基于卷积神经网络的显微煤粒分类算法
金厚鑫,曹乐*,阚秀,孙维周
(1.上海工程技术大学电子电气工程学院,上海 201620;2.安徽工业大学冶金工程学院,安徽 马鞍山 243002)
摘 要 煤岩组分的精确表征直接关系到煤炭的质量以及工艺性能的界定。传统的煤岩表征技术过程复杂,鲁棒性不够,严重阻碍了煤岩自动化分析系统的发展。针对上述问题,本文提出了一种结合多阶段注意力与模型融合的卷积神经模型来实现显微煤粒的自动化分类。该方法先使用语义分割算法、颗粒分离算法等预处理手段提取出单独的显微煤粒,接着采用多阶段注意力机制和多分支模型融合的卷积神经网络模型实现显微煤粒的分类。此外,本文在模型训练过程中使用了迁移学习的方法大幅提升了模型的训练效率。实验结果表明,本文方法的精度高达99.12%,可应用于煤岩显微自动化分析系统中显微煤粒的提取和分类。
关键词 显微煤粒;自动化;深度学习;注意力机制;分支融合;EfficientNet
中图分类号:TQ52;TP391.41;TP183 文献标识码:A doi:10.3969/j.issn.1000-6281.2022.02.006
Microscopic coal grain classification algorithm based on convolutional neural network
JIN Hou-xin1,CAO Le1*,KAN Xiu1,SUN Wei-zhou2
(1. School of Electronic and Electrical Engineering, School of Electronic and Electrical Engineering, Shanghai 201620;2. School of Metallurgical Engineering,Anhui University of Technology, Ma'anshan Anhui 243002, China)
Abstract The accurate characterization of the microstructure of industrial coal is directly related to the definition of coal quality and process performance. Traditional coal characterization techniques have complex processes and insufficient robustness, which seriously hinder the development of automated coal analysis systems. To address these problems, this paper proposes a convolutional neural model combining multi-stage attention and model fusion to achieve automated classification of microscopic coal particles. The method first extracts individual microscopic coal particles using pre-processing means such as semantic segmentation algorithm and particle separation algorithm, followed by a convolutional neural network model with multi-stage attention mechanism and multi-branch fusion to realize the classification of microscopic coal particles. In addition, we use the migration learning method in the model training process to significantly improve the training efficiency of the model. The experimental results show that the accuracy of the method can reach 99.12%, which can be applied to the extraction and classification of coal particles in the automatic coal particle analysis system.
Keywords coal maceral particles;automation;deep learning;attention mechanism;branch fusion;EfficientNet
“全文下载请到同方知网,万方数据库或重庆维普等数据库中下载!”
[1] ALPANA, CHAND S. Discovery of smart technology in manufacturing industries:automated characterization of coal using machine learning[C]//2020 International Conference on Smart Electronics and Communication (ICOSEC), India, 2020: 85-89.
[2] 宋孝忠, 张群. 煤岩显微组分组图像自动识别系统与关键技术[J]. 煤炭学报, 2019, 44(10): 3085-3097.
[3] 蒋敏, 奚峥皓. 基于自适应Gamma校正的煤岩显微图像增强研究[J]. 电子显微学报, 2020, 39(1): 46-52.
[4] ALPANA, MOHAPATRA S. Machine learning approach for automated coal characterization using scanned electron microscopic images[J]. Computers in Industry, 2016, 75(C):34-35.
[5] 张素新, 肖红艳. 煤储层中微孔隙和微裂隙的扫描电镜研究[J]. 电子显微学报, 2000, 19(4):531-532.
[6] 于冰, 卢兆林, 王帅,等. 基于EDS的煤中微细粒矿物相分布研究[J]. 电子显微学报, 2013, 32(3):244-249.
[7] 郭亮, 姚伯元.利用煤镜质组反射率分布图计算其镜质组含量[J]. 海南大学学报:自然科学版, 2015, 33(3):227-233.
[8] MLYNARCZUK M, SKIBA M. The application of artificial intelligence for the identification of the maceral groups and mineral components of coal[J]. Computers& Geosciences, 2017, 103(JUN): 133-141.
[9] TIWARY A K, GHOSH S, SINGH R, et al. Automated coal petrography using random forest[J]. International Journal of Coal Geology, 2020, 232: 103629.
[10] ZHANG Z L, YANG J G, WANG Y L, et al. Ash content prediction of coarse coal by image analysis and GA-SVM[J]. Powder Technology, 2014, 268:429-435.
[11] G O’BRIEN, JENKINS B, OFORI P, et al. Semi-automated petrographic assessment of coal by coal grain analysis[J]. Minerals Engineering, 2007, 20(5): 428-434.
[12] GHAZI M M, YANIKOGLU B, APTOULA E. Plant identification using deep neural networks via optimization of transfer learning parameters[J]. Neurocomputing, 2017, 235(APR.26): 228-235.
[13] TAN M, LE Q. EfficientNet:Rethinking model scaling for convolutional neural networks[C]//Proceedings of the International Conference on Machine Learning, Long Beach, New York, 2019: 6105-6114.
[14] ASTM D2799-13. Standard test method for microscopical determination of the maceral composition of coal, ASTM international[S]. West Conshohocken, PA: 2013.
[15] 曹乐, 金厚鑫, 阚秀, 等. 基于改进U-net网络的煤岩全组分提取方法及其应用[P]. 中国:CN113177965A, 2021-07-27.
[16] 刘伟华, 隋青美. 基于凹点搜索的重叠粉体颗粒的自动分离算法[J]. 电子测量与仪器学报, 2010, 24(12): 1095-1100.
[17] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. Las Vegas, America, 2016: 770-778.
[18] ZAGORUYKO S,KOMODAKIS N. Wide residual networks[C]//Proceedings of the British Machine Vision Conference (BMVC). Bristol, UK, 2016: 87.1-87.12.
[19] WOO S, PARK J, LEE J Y, et al. CBAM:Convolutional block attention module[C]//European Conference on Computer Vision (Eccv), Munich,2018:3-1,9.
[20] HU J, SHEN L, SUN G. Squeeze-and-excitation networks [C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018: 7132-7141.
[21] SANDLER M, HOWARD A, ZHU M, et al. Mobilenetv2: inverted residuals and linear bottlenecks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018: 4510-4520.
[22] 宋燕, 王勇. 多阶段注意力胶囊网络的图像分类[J]. 自动化学报, 2021, 47:1-14.
[23] WANG X, GIRSHICK R, GUPTA A, et al. Non-local neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake, USA, 2018:7794−7803.
[24] BUSLAEV A, PARINOV A, KHVEDCHENYA E, et al. Albumentations:fast and flexible image augmentations[J]. Information-an International Interdisciplinary Journal, 2020, 11(2):125.
[25] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet classification with deep convolutional neural networks[C]//Advances in Neural Information Processing Systems 25, 2012:1097-1105.
[26] SIMONYAN K, ZISSERMAN A. Very deep convolutional Networks for large-scale image recognition[J]. Computer Science, 2014, 1409-1556.
[27] SZEGEDY C, VANHOUCKE V, IOFFE S, et al. Rethinking the inception architecture for computer vision[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2016: 2818-2826.
[28] HUANG G, LIU Z, Van Der MAATEN L, et al. Densely connected convolutional networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 4700-4708.