To address the classification challenge of rock thin section images due to large intra-class variations and high inter-class similarities, a dual-path classification model WD-DDPM that integrates local frequency domain features and global semantics was proposed. Based on ResNet-50, a multi-scale downsampling module (WTMDB) based on wavelet transform to replace the traditional downsampling operation was designed. By performing wavelet decomposition on the image, it retains the low-frequency contour information while enhancing the edge and texture details in the high-frequency sub-bands using an attention mechanism, effectively strengthening the model’s ability to extract local discriminative features. Additionally, a U-Net encoder from the diffusion model was introduced as a parallel path to extract global context features containing overall structure and long-range dependencies. Finally, by fusing the dual-path features of local details and global semantics, a complementary image representation was formed and sent to the classifier for lithology identification. The experimental results on a dataset containing 108 rock thin slice samples show that the accuracy of the proposed method reaches 97.5%, the recall rate is 94.7%, and the F1 score is 96.0%, which is significantly better than mainstream convolutional neural network models such as VGG16 and MobileNetV3 (under the same experimental conditions), verifying its superior performance and practical value in the complex rock image classification task.
DU Rui-shan, MU Wen-xuan, MENG Ling-dong.
Classification of Rock Thin Slice Images by Fusing Diffusion and Wavelet Residual Networks[J].
Science Technology and Engineering, 2026, 26(13): 5566-5573 DOI:10.12404/j.issn.1671-1815.2503159
LiY, LiJ, ShenY, et al. Application of thin slice technology of loose cuttings sample in early stage of experimental geology[J]. Journal of Physics: Conference Series, 2024, 2834(1): 012129.
[2]
KellyJ, HollisP S, DanaP D C, et al. Characterization of a metamorphosed volcanic stratigraphy and VMS alteration halos using rock chip petrography and lithogeochemistry: a case study from King North, Yilgarn Craton, western Australia[J]. Minerals, 2024, 14(5): 481.
[3]
BehnamE, SirousH, AliK, et al. Simulating reservoir capillary pressure curves using image processing and classification machine learning algorithms applied to petrographic thin sections[J]. Journal of African Earth Sciences, 2024, 209: 105098.
ZhouYueyang, XiaZhujun, LiChao, et al. Technology and application of wellsite rock thin section identification[J]. China Petroleum and Chemical Standard and Quality, 2021, 41(10): 180-181.
LiuMingwei, ZhenMiao, ZhangTiean, et al. Design and application of an auxiliary system for rock thin section identification based on VB6.0[J]. Experiment Science and Technology, 2025, 23(4): 54-59.
RenShujie, HuYong, HeWenxiang, et al. Sandstone component microscopic image recognition based on deep learning[J]. Science Technology and Engineering, 2024, 24(9): 3727-3736.
LiLinjuan, ZhangZhaoxiang, HeYun. A robust image classification algorithm based on domain adaptation[J]. Science Technology and Engineering, 2024, 24(36): 15558-15566.
CuiXintong, WangYing, DengZhennan, et al. Defogging method for diffusion models integrating parallel multi-convolution attention[J]. Computer Engineering and Design, 2025, 46(7): 2081-2088.
[14]
WangZ Y, LiD Y, ZhangM Y, et al. Enhancing hyperspectral images via diffusion model and group-autoencoder super-resolution network[J]. arXiv Preprint, arXiv: 2402.17285, 2024.
ZhouZhiqi, LiPeijun. Lithological classification of multispectral images using spectral-spatial clustering[J]. Acta Scientiarum Naturalium Universitatis Pekinensis, 2025, 61(3): 525-534.
ZhouChengyang, LiuWei, WuTianrun, et al. Rock thin section image classification based on a mixture of experts model[J]. Journal of Jilin University (Science Edition), 2024, 62(4): 905-914.
[19]
MaW, HanT, XuZ, et al. Feature fusion of single and orthogonal polarized rock images for intelligent lithology identification[J]. AI in Civil Engineering, 2025, 4(1): 5.
LiuJinzi, DuWenying, DongChi, et al. Intelligent classification and identification of complex lithologic reservoir rock samples based on color space and deep residual network ResNet-50[J]. Science Technology and Engineering, 2023, 23(29): 12628-12637.
LiTianping, LiGongquan. An automatic pore segmentation method for rock cast thin sections based on the integration of multiple deep learning models[J]. Science Technology and Engineering, 2023, 23(21): 9168-9175.
[24]
YuC, WuW, ZhengJ, et al. A multi-stage adaptive otsu thresholding algorithm for pore segmentation in rock thin-section images[J]. Earth Science Informatics, 2025, 18(2): 239.
ChengGuojian, LiZongxiang, LiQiushi, et al. Lithology identification method based on rock thin section images and improved EfficientNet modeling[J]. Journal of Xi’an Shiyou University (Natural Science Edition), 2025, 40(2): 124-134.
LiShunyong, LiQinghui, XingYuman. Intelligent rock recognition based on lightweight network and transfer learning[J]. Science Technology and Engineering, 2025, 25(5): 1774-1782.
[29]
ZhuX, LiM, LanZ, et al. RockNet: deep progressive lithology recognition model based on feature saliency and fusion[J]. Neurocomputing, 2025, 616: 128898.
[30]
HeK, ZhangX, RenS, et al. Deep residual learning for image recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
[31]
HowardA, SandlerM, ChuG, et al. Searching for MobileNetv3[C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 1314-1324.
[32]
SimonyanK, ZissermanA. Very deep convolutional networks for large-scale image recognition[J]. arXiv Preprint, arXiv: 1409.1556, 2014.
[33]
SzegedyC, LiuW, JiaY, et al. Going deeper with convolutions[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Boston: IEEE, 2015: 1-9.
[34]
HuangG, LiuZ, van der MaatenL, et al. Densely connected convolutional networks[C]// Proceedings of the IEEE Conference On Computer Vision and Pattern Recognition. New York: IEEE, 2017: 4700-4708.