The Study of Coal Macerals Segmentation and Quantitative Analysis Based on MSR-UNet

doi:10.11871/jfdc.issn.2096-742X.2023.06.012

Abstract

Abstract:

[Objective] The utilization rate of coal macerals is mainly determined by the quality of coal, and the analysis of its composition is an important basis for judging the quality of coal. This paper proposes an improved UNet for the segmentation of coal maceral micro-components. The purpose is to improve the accuracy of the segmentation of coal micro-components so as to realize the automatic analysis of coal maceral micro-components. [Methods] Firstly, a multi-scale contextual attention module is proposed. It can improve the ability of the network to extract key features by capturing high-level features with spatial contextual information. Secondly, a squeeze and excitation module is introduced in the jump connection layer to improve the network's ability to capture important information about low-level features. Finally, the dice loss function and the focal loss function are selected to train the network to improve the sensitivity of the network to small target components and the ability to distinguish similar components. [Results] The experimental results show that the proposed method performs well in segmenting the microscopic component images of coal rocks with the PA, IoU, and Dice of 91.24%, 83.01%, and 84.70%, respectively. The mean absolute error for each component segmentation is 2.95%, 5.43%, and 6.19%, respectively. [Conclusions] The algorithm in this paper has great potential in realizing the use of computer-aided automatic analysis of coal macerals quality.

Key words: coal maceral micro-components, image segmentation, UNet, multiscale contextual attention, squeeze and excitation

JI Jingjing, XI Zhenghao, LI Zhongfeng. The Study of Coal Macerals Segmentation and Quantitative Analysis Based on MSR-UNet[J]. Frontiers of Data and Computing, 2023, 5(6): 126-137, https://cstr.cn/32002.14.jfdc.CN10-1649/TP.2023.06.012.

Figures/Tables 12

Fig.1

Fig.2

Fig.3

Fig.4

Table 1

Table 2

Fig.5

Table 3

Table 4

Fig.6

Table 5

Fig.7

References 28

[1]	陈浮, 于昊辰, 卞正富, 等. 碳中和愿景下煤炭行业发展的危机与应对[J]. 煤炭学报, 2021, 46(6): 1808-1820.
[2]	王晓磊, 陈贵锋, 李文博, 等. 双碳背景下煤炭清洁高效利用方向构建[J]. 煤质技术, 2021, 36(6):1-5.
[3]	CHEN Y, QIN Y, LI Z, SHI Q, et al. Differences in desorption rate and composition of desorbed gases between undeformed and mylonitic coals in the zhina coalfield, southwest china[J]. Fuel, 2019, 239: 905-916. doi: 10.1016/j.fuel.2018.11.085
[4]	杨兆中, 杨苏, 张健, 等. 800m以深直井煤储层压裂特征分析[J]. 煤炭学报, 2016, 41(1): 100-104.
[5]	宋孝忠. 烟煤显微组分组图像自动识别技术研究及应用[D]. 北京: 煤炭科学研究总院, 2020.
[6]	MLYNARCZUK M, SKIBA M. The application of artificial intelligence for the identification of the maceral groups and mineral components of coal[J]. Computers Geosci, 2017, 103: 133-141. doi: 10.1016/j.cageo.2017.03.011
[7]	WANG H, LEI M, CHEN Y, LI M, ZOU L. Intelligent identification of maceral components of coal based on image segmentation and classification[J]. Appl Sci, 2019, 9(16): 3245-3260. doi: 10.3390/app9163245
[8]	SKIBA M, MLYNARCZUK M. Identification of macerals of the inertinite group using neural classifiers, based on selected textural features[J]. Arch Min Sci, 2018, 63(4): 827-837.
[9]	赵伟昱, 张宏海, 仲波. 基于深度学习的遥感影像地块分割方法[J]. 数据与计算发展前沿, 2021, 3(2):133-141.
[10]	LONG J, SHELHAMER E, DARRELL T. Fully convolutional networks for semantic segmentation[C]// International Conference on Medical Image Computing and Computer-assisted Intervention. Boston, USA, New York: IEEE, 2015: 3431-3440.
[11]	RONNEBERGER O, FISCHER P, BROX T. U-net: Convolutional networks for biomedical image segmentation[C]// Proceedings of Medical Image Computing and Computer-assisted Intervention, Singapore: Springer, Cham, 2015: 234-241.
[12]	CHEN L C, PAPANDREOU G, KOKKINOS I, et al. Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 40(4): 834-848. doi: 10.1109/TPAMI.2017.2699184
[13]	KONG X, SUN G, WU Q, et al. Hybrid pyramid u-net model for brain tumor segmentation[C]// International Conference on Intelligent Information Processing, Barcelona, Spain, Singapore: Springer, 2018:346-355.
[14]	LI X, ZHONG Z, WU J, et al. Expectation-maximization attention networks for semantic segmentation[C]// Proceeding of the International Conference on Computer Vision, Seoul, Korea, New York: IEEE, 2019: 9167-9176.
[15]	ZHANG H, ZHANG H, WANG C, et al. Co-occurrent features in semantic segmentation[C]// Proceeding of the Conference on Computer Vision and Pattern Recognition, Seoul, Korea, New York: IEEE, 2019: 548-557.
[16]	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, USA, New York: IEEE, 2018: 7132-7141.
[17]	WOO S, PARK J, LEE J Y, et al. Cbam: Convolutional block attention module[C]// Proceedings of the European conference on computer vision, Munich, Germany, Singapore: Springer, 2018: 3-19.
[18]	RUNDO L, HAN C, NAGANO Y, et al. USE-Net: Incorporating Squeeze-and-Excitation blocks into UNet for prostate zonal segmentation of multi-institutional MRI datasets[J]. Neurocomputing, 2019, 365: 31-43. doi: 10.1016/j.neucom.2019.07.006
[19]	GU Z, CHENG J, FU H, et al. Ce-net: Context encoder network for 2d medical image segmentation[J]. IEEE transactions on medical imaging, 2019, 38(10): 2281-2292. doi: 10.1109/TMI.2019.2903562 pmid: 30843824
[20]	CHEN L C, PAPANDREOU G, KOKKINOS I, et al. Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFS[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 40(4): 834-848. doi: 10.1109/TPAMI.2017.2699184
[21]	MILLETARI F, NAYAB N, AHMADI S A. V-net: Fully convolutional neural networks for volumetric medical image segmentation[C]// Proceedings of 3D Vision, California, USA, New York: IEEE, 2016: 565-571.
[22]	ZHU W, HUANG Y, ZENG L, et al. Anatomynet: Deep learning for fast and fully automated whole-volume segmentation of head and neck anatomy[J]. Med Phys, 2019, 46(2): 576-89. doi: 10.1002/mp.13300 pmid: 30480818
[23]	LI X, WANG W, HU X, et al. Generalized Focal Loss V2: Learning Reliable Localization Quality Estimation for Dense Object Detection[C]// Proceedings of Computer Vision and Pattern Recognition, Nashville, USA, New York: IEEE, 2021: 11627-11636.
[24]	WILSON M A, PUGMIRE R J, KARAS J, et al. Carbon distribution in coals and coal macerals by cross polarization magic angle spinning carbon-13 nuclear magnetic resonance spectrometry[J]. Analytical Chemistry, 56(6): 933-943. doi: 10.1021/ac00270a018
[25]	CSURKA G, LARLUS D, PERRONNIN F. What is a good evaluation measure for semantic segmentation[C]// Proceeding of British Machine Vision Conference, London, England, London: the British Machine Vision Conference, 2013: 10-5244.
[26]	LONG J, SHELHAMER E, DARRELL T. Fully Convolutional Networks for Semantic Segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 39(4): 640-651. doi: 10.1109/TPAMI.2016.2572683
[27]	BADRINARAYANAN V, HANDA A, CIPOLLA R. Segnet: A deep convolutional encoder-decoder architecture for robust semantic pixel-wise labelling[J]. arXiv preprint arXiv: 1505.07293, 2015. https://arxiv.org/abs/1505.07293.
[28]	XU J, LIU J, ZHANG X, LING P, et al. Chemical imaging of coal in micro-scale with raman mapping technology[J]. Fuel, 2020, 264: 116826. doi: 10.1016/j.fuel.2019.116826

平台硬件与软件配置
显存	12G	框架	Pytorch1.7
CPU	Intel core i7-7500	Cuda	11.0
内存	2T	语言	Python3.7
GPU	NVIDIA Titan X	操作系统	Windows10 Professional

模型	PA	IoU	Dice
Kmeans	0.7687	0.6713	0.6895
two-step strategy	0.8062	0.7581	0.7614
FCN	0.8570	0.7609	0.7741
SegNet	0.8115	0.7224	0.7376
UNet	0.8863	0.7931	0.8065
MSR-UNet	0.9124	0.8301	0.8470

模型	PA	IoU	Dice
VGG-UNet	0.8987	0.8073	0.8125
Mobile-UNet	0.8962	0.7981	0.8016
Res-UNet	0.9076	0.8179	0.8241

模型	PA	IoU	Dice
Res-UNet	0.9076	0.8179	0.8241
Res-UNet+SE	0.9093	0.8224	0.8376
Res-UNet+MCA	0.9108	0.8267	0.8395
MSR-UNet	0.9110	0.8268	0.8419

L_D	L_F	PA	IoU	Dice
		0.9110	0.8268	0.8419
√		0.9117	0.8271	0.8422
	√	0.9119	0.8273	0.8425
√	√	0.9124	0.8301	0.8470