Intelligent Prediction of Surrounding Rock Classification by Fusing Image-Text Multimodal Detection Information
-
摘要: 深部地质工程地质条件复杂、隐蔽性强，施工过程存在较多安全隐患，准确预报围岩级别对保障高效施工至关重要.针对破碎岩体特征的不易界定性提出了融合图像-文本多模态探测信息的围岩级别智能预报方法. 使用依据里程范围对齐的综合物探图像，由图文跨模态匹配模型训练得到物探图像响应特征识别模型，用于提取破碎岩体响应特征；由目标检测模型对破碎岩体响应特征进行定位；进而融合物探图像的文本描述特征、破碎岩体目标检测框序列特征及相关地质参数，构建多维统一特征向量，利用多输出随机森林模型实现探测里程范围内围岩级别的准确预报. 通过文本预报结论和目标框预报结论的相互验证，有效避免了不易界定的破碎岩体特征的误报和漏报，实现了由多源物探图像到围岩级别等级的预报，在实际测试中准确率达到92%，为深部地质工程超前智能预报工作奠定了重要基础.Abstract: Deep geotechnical engineering involves complex, hidden geology and high risks, making accurate rock classification critical. This study presents an intelligent prediction method that fuses image-text multimodal geophysical data via mileage-aligned survey images. A cross-modal matching model extracts fractured-rock features, localized by an object detector. Text descriptors, bounding-box sequences, and geological parameters are combined into a unified vector; a multi-output random forest predicts rock classes along mileage. Mutual correction between text and detection reduces false alarms and missed detections. Field tests show 92% accuracy, supporting intelligent advance forecasting.
-
图 2 图像文本标签与岩体完整性分布相关性
a.施作面较完整时文本标签对应岩体完整分布情况；b.施作面较破碎时文本标签对应岩体完整分布情况；c.施作面破碎时文本标签对应岩体完整分布情况
Fig. 2. Correlation between image-text labels and rock integrity distribution
图 3 图像目标框类别与岩体完整性分布相关性
a. 施作面较完整时目标框类别对应岩体完整分布情况；b. 施作面较完整时目标框类别对应岩体完整分布情况；c. 施作面较完整时目标框类别对应岩体完整分布情况
Fig. 3. Correlation between image bounding box categories and rock integrity distribution
图 5 某两段探测里程内围岩级别实际分布情况及预报效果
Fig. 5. Actual distribution and prediction performance of surrounding rock grades within two survey sections
表 1 实际施工过程围岩级别对照关系表
Table 1. Reference table of surrounding rock classifications in actual construction process
围岩级别 岩石硬度/岩体完整性 存在风险 Ⅰ级 岩体完整 岩爆 Ⅱ级 极硬岩：岩体较完整 岩爆、局部掉块 硬岩：岩体完整 局部掉块 Ⅲ级 极硬岩：岩体较破碎 掉块 硬岩：岩体较完整 掉块、坍塌 较软岩：岩体完整 掉块、小坍塌 Ⅳ级 极硬岩：岩体破碎 掉块、小坍塌 硬岩：岩体较破碎 掉块、小坍塌 较软岩：较完整或较破碎 掉块、坍塌 软岩：岩体较完整或完整 掉块、坍塌 Ⅴ级 较软岩：岩体破碎 掉块、坍塌、突泥 软岩：岩体破碎或较破碎 掉块、坍塌、突泥 极软岩：极破碎岩 掉块、坍塌、突泥 Ⅵ级 断层；绿泥石；软黏土 大型坍塌、突涌水 
下载: 导出CSV
表 2 物探图像跨模态模型响应特征识别效果
Table 2. Recognition performance of multimodal model for geophysical image response features
物探图像类型 实际标签 预报结论 GPR-TSP-TEM 同相轴不连续-波速下降-较高阻 同相轴不连续-波速下降-较高阻 GPR 同相轴不连续 同相轴连续性差 GPR-TSP 同相轴连续性差-波速下降 同相轴连续性差-波速下降 GPR-TEM 同相轴连续性一般-较高阻 同相轴连续性一般-较高阻 TSP 波速下降 波速下降 TEM 较低阻 较低阻
下载: 导出CSV
表 3 实际隧道段落内围岩级别预报效果
Table 3. Prediction performance of surrounding rock grades within actual tunnel sections
预报方法 预报准确率 本文所提方法 92% CNN-LSTM模型 84% GoogleNet-LSTM模型 87% SVM模型 79% 本文所提方法去除文本标签 89%
下载: 导出CSV
-
An, J. F., Su, P., Liu, J. Q., et al., 2025. Implementation of YOLO-CLIP Fusion Algorithm for Fall Detection. Signal, Image and Video Processing, 19(10): 937. https://doi.org/10.1007/s11760-025-04397-w Dong, Z., Zhang, X. H., Yang, W. J., et al., 2025. Using a Hybrid of the U-Net and Multilayer Perceptron Models to Automatically Predict the Geological Strength Index from Tunnel Face Rock Joint Images. Rock Mechanics and Rock Engineering, 58(11): 12599-12616. https://doi.org/10.1007/s00603-025-04773-5 Fang, Z. H., Tan, Z. H., Zhu, Z., et al., 2023. Detection of Highway Disease Images Using Ground-Penetrating Radar Based on Faster R-CNN. Construction Technology (Chinese and English), 52(24): 76-82(in Chinese with English abstract). Gan, C., Wang, Y., Cao, W. H., et al., 2025. Real-Time Formation Drillability Sensing-Based Hybrid Online Prediction Method for the Rate of Penetration (ROP) and Its Industrial Application for Drilling Processes. Control Engineering Practice, 164(4): 106487. https://doi.org/10.1016/j.conengprac.2025.106487 Gao, S. Q., Wang Y. Q., Mou Y. C., 2021. Advanced Geological Forecasting and Image Analysis of Karst Tunnels Using Geological Radar. Journal of Engineering Geophysics, 18(5): 642-646(in Chinese with English abstract). doi: 10.1093/jge/gxab041 Jiang, Y., Wang, H. L., Chen, Z., 2024. Intelligent Prediction Algorithm for Advanced Forecasting Images of Adverse Tunnel Geological Bodies Based on Deep Learning. Modern Tunnel Technology, 61(3): 148-156(in Chinese with English abstract). Li, J. F., Sun, S. Y., Zhang, K., et al., 2025a. Single-Stage Zero-Shot Object Detection Network Based on CLIP and Pseudo-Labeling. International Journal of Machine Learning and Cybernetics, 16(2): 1055-1070. https://doi.org/10.1007/s13042-024-02321-1 Li, X. F., Zhang, X. P., Liu, Q. S., et al., 2025b. Evaluation of Rock Mass Quality and Its Mechanical Properties through Digital Drilling Process Monitoring. Journal of Rock Mechanics and Geotechnical Engineering, 17(7): 4490-4511. https://doi.org/10.1016/j.jrmge.2025.01.024 Lin, Z., Feng, S., Fan, G. Y., et al., 2021. Application of Geological Radar in Advanced Geological Forecasting of a Tunnel. Exploration Science and Technology, (3): 47-51+64(in Chinese with English abstract). Liu, C., Li, J., Liu, Z. N., et al., 2025. A Comprehensive Review of Data Processing and Target Recognition Methods for Ground Penetrating Radar Underground Pipeline B-Scan Data. Discover Applied Sciences, 7(4): 310. https://doi.org/10.1007/s42452-025-06791-y Long, G. Y., Zheng, L., 2024. Research on Surrounding Rock Classification Based on Tunnel Advanced Geological Detection. Science and Technology Innovation and Application, 14(6): 124-127(in Chinese with English abstract). Qin, Z. X., Jiang, Y. N., Xu, L., et al., 2023. Automatic Anomaly Detection of Ground-Penetrating Radar Road Images Based on YOLO Algorithm. Science Technology and Engineering, 23(27): 11505-11512(in Chinese with English abstract). Sabri, M. S., Jaiswal, A., Verma, A. K., et al., 2025. Systematic Review of RMR, Q-System, and GSI in Tunnel Classification: Origin, Advancement, and Limitations. Indian Geotechnical Journal, 23(1958): 1-23. https://doi.org/10.1007/s40098-025-01401-5 Saeed, F., Aldera, S., Al-Shamma'a, A. A., et al., 2024. Rapid Adaptation in Photovoltaic Defect Detection: Integrating CLIP with YOLOv8n for Efficient Learning. Energy Reports, 12(18): 5383-5395. https://doi.org/10.1016/j.egyr.2024.11.033 Sun, H. Y., Sheng, L. Y., Dai, Y. M., et al., 2025. 3D Geological Modeling of Tunnel Alignment in the Complex Mountainous Region of Yongshan, China, Based on Multisource Data Fusion. Engineering Geology, 354(3): 108209. https://doi.org/10.1016/j.enggeo.2025.108209 Wang, W. W., Mou, Y. C., 2021. Application of TSP Tunnel Advanced Geological Forecasting Technology in Large Faults. Journal of Engineering Geophysics, 18(2): 186-193(in Chinese with English abstract). Xu, Z. H., Wang, Z. Y., Li, S. C., et al., 2025. GeoPredict-LLM: Intelligent Tunnel Advanced Geological Prediction by Reprogramming Large Language Models. Intelligent Geoengineering, 1(1): 49-57. https://doi.org/10.1016/j.ige.2024.10.005 Yu, X. Y., Dong, N., Zhu, L. H., et al., 2025. CLIP-Driven Semantic Discovery Network for Visible-Infrared Person Re-Identification. IEEE Transactions on Multimedia, 27: 4137-4150. https://doi.org/10.1109/tmm.2025.3535353 Zhang, S., Zhen, D. Y., Liu, K., et al., 2024. Application of 3D visualization of TSP and Transient Electromagnetic Results in Tunnel Fault Water-Bearing Forecasting. China Railway, (5): 75-81(in Chinese with English abstract). Zhang, X. J., 2023. Application of Transient Electromagnetic Method in Advanced Geological Forecasting of Water Diversion Tunnels. Heilongjiang Traffic Science and Technology, 46(10): 102-105(in Chinese with English abstract). Zhang, Y. L., Zhou, J., Li, J. L., et al., 2025. Advancing Overbreak Prediction in Drilling and Blasting Tunnel Using MVO, SSA and HHO-Based SVM Models with Interpretability Analysis. Geomechanics and Geophysics for Geo-Energy and Geo-Resources, 11(1): 53. https://doi.org/10.1007/s40948-025-00963-1 Zhao, Z. P., Chen, J. X., Geng, Q., et al., 2024. Construction Technology and Application of TBM in Shengli Tunnel of Tianshan Mountain under Poor Geological Conditions. Stavební obzor-Civil Engineering Journal, 33(2): 168-180. https://doi.org/10.14311/cej.2024.02.0012 Zheng, P. F., Zhang, A. X., Shi, Z. S., et al., 2025. TLAD-YOLO: Lightweight Network for Intelligent Detection of Railway Tunnel Lining Anomalies Using Ground Penetrating Radar. Journal of Applied Geophysics, 241(12): 105869. https://doi.org/10.1016/j.jappgeo.2025.105869 Zou, Y., Dong, X. J., Feng, T., et al., 2025. Research on Spatial Prediction Technology for Mitigating Tunnel Inrush Disasters under Complex Geological Conditions in China's Hengduan Mountain Range. Scientific Reports, 15(1): 1850-1850. https://doi.org/10.1038/s41598-025-85796-4 房振华, 谭治海, 朱哲, 等, 2023. 基于Faster-RCNN的探地雷达公路病害图像检测, 施工技术(中英文), 52(24): 76-82. 高树全, 王玉琴, 牟元存, 2021. 岩溶隧道地质雷达超前地质预报及图像分析, 工程地球物理学报, 18(5): 642-646. 蒋源, 王海林, 陈兆, 2024. 基于深度学习的隧道不良地质体超前预报图像智能预测算法, 现代隧道技术, 61(3): 148-156. 林志, 冯森, 范国宇, 等, 2021. 地质雷达在某隧道超前地质预报中的应用, 勘察科学技术, (3): 47-51+64. 龙贵云, 郑亮, 2024. 基于隧道超前地质探测的围岩分级研究, 科技创新与应用, 14(6): 124-127. 覃紫馨, 姜彦南, 徐立, 等, 2023. 基于YOLO算法的探地雷达道路图像异常自动检测, 科学技术与工程, 23(27): 11505-11512. 王汪汪, 牟元存, 2021. TSP隧道超前地质预报技术在宽大断层内部的应用, 工程地球物理学报, 18(2): 186-193. 张硕, 甄大勇, 刘康, 等, 2024. TSP与瞬变电磁成果三维化在隧道断层含水预报中的应用, 中国铁路, (5): 75-81. 张旭杰, 2023. 瞬变电磁法在引水隧洞超前地质预报中的应用, 黑龙江交通科技, 46(10): 102-105. -
点击查看大图
图(5) / 表(3)
计量
- 文章访问数: 118
- HTML全文浏览量: 25
- PDF下载量: 19
- 被引次数: 0
