Keywords: 
• Intelligent Analysis of Sedimentary Rocks /
• Rock Thin Section /
• Convolutional Neural Networks (CNNs) /
• Large Models /
• Artificial Intelligence (AI) /
• Deep Learning

Abstract: Accurate identification of mineral composition and microstructural features in sedimentary rock thin sections is an important basis for provenance analysis, reconstruction of depositional environments, and reservoir evaluation. Although traditional manual identification based on polarized light microscopy is highly reliable, it is limited by low efficiency, strong subjectivity, and insufficient standardization. Therefore, it is difficult to meet the growing demand for high-throughput, quantitative, and reproducible analysis in data-driven sedimentological research. Focusing on the development history and technical evolution of intelligent mineral identification in sedimentary rocks, recent progress in traditional image processing, machine learning, and deep learning methods for rock thin-section image analysis is systematically summarized. The applications of convolutional neural networks and Vision Transformers in image classification, object detection, and image segmentation are reviewed, with particular emphasis on their adaptability to the recognition of sedimentological features. The roles of weakly supervised, semi-supervised, and unsupervised learning in reducing dependence on manual annotation are further discussed. The potential of generative models in data augmentation, class balancing, and rare-sample supplementation is also analyzed. In addition, the development trends of vision foundation models, large language models, and multimodal frameworks are examined in terms of geological knowledge integration, model generalization, and interpretive consistency. Intelligent mineral identification is shifting from the application of single models toward an integrated analytical paradigm that combines low-label learning, generative models, and multimodal large models. Future research should strengthen the construction of standardized rock thin-section datasets, the incorporation of geological prior knowledge, cross-domain model adaptation, and interpretability evaluation. A sedimentary rock intelligent analysis system with high generalization ability and geological interpretive consistency is expected to provide efficient and reliable technical support for sedimentological research.