Quantitative analysis of coal particle size based on bi-level routing attention mechanism
-
摘要: 煤粒粒度分布特征与煤中甲烷气体传播规律的分析密切相关。目前，基于图像分割的煤粒粒度分析方法已成为获取煤粒粒度的主流方案之一，但存在上下文信息丢失、煤粒特征融合不当造成煤粒漏分割和过分割等问题。针对上述问题，设计了一种基于双层路由注意力机制（BRA）的煤粒粒度分析模型。在残差U型网络ResNet−UNet中嵌入BRA模块，得到B−ResUNet网络模型：为减少在煤粒分割过程中出现的漏分割问题，在ResNet−UNet网络的上采样前添加BRA模块，使网络根据上一层的特征调整当前特征层的重要性，增强特征的表达能力，提高长距离信息的传递能力；为减少在煤粒分割过程中出现的过分割问题，在ResNet−UNet网络的特征拼接模块后添加BRA模块，通过动态选择和聚合重要特征，实现更有效的特征融合。对分割出的煤粒进行特征信息提取，针对实验分析中采用的煤粒数据集的煤粒粒度与细胞大小相当，为精确表征煤粒粒度，采用等效圆粒径获取煤粒粒度及粒度分布。实验结果表明：① B−ResUNet网络模型的准确率、平均交并比、召回率较ResNet−UNet基础网络分别提高了0.6%，14.3%，35.9%，准确率达99.6%，平均交并比达92.6%，召回率达94.4%，B−ResUNet网络模型在煤样中具有较好的分割效果，能够检测出较为完整的颗粒结构。② 在上采样前和特征拼接后均引入BRA模块时，网络对煤粒的边缘区域给予了足够的关注，且对一些不太重要的区域减少了关注度，从而提高了网络的计算效率。③ 煤粒的粒度大小在1～2 mm内呈相对均衡的分布趋势，粒度在1～2 mm内的煤粒占比最大为99.04%，最小为90.59%，表明基于BRA的图像处理方法在粒度分析方面具有较高的准确性。Abstract: The distribution features of coal particle size are closely related to the analysis of methane gas propagation in coal. At present, the coal particle size analysis method based on image segmentation has become one of the mainstream solutions to obtain coal particle size. But there are problems such as loss of contextual information, improper fusion of coal particle features resulting in missed segmentation and over-segmentation of coal particles. In order to solve the above problems, a coal particle size analysis model based on bi-level routing attention (BRA) is designed. The BRA module is embedded in the residual U-shaped network ResNet-UNet to obtain the B-ResUNet network model. To reduce the problem of missed segmentation in coal particle segmentation, a BRA module is added before upsampling in the ResNet-UNet network. It allows the network to adjust the importance of the current feature layer based on the features of the previous layer, enhance the expression capability of features, and improve the transmission capability of long-distance information. To reduce the problem of over segmentation in coal particle segmentation, a BRA module is added after the feature concatenation module of the ResNet-UNet network. By dynamically selecting and aggregating important features, more effective feature fusion is achieved. The feature information from the segmented coal particles is extracted. The coal particle size of the coal particle dataset used in the experimental analysis is equivalent to the cell size. In order to accurately characterize the coal particle size, equivalent circular particle size is used to obtain the coal particle size and size distribution. The experimental results show the following points. ① The accuracy, average intersection to union ratio, and recall of the B-ResUNet network model have been improved by 06.%, 14.3%, and 35.9% compared to the ResNet-UNet basic network, with an accuracy of 99.6%, an average intersection to union ratio of 92.6%, and a recall of 94.4%. The B-ResUNet network model has good segmentation performance in coal samples and can detect relatively complete particle structures. ② When the BRA module is introduced before upsampling and after feature concatenation, the network pays sufficient attention to the edge areas of coal particles and reduces attention to some less important areas, thereby improving the computational efficiency of the network. ③ The particle size of coal particles shows a relatively balanced distribution trend within 1-2 mm, with the maximum proportion of coal particles within 1-2 mm being 99.04% and the minimum being 90.59%. It indicates that the image processing method based on BRA has high accuracy in particle size analysis.
-
图 6 不同网络模型的语义分割结果
Figure 6. Semantic segmentation results for the different network models
表 1 不同网络模型评价指标对比
Table 1. Comparison of the evaluation indexes of different network models
% 模型 准确率 平均交并比 召回率 PAN 98.3 62.8 27.8 PSPNet 98.4 66.6 36.9 U−Net 99.1 79.4 62.0 Link−Net 97.9 68.6 60.3 ResNet−UNet 99.0 78.3 58.5 B−ResUNet 99.6 92.6 94.4 
下载: 导出CSV
表 2 各网络模型性能
Table 2. Network performan
% 模型 准确率 平均交并比 召回率 ResNet−UNet 99.0 78.3 58.5 ResNet−采样BRA 99.4 87.2 79.0 ResNet−拼接BRA 99.2 82.6 66.5 B−ResUNet 99.6 92.6 94.4
下载: 导出CSV
表 3 不同方法测量粒度的准确率
Table 3. Accuracy of particle size measurement by different methods
% 测量方法 准确率 第1组 第2组 第3组 第4组 第5组 第6组 LPA方法 62.18 57.78 56.44 38.75 67.35 68.78 形态学方法 84.50 87.83 87.82 92.0 88.48 94.29 本文方法 97.42 97.37 89.80 95.56 96.47 96.15
下载: 导出CSV
-
[1] 邢震,韩安,陈晓晶,等. 基于工业互联网的智能矿山灾害数字孪生研究[J]. 工矿自动化,2023,49(2):23-30,55.XING Zhen,HAN An,CHEN Xiaojing,et al. Research on intelligent mine disaster digital twin based on industrial Internet[J]. Journal of Mine Automation,2023,49(2):23-30,55. [2] 张哲,魏晨慧,刘书源,等. 煤粒尺寸对气体扩散过程影响的数值模拟研究[J]. 矿业研究与开发,2021,41(7):85-92.ZHANG Zhe,WEI Chenhui,LIU Shuyuan,et al. Numerical simulation study of the influence of coal particle size on gas diffusion process[J]. Mining Research and Development,2021,41(7):85-92. [3] 马卫国,曾立,曾琦,等. 真空过滤数值模拟和试验验证[J]. 流体机械,2022,50(12):49-55.MA Weiguo,ZENG Li,ZENG Qi,et al. Numerical simulation and experimental verification of vacuum filtration[J]. Fluid Machinery,2022,50(12):49-55. [4] 李文凯,吴玉新,黄志民,等. 激光粒度分析和筛分法测粒径分布的比较[J]. 中国粉体技术,2007(5):10-13.LI Wenkai,WU Yuxin,HUANG Zhimin,et al. Measurement results comparison between laser particle analyzer and sieving method in particle size distribution[J]. China Powder Science and Technology,2007(5):10-13. [5] LIU Jingjing,CHENG Deqiang,LI Yunlong,et al. Quantitative evaluation of the influence of coal particle size distribution on gas diffusion coefficient by image processing method[J]. Fuel,2022,314:122946. doi: 10.1016/j.fuel.2021.122946 [6] GUIDA G,VIGGIANI G M B,CASINI F. Multi-scale morphological descriptors from the fractal analysis of particle contour[J]. Acta Geotech,2020,15(5):1067-1080. doi: 10.1007/s11440-019-00772-3 [7] SU D,YAN W M. Prediction of 3D size and shape descriptors of irregular granular particles from projected 2D images[J]. Acta Geotech,2020,15(6):1533-1555. doi: 10.1007/s11440-019-00845-3 [8] LAI Zhengshou,CHEN Qiushi. Reconstructing granular particles from X-ray computed tomography using the TWS machine learning tool and the level set method[J]. Acta Geotech,2019,14(1):1-18. doi: 10.1007/s11440-018-0759-x [9] 程德强,钱建生,郭星歌,等. 煤矿安全生产视频AI识别关键技术研究综述[J]. 煤炭科学技术,2023,51(2):349-365.CHENG Deqiang,QIAN Jiansheng,GUO Xingge,et al. Review on key technologies of AI recognition for videos in coal mine[J]. Coal Science and Technology,2023,51(2):349-365. [10] 李颖,李秀宇,卢兆林,等. 基于深度学习的煤粉颗粒CT图像分割方法[J]. 计算机工程与设计,2022,43(8):2252-2259.LI Ying,LI Xiuyu,LU Zhaolin,et al. Coal particle CT image segmentation method based on deep learning[J]. Computer Engineering and Design,2022,43(8):2252-2259. [11] 徐江川,金国强,朱天奕,等. 基于深度学习U−Net模型的石块图像分割算法[J]. 工业控制计算机,2018,31(4):98-99,102.XU Jiangchuan,JIN Guoqiang,ZHU Tianyi,et al. Segmentation of rock images based on U-Net[J]. Industrial Control Computer,2018,31(4):98-99,102. [12] 王征,张赫林,李冬艳. 特征压缩激活作用下U−Net网络的煤尘颗粒特征提取[J]. 煤炭学报,2021,46(9):3056-3065.WANG Zheng,ZHANG Helin,LI Dongyan. Feature extraction of coal dust particles based on U-Net combined with squeeze and excitation module[J]. Journal of China Coal Society,2021,46(9):3056-3065. [13] ZHU Lei,WANG Xinjiang,KE Zhanghan,et al. BiFormer:vision transformer with bi-level routing attention[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition,Vancouver,2023:10323-10333. [14] 梅昕苏. 基于多粒度Top−k查询的流式数据事件获取方法研究[D]. 沈阳:辽宁大学,2019.MEI Xinsu. Research on streaming data event acquisition method based on multi-granular Top-k query[D]. Shenyang:Liaoning University,2019. [15] LIU Huajie,XU Ke. Recognition of gangues from color images using convolutional neural networks with attention mechanism[J]. Measurement,2023,206:1-13. [16] 伊建峰,黎思成,吕珊,等. 基于频域数据增强及YOLOv7的动火作业检测模型[J]. 计算机应用,2023,43(增刊2):285-290.YI Jianfeng,LI Sicheng,LYU Shan,et al. Hot work detection model based on frequency domain data enhancement and YOLOv7[J]. Journal of Computer Applications,2023,43(S2):285-290. [17] ZHOU Buzhuang,YANG Shengqiang,JIANG Xiaoyuan,et al. Experimental study on oxygen adsorption capacity and oxidation characteristics of coal samples with different particle sizes[J]. Fuel,2023,331. DOI: 10.1016/J.FUEL.2022.125954. [18] MIYAKAWA T,TAKETANI F,TOBO Y,et al. Measurements of aerosol particle size distributions and INPs over the Southern Ocean in the late austral summer of 2017 on board the R/V Mirai:importance of the marine boundary layer structure[J]. Earth and Space Science,2023,10(3). DOI: 10.1029/2022EA002736. [19] TANG Songlei,LIU Qiang,TANG Hong,et al. Study on the movement of pulverized coal particles in fractal fracture network[J]. ACS Omega,2023. DOI: 10.1021/acsomega.3c02902. [20] REN Sucheng,ZHOU Daquan,HE Shengfeng,et al. Shunted self-attention via multi-scale token aggregation[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition,New Orleans,2022:10853-10862. [21] LIU Jingjing,LIU Ruihang,ZHANG Haoxiang,et al. Fast image processing method for coal particle cluster box dimension measurement and its application in diffusion coefficient testing[J]. Fuel,2023,352. DOI: 10.1016/J.FUEL.2023.129050. [22] RUSSELL B,TORRALBA A,MURPHY K P,et al. LabelMe:a database and web-based tool for image annotation[J]. International Journal of Computer Vision,2008,77(1/2/3):157-173. [23] LIU Shu,QI Lu,QIN Haifang,et al. Path aggregation network for instance segmentation[C]. IEEE/CVF Conference on Computer Vision and Pattern Recognition,Salt Lake City,2018:8759-8768. [24] ZHU Xiliang,CHENG Zhaoyun,WANG Sheng,et al. Coronary angiography image segmentation based on PSPNet[J]. Computer Methods and Programs in Biomedicine,2021. DOI: 10.1016/J.CMPB.2020.105897. [25] DU Getao,GAO Xu,LIANG Jimin,et al. Medical image segmentation based on U-net:a review[J]. Journal of Imaging Science and Technology,2020,64(2). DOI: 10.2352/J.ImagingSci.Technol.2020.64.2.020508. [26] CAI Junxiong,MU Taijiang,LAI Yukun,et al. LinkNet:2D-3D linked multi-modal network for online semantic segmentation of RGB-D videos[J]. Computers & Graphics,2021,98:37-47. [27] QIN Jiayin,SUN Yibo,WU Luji. Research on gear surface damage identification based on the ResNet Network[C]. The 2nd International Conference on Mechanical Automation and Electronic Information Engineering,Guizhou,2023. DOI: 10.1088/1742-6596/2419/1/012090. -
点击查看大图
图(8) / 表(3)
计量
- 文章访问数: 110
- HTML全文浏览量: 26
- PDF下载量: 22
- 被引次数: 0
