer at the School of Information and Control Engineering, China University of Mining and Technology. Her research interests include intelligent modeling of complex industrial processes, intelligent detection, and intelligent control

LIU Xin　Professor at the School of Information and Control Engineering, China University of Mining and Technology. His research interests include system identification, data-driven industrial modeling, and soft sensing

XIONG Feng　Master student at the School of Information and Control Engineering, China University of Mining and Technology. His main research interest is deep learning-based industrial vision system

ZOU Guo-Feng　Associate professor at the School of Electrical and Electronic Engineering, Shandong University of Technology. His research interests include intelligent detection of low-voltage electrical faults and power vision technology

DAI Wei　Professor at the School of Information and Control Engineering, China University of Mining and Technology. His research interests include modeling, operational optimization, and control of complex industrial processes. Corresponding author of this paper

摘要

摘要: 针对浮选精煤泡沫分割中因低质图像数据导致的目标漏检及误分割问题, 提出一种基于密度感知引导的煤泥浮选泡沫分割方法. 首先, 构建跨尺度区域密度感知模块, 设计层次化密度估计子模块提取多尺度差异化区域密度信息, 并提出基于全局语义引导的跨尺度聚合方式, 融合生成具有区域差异感知能力的密度引导表征. 其次, 设计基于密度引导的分支增强模块, 建立基于分布感知的动态密度注意力机制, 构成以密度分布为先验、动态调节双分支空间特征响应的增强策略, 降低泡沫漏检率. 最后, 设计基于密度引导的互信息约束优化模块, 提出以互信息最大化为目标的语义耦合策略, 形成强化密度与分割表征间统计依赖的联合优化方法, 提升泡沫边界的分割判别能力. 在两个实际浮选泡沫数据集上的实验结果表明, 所提方法有效提升了泡沫分割性能.
- 浮选精煤泡沫 /
- 图像分割 /
- 密度感知 /
- 动态密度注意力 /
- 互信息约束
Abstract: To address the challenges of missed detection and mis-segmentation of targets in clean coal flotation foam segmentation caused by low-quality image data, this paper proposes a coal flotation foam segmentation method with density-aware guidance. First, a cross-scale regional density perception module is constructed, in which a hierarchical density estimation submodule is designed to extract multi-scale differential regional density information. Furthermore, a global semantic-guided cross-scale aggregation approach is proposed to integrate and generate density-guided representation with regional difference awareness capability. Second, a density-guided branch enhancement module is designed, in which a distribution-aware dynamic density attention mechanism is established to construct an enhancement strategy that uses density distribution as a prior to dynamically regulate the spatial feature responses of dual branches, thereby reducing the missed detection rate of foam. Finally, a density-guided mutual information constraint optimization module is designed, introducing a semantic coupling strategy based on mutual information maximization to jointly optimize and strengthen the statistical dependence between density and segmentation representation, improving the discriminative ability of foam boundary segmentation. Experimental results on two real flotation foam datasets demonstrate that the proposed method effectively improves foam segmentation performance.
- clean coal flotation foam /
- image segmentation /
- density-aware /
- dynamic density attention /
- mutual information constraint

HTML全文

图 1 浮选精煤泡沫形成原理、图像特征及分割特性分析图

Fig. 1 Analysis diagram of formation principle, image features, and segmentation characteristics of clean coal flotation froth

下载: 全尺寸图片幻灯片

图 2 基于密度感知引导的煤泥浮选泡沫分割网络整体架构图

Fig. 2 Overall architecture diagram of coal flotation foam segmentation network based on density-aware guidance

下载: 全尺寸图片幻灯片

图 3 跨尺度区域密度感知模块结构

Fig. 3 Structure of cross-scale regional density-aware module

下载: 全尺寸图片幻灯片

图 4 密度感知注意力结构

Fig. 4 Structure of density-aware attention

下载: 全尺寸图片幻灯片

图 5 浮精泡沫采集现场图

Fig. 5 Field image of clean coal foam collection

下载: 全尺寸图片幻灯片

图 6 数据标注示意图

Fig. 6 Illustration of data annotation

下载: 全尺寸图片幻灯片

图 7 参数调优结果图

Fig. 7 Result plot of parameter tuning

下载: 全尺寸图片幻灯片

图 8 不同数据集上模型训练验证曲线图

Fig. 8 Plot of model training and validation curves on different datasets

下载: 全尺寸图片幻灯片

图 9 可视化分割对比图

Fig. 9 Diagram of visual segmentation comparison

下载: 全尺寸图片幻灯片

图 10 泡沫密度估计结果图

Fig. 10 Diagram of foam density estimation results

下载: 全尺寸图片幻灯片

图 11 泛化性验证可视化分割结果图

Fig. 11 Diagram of visual segmentation results for generalization validation

下载: 全尺寸图片幻灯片

表 1 各模块消融实验验证结果(%)

Table 1 Ablation experiment verification results of each module (%)

基线	MA	MB	MC	Froth-Plant1		Froth-Plant2
基线	MA	MB	MC	AP	mIoU	AP	mIoU
√	—	—	—	63.27	50.36	73.36	57.20
√	√	—	—	65.14	51.94	74.49	57.58
√	√	√	—	67.84	52.32	75.88	58.03
√	√	—	√	66.52	53.55	75.01	58.62
√	√	√	√	68.27	53.78	76.15	58.84

下载: 导出CSV

表 2 与主流方法的对比实验结果

Table 2 Comparative experimental results with mainstream methods

方法类别	方法名称	Froth-Plant1			Froth-Plant2
方法类别	方法名称	AP (%)	mIoU (%)	FPS	AP (%)	mIoU (%)	FPS
两阶段	Mask R-CNN^[25]	69.30	49.48	10.6	75.70	56.66	11.3
	Cascade Mask R-CNN^[26]	68.90	50.11	7.1	78.20	58.65	7.9
	Hybrid Task Cascade^[27]	69.40	50.92	5.2	77.30	54.65	5.7
	Mask DINO^[28]	67.61	53.36	10.7	72.84	54.96	12.1
	Contourformer^[29]	68.85	44.15	15.3	73.27	39.07	16.5
	Grounded SAM^[34]	56.55	53.12	0.1	68.71	54.80	0.2
单阶段	EmbedMask^[30]	61.30	45.67	18.2	75.10	51.99	19.4
	BoxInst^[36]	64.70	45.17	15.6	71.30	55.71	16.7
	SOLO^[37]	66.30	48.96	18.7	72.40	52.08	19.5
	SparseInst^[38]	65.90	52.27	20.5	74.20	57.82	20.7
	SimCIS^[32]	61.12	51.90	19.0	73.30	56.18	20.3
	FastSAM^[33]	69.18	52.39	22.3	73.26	57.12	23.6
	YOLACT^[31]	63.27	50.36	24.9	73.36	57.20	25.6
	本文方法	68.27	53.78	23.6	76.15	58.84	24.3

下载: 导出CSV

表 3 泛化性验证实验结果(%)

Table 3 Results of generalization validation experiments(%)

数据集	AP	mIoU
PanNuke	73.88	63.33
bubble_size_distribution	64.76	47.77
BubbleBench	64.01	51.96

下载: 导出CSV

参考文献(38)

[1]	Polat M, Polat H, Chander S. Physical and chemical interactions in coal flotation. International Journal of Mineral Processing, 2003, 72(1–4): 199−213 doi: 10.1016/s0301-7516(03)00099-1
[2]	王红艳, 孙卓琪, 王兰豪, 南静, 代伟. 基于多源信息融合的浮选精煤灰分智能检测方法. 煤炭科学技术, 2025, 53(12): 280−292 doi: 10.12438/cst.2025-0723 Wang Hong-Yan, Sun Zhuo-Qi, Wang Lan-Hao, Nan Jing, Dai Wei. Intelligent detection method for flotation clean coal ash based on multi-source information fusion. Coal Science and Technology, 2025, 53(12): 280−292 doi: 10.12438/cst.2025-0723
[3]	Han N, Li Y F, Zhang Z Y, Gong J B, Zhang B L, Li Y F. A Bayesian optimization stacked ensemble learning method: Predicting ash content in flotation cleaned coal using X-ray fluorescence oxides data. Minerals Engineering, 2026, 242: Article No. 110227 doi: 10.1016/j.mineng.2026.110227
[4]	Lu F C, Liu H Z, Lv W B. Deep correlation and precise prediction between static features of froth images and clean coal ash content in coal flotation: An investigation based on deep learning and maximum likelihood estimation. Measurement, 2024, 224: Article No. 113843 doi: 10.1016/j.measurement.2023.113843
[5]	Yang T, Zhang Z Y, Chen P, Gui D J, Li Y F. Study on the prediction of cleaning coal ash content based on principal component analysis and machine learning. International Journal of Coal Preparation and Utilization, 20251−17
[6]	赵洪伟, 谢永芳, 蒋朝辉, 徐德刚, 阳春华, 桂卫华. 基于泡沫图像特征的浮选槽液位智能优化设定方法. 自动化学报, 2014, 40(6): 1086−1097 Zhao Hong-Wei, Xie Yong-Fang, Jang Zhao-Hui, Xu De-Gang, Yang Chun-Hua, Gui Wei-Hua. An intelligent optimal setting approach based on froth features for level of flotation cells. Acta Automatica Sinica, 2014, 40(6): 1086−1097
[7]	Massinaei M, Jahedsaravani A, Taheri E, Khalilpour J. Machine vision based monitoring and analysis of a coal column flotation circuit. Powder Technology, 2019, 343: 330−341 doi: 10.1016/j.powtec.2018.11.056
[8]	Tan J K, Liang L, Peng Y L, Xie G Y. Correction of froth gray value in the prediction of clean coal ash content in coal flotation. International Journal of Coal Preparation and Utilization, 2022, 42(9): 2742−2755 doi: 10.1080/19392699.2021.1900135
[9]	汪海洋, 潘德炉, 夏德深. 二维Otsu自适应阈值选取算法的快速实现. 自动化学报, 2007(9): 968−971 Wang Hai-Yang, Pan De-Lu, Xia De-Shen. A fast algorithm for two-dimensional Otsu adaptive threshold algorithm. Acta Automatica Sinica, 2007(9): 968−971
[10]	Brar K, Goyal B, Dogra A, Mustafa M, Majumdar R, Alkhayyat A, et al. Image segmentation review: Theoretical background and recent advances. Information Fusion, 2025, 114: Article No. 102608 doi: 10.1016/j.inffus.2024.102608
[11]	桂卫华, 阳春华, 徐德刚, 卢明, 谢永芳. 基于机器视觉的矿物浮选过程监控技术研究进展. 自动化学报, 2013, 39(11): 1879−1888 Gui Wei-Hua, Yang Chun-Hua, Xu De-Gang, Lu Ming, Xie Yong-Fang. Machine-vision-based online measuring and controlling technologies for mineral flotation: A review. Acta Automatica Sinica, 2013, 39(11): 1879−1888
[12]	Minaee S, Boykov Y, Porikli F, Plaza A, Kehtarnavaz N, Terzopoulos D. Image segmentation using deep learning: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 44(7): 3523−3542
[13]	路迈西, 王勇, 王凡, 刘文礼. 利用阈值分割技术提取煤泥浮选泡沫图像的物理特征. 煤炭科学技术, 2002(8): 34−37 Lu Mai-Xi, Wang Yong, Wang Fan, Liu Wen-Li. Threshold segmentation technology applied to get physical features of foam image from slime floatation. Coal Science and Technology, 2002(8): 34−37
[14]	Shilov R E, Logunova O S, Lednov A V. Methods and algorithms for segmentation of the image in control flotation process. In: Proceedings of the International Russian Automation Conference. Sochi, Russia: IEEE, 2018. 1–4
[15]	Jahedsaravani A, Massinaei M, Marhaban M H. An image segmentation algorithm for measurement of flotation froth bubble size distributions. Measurement, 2017, 111: 29−37 doi: 10.1016/j.measurement.2017.07.023
[16]	Jia R D, Ren M X, Lang D, Zheng J, He D K, Yu F. A semantic segmentation-based algorithm for fast flotation bubble size distribution measurement. Chemical Engineering Research and Design, 2024, 208: 795−807 doi: 10.1016/j.cherd.2024.07.041
[17]	Zeng T, Wu B, Ji S W. DeepEM3D: Approaching human-level performance on 3D anisotropic EM image segmentation. Bioinformatics, 2017, 33(16): 2555−2562 doi: 10.1093/bioinformatics/btx188
[18]	罗会兰, 黎宵. 基于上下文和浅层空间编解码网络的图像语义分割方法. 自动化学报, 2022, 48(7): 1834−1846 Luo Hui-Lan, Li Xiao. Image semantic segmentation method based on context and shallow space encoder-decoder network. Acta Automatica Sinica, 2022, 48(7): 1834−1846
[19]	Yu C Q, Wang J B, Peng C, Gao C X, Yu G, Sang N. Bisenet: Bilateral segmentation network for real-time semantic segmentation. In: Proceedings of the European Conference on Computer Vision. Munich, Germany: Springer, 2018. 334–349
[20]	Strudel R, Garcia R, Laptev I, Schmid C. Segmenter: Transformer for semantic segmentation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. Montreal, Canada: IEEE, 2021. 7242–7252
[21]	Wan Z F, Zhang P P, Wang Y H, Yong S L, Stepputtis S, Sycara K, et al. Sigma: Siamese mamba network for multi-modal semantic segmentation. In: 2025 IEEE/CVF Winter Conference on Applications of Computer Vision. Waikoloa, USA: IEEE, 2025. 1734–1744
[22]	Xu Z Z, Wu D Y, Yu C Q, Chu X X, Sang N, Gao C X. Sctnet: Single-branch CNN with Transformer semantic information for real-time segmentation. In: Proceedings of the AAAI Conference on Artificial Intelligence. Vancouver, Canada: AAAI Press, 2024. 38(6): 6378–6386
[23]	Wu R K, Liu Y H, Ning G C, Liang P C, Chang Q. UltraLight VM-UNET: Parallel vision mamba significantly reduces parameters for skin lesion segmentation. Patterns, 2025, 6(11): Article No. 101298 doi: 10.1016/j.patter.2025.101298
[24]	Fu Y X, Lou M, Yu Y Z. SegMAN: Omni-scale context modeling with state space models and local attention for semantic segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, USA: IEEE, 2025. 19077–19087
[25]	He K M, Gkioxari G, Dollár P, Girshick R. Mask R-CNN. In: Proceedings of the IEEE International Conference on Computer Vision. Venice, Italy: IEEE, 2017. 2961–2969
[26]	Cai Z, Vasconcelos N. Cascade R-CNN: High quality object detection and instance segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 43(5): 1483−1498
[27]	Chen K, Pang J M, Wang J Q, Xiong Y, Li X X, Sun S Y, et al. Hybrid task cascade for instance segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, CA, USA: IEEE, 2019. 4974–4983
[28]	Li F, Zhang H, Xu H Z, Liu S L, Zhang L, Ni L M, et al. Mask DINO: Towards a unified Transformer-based framework for object detection and segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver, Canada: IEEE, 2023. 3041–3050
[29]	Yao W W, Li C, Xiong M J, Dong W B, Chen H, Xiao X. ContourFormer: Real-time contour-based end-to-end instance segmentation transformer. In: 2025 International Joint Conference on Neural Networks. Rome, Italy: IEEE, 2025. 1–8
[30]	Ying H, Huang Z J, Liu S, Shao T J, Zhou K. EmbedMask: Embedding coupling for one-stage instance segmentation. In: Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence. Montreal, Canada: IJCAI, 2021. 1266–1273
[31]	Bolya D, Zhou C, Xiao F Y, Lee Y J. YOLACT: Real-time instance segmentation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. Seoul, Republic of Korea: IEEE, 2019. 9157–9166
[32]	Zhu Y C, Shi C, Wang D Y, Tang J J, Wei Z X, Wu Y, et al. Rethinking query-based transformer for continual image segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, USA: IEEE, 2025. 4595–4606
[33]	Zhao X, Ding W C, An Y Q, Du Y L, Yu T, Li M, et al. Fast segment anything. arXiv preprint arXiv: 2306.12156, 2023.
[34]	Ren T H, Liu S L, Zeng A L, Lin J, Li K C, Cao H, et al. Grounded sam: Assembling open-world models for diverse visual tasks. arXiv preprint arXiv: 2401.14159, 2024.
[35]	曹斌芳, 谢永芳, 阳春华, 桂卫华, 王晓丽. 基于多源数据的铝土矿浮选生产指标集成建模方法. 控制理论与应用, 2014, 31(9): 1252−1261 Cao Bin-Fang, Xie Yong-Fang, Yang Chun-Hua, Wang Xiao-Li. Integrated modeling for production index of bauxite flotation based on multi-source data. Control Theory & Applications, 2014, 31(9): 1252−1261
[36]	Tian Z, Shen C H, Wang X L, Chen H. BoxInst: High-performance instance segmentation with box annotations. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Virtual Event: IEEE, 2021. 5443–5452
[37]	Wang X L, Kong T, Shen C H, Jiang Y N, Li L. SOLO: Segmenting objects by locations. In: Proceedings of the European Conference on Computer Vision. Glasgow, UK: Springer, 2020. 649–665
[38]	Cheng T H, Wang X G, Chen S Y, Zhang W Q, Zhang Q, Huang C, Zhang Z X, Liu W Y. Sparse instance activation for real-time instance segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans, USA: IEEE, 2022. 4433–4442