Traditional rock slice image classification relies on a large number of manually labeled image samples, which is subject to the experience and ability of the labelers. This practice limits the scalability of classification enhancement as increasing unlabeled rock slice image samples does not contribute effectively. In order to achieve effective utilization of unlabeled data information, the hierarchy consistency mean teacher (HCMT) model adds a hierarchy consistency regularization term to the unsupervised loss of the mean teacher (MT) model to constrain the hierarchical structure of the teacher-student model. Ablation experiments and comparative analyses reveal that the introduction of hierarchy consistency regularization method improves the classification performance of the MT model by using the effective information of unlabeled data. As a result, the HCMT model achieves comparable classification capability in both half-labeled and fully labeled dataset. The experiments show the potential of the semi-supervised learning model to improve the classification ability of the model by using a large number of unlabeled rock slice image data.
[1] Klein C, Philpotts A. Earth materials: introduction to mineralogy and petrology [M]. 2ed. Cambridge: Cambridge University Press, 2017.
[2] 付光明, 严加永, 张昆, 等. 岩性识别技术现状与进展[J]. 地球物理学进展, 2017, 32(1): 26-40. Fu G M, Yan J Y, Zhang K, et al. Current status and progress of lithology identification technology [J]. Progress in Geophysics, 2017, 32(1): 26-40. (in Chinese)
[3] Mackenzie W, Adams A, Brodie K. Rocks and minerals in thin section [M]. 2ed. New York: CRC Press, 2017.
[4] 苏程, 朱孔阳. 岩石薄片图像智能分析研究进展[J]. 矿物岩石地球化学通报, 2023, 42(1): 13-25, 6. Su C, Zhu K Y. Research progress of intelligent image analysis for petrographic thin section images [J]. Bulletin of Mineralogy, Petrology and Geochemistry, 2023, 42(1): 13-25, 6. (in Chinese)
[5] Reichstein M, Camps-Valls G, Stevens B, et al. Deep learning and process understanding for data-driven earth system science [J]. Nature, 2019, 566(7743): 195-204.
[6] Lee D H. Pseudo-label: the simple and efficient semi-supervised learning method for deep neural networks [J]. Neural Networks, 2015, 64: 59-63.
[7] 刘烨, 吕锦涛. 基于超像素与半监督的岩石图像分割与识别[J]. 工程科学与技术, 2023, 55(2): 171-183. Liu Y, Lyu J T. Semi-supervised rock image segmentation and recognition based on superpixel [J]. Advanced Engineering Sciences, 2023, 55(2): 171-183. (in Chinese)
[8] Baykan N A, Yilmaz N. Mineral identification using color spaces and artificial neural networks [J]. Computers & Geosciences, 2010, 36(1): 91-97.
[9] Yann L C, Bengio Y, Hinton G. Deep learning [J]. Nature, 2015, 521(7553): 436-444.
[10] Verma V, Lamb A, Kannala J, et al. Interpolation consistency training for semi-supervised learning [C]//The Twenty-Eighth International Joint Conference on Artificial Intelligence, 2019: 3635-3641.
[11] Liu F B, Tian Y, Cordeiro F R, et al. Self-supervised mean teacher for semi-supervised chest X-ray classification [M]//Machine Learning in Medical Imaging. Cham: Springer, 2021.
[12] Saha S, Ebel P, Zhu X X. Self-supervised multisensor change detection [J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 60: 1-10.
[13] Perlaza S M, Bisson G, Esnaola I, et al. Empirical risk minimization with relative entropy regularization: optimality and sensitivity analysis [C]//IEEE International Symposium on Information Theory (ISIT), 2022: 684-689.
[14] Berthelot D, Carlini N, Goodfellow I, et al. MixMatch: a holistic approach to semisupervised learning [DB/OL]. 2019[2023-11-22]. https://arxiv.org/abs/1905.02249
[15] Park D S, Zhang Y, Jia Y, et al. Improved noisy student training for automatic speech recognition [C]//Interspeech 2020, 2020: 2817-2821.
[16] Dou Q, Yu L Q, Chen H, et al. 3D deeply supervised network for automated segmentation of volumetric medical images [J]. Medical Image Analysis, 2017, 41: 40-54.
[17] Zhao Z Y, Zeng Z, Xu K X, et al. DSAL: deeply supervised active learning from strong and weak labelers for biomedical image segmentation [J]. IEEE Journal of Biomedical and Health Informatics, 2021, 25(10): 3744-3751.
[18] 赖文, 蒋璟鑫, 邱检生, 等. 南京大学岩石教学薄片显微图像数据集[J]. 中国科学数据, 2020, 5(3): 26-38. Lai W, Jiang J X, Qiu J S, et al. Rock teaching thin section microscopic image dataset in Nanjing University [J]. Chinese Scientific Data, 2020, 5(3): 26-38. (in Chinese)
[19] 张世杰, 胡修棉. 新疆塔里木盆地西部晚白垩世-始新世岩石薄片偏光显微图像数据集[J]. 中国科学数据, 2020, 5(3): 59-69. Zhang S J, Hu X M. Polarized light micrograph dataset of Late Cretaceous-Eocene rock thin sections from western Tarim Basin, Xinjiang [J]. Chinese Scientific Data, 2020, 5(3): 59-69. (in Chinese)
[20] 韩中, 胡修棉. 西藏特提斯喜马拉雅早-中侏罗世岩石薄片偏光显微图像数据集[J]. 中国科学数据, 2020, 5(3): 107-115. Han Z, Hu X M. A photomicrograph dataset of Early-Middle Jurassic rocks in the Tibetan Tethys Himalaya [J]. Chinese Scientific Data, 2020, 5(3): 107-115. (in Chinese)
[21] 钱红杉, 邢凤存, 张春林, 等. 鄂尔多斯盆地中寒武统徐庄组岩石薄片显微图像数据集[J]. 中国科学数据, 2020, 5(3): 200-210. Qian H S, Xing F C, Zhang C L, et al. A dataset of microscope images of Middle Cambrian rock sections from Xuzhuang Formation in Ordos Basin [J]. China Scientific Data, 2020, 5(3): 200-210. (in Chinese)
[22] Cheng G J, Guo W H. Rock images classification by using deep convolution neural network [J]. Journal of Physics: Conference Series, 2017, 887: 012089.
[23] Seo W, Kim Y, Sim H, et al. Classification of igneous rocks from petrographic thin section images using convolutional neural network [J]. Earth Science Informatics, 2022, 15(2): 1297-1307.
