基于层序智能提取的构造地质图三维地质建模方法研究

doi:10.13745/j.esf.sf.2026.2.65

地学前缘 ›› 2026, Vol. 33 ›› Issue (4): 95-105.DOI: 10.13745/j.esf.sf.2026.2.65

基于层序智能提取的构造地质图三维地质建模方法研究

郭甲腾¹^,^*(), 李俊昆¹, 赵域浩¹, 王徐磊¹, 王路远¹, 向楷文¹, 熊自明², 李丰丹³

1.东北大学资源与土木工程学院, 辽宁沈阳 110819
2.陆军工程大学爆炸冲击防灾减灾全国重点实验室, 江苏南京 210007
3.中国地质调查局自然资源综合调查指挥中心, 北京 100055

收稿日期:2026-01-05 修回日期:2026-03-25 出版日期:2026-07-25 发布日期:2026-04-20
通信作者: *郭甲腾(1980—),男,博士,教授,博士生导师,主要从事三维地学建模理论与应用研究。E-mail: guojiateng@mail.neu.edu.cn
基金资助:
国家自然科学基金项目(42172327);国家自然科学基金项目(42477197);国家自然科学基金项目(52278419);爆炸冲击防灾减灾全国重点实验室资助项目(LGD-SKL-202209)

Research on 3D geological modeling method of structural geological maps based on intelligent extraction of layer sequence

GUO Jiateng¹^,^*(), LI Junkun¹, ZHAO Yuhao¹, WANG Xulei¹, WANG Luyuan¹, XIANG Kaiwen¹, XIONG Ziming², LI Fengdan³

1. School of Resources & Civil Engineering, Northeastern University, Shenyang 110819, China
2. State Key Laboratory of Disaster Prevention and Mitigation of Explosion and Impact, Army Engineering University of PLA, Nanjing 210007, China
3. Comprehensive Survey Command Center for Natural Resources, China Geological Survey, Beijing 100055, China

Received:2026-01-05 Revised:2026-03-25 Online:2026-07-25 Published:2026-04-20

摘要/Abstract

摘要：

基于地质图自动构建三维地质模型是推动数字地球技术发展的关键技术。目前主流三维地质建模方法难以仅通过图中的地质界线、产状等几何信息确定准确的地下地质构造,往往需要通过人工确定地层切割顺序或划定势场等方式才能实现建模,无法将图中体现的深层地质语义自动、智能地融入建模过程中。为此,本文提出一种基于多模态大模型的构造地质图三维地质建模方法。首先,采用地质构造相似案例检索与先验钻孔注入的方式对多模态大模型进行提示调优,实现地质图目标点位的地层层序智能提取;其次,基于图上的地质界线与产状几何数据,自动计算对应的地层厚度信息,完成虚拟钻孔的构建;最后,通过虚拟钻孔与地质图联合提取出地上下地层约束空间点集,采用神经网络对空间点地层属性进行分类预测训练并完成三维模型的构建。实验结果表明,与主流建模方法相比,该方法实现了针对复杂构造地质图的虚拟钻孔自动化智能提取,并能构建地表-地下无缝衔接的三维地质模型。

关键词: 智能三维地质建模, 多模态大模型, 地质构造三维建模, 提示调优与检索增强生成, 数字地质调查

Abstract:

3D modeling from geological maps is a key technology for advancing digital earth development. Traditional modeling methods based on geological maps often struggle to accurately reconstruct subsurface structures using only geometric information—such as geological boundaries and attitudes. These methods typically require manual determination of stratigraphic sequences or potential field delineations, making it difficult to automatically integrate the deep geological semantics embedded in the maps into the modeling process. To address this limitation, we propose a novel 3D geological modeling framework based on a large multimodal model (LMM). First, through case retrieval based on geological structure similarity and pre-drilling data injection, the LMM is rapidly optimized to intelligently extract the stratum sequence at target positions on the geological map. Second, using geometric data from geological boundaries and attitudes, formation thickness information is automatically calculated and virtual boreholes are constructed. Finally, spatial constraint point sets for surface and subsurface strata are extracted from the virtual boreholes and geological map; stratum attributes are then classified and predicted using a neural network, and a 3D model is constructed. Experiments show that, compared with mainstream modeling methods, the proposed approach achieves automatic and intelligent extraction of virtual boreholes from complex structural geological maps and constructs three-dimensional geological models with seamless surface-subsurface integration.

Key words: intelligent 3D geological modeling, large multimodal model, 3D modeling of geological structures, prompt optimization and retrieval-augmented generation, digital geological survey

中图分类号:

郭甲腾, 李俊昆, 赵域浩, 王徐磊, 王路远, 向楷文, 熊自明, 李丰丹. 基于层序智能提取的构造地质图三维地质建模方法研究[J]. 地学前缘, 2026, 33(4): 95-105.

GUO Jiateng, LI Junkun, ZHAO Yuhao, WANG Xulei, WANG Luyuan, XIANG Kaiwen, XIONG Ziming, LI Fengdan. Research on 3D geological modeling method of structural geological maps based on intelligent extraction of layer sequence[J]. Earth Science Frontiers, 2026, 33(4): 95-105.

图/表 12

图1 基于智能层序提取的三维地质建模总体流程

Fig.1 The overall process of 3D geological modeling based on intelligent sequence extraction

图2 提示词调优整体架构

Fig.2 The overall framework of prompt tuning

图3 基于钻孔和地质图的空间点采样

Fig.3 Spatial point sampling based on boreholes and geological maps

图4 野鸭湖地区地形地质图

Fig.4 Topographic and geological map of the Yeyahu area

表1 基座大模型先验钻孔层序提取测试结果

Table 1 Test results of prior drilling sequence extraction for LMMs

多模态大模型	是否注入相似案例	地层层序完全正确 (钻孔数量/个)	层序部分正确但底部缺层 (钻孔数量/个)	地层层序错误 (钻孔数量/个)
GPT-5.1	是	5	1	0
Gemini-2.5-pro	是	3	1	2
Claude-sonnet-4.5	是	3	2	1
GPT-5-mini	是	1	2	3
GPT-5.1	否	3	2	1
Gemini-2.5-pro	否	2	1	3
Claude-sonnet-4.5	否	2	2	2
GPT-5-mini	否	0	3	3

图5 实验所提取的虚拟钻孔数据可视化

Fig.5 The virtual boreholes data extracted from the experiment

图6 野鸭湖三维地质模型

Fig.6 Three-dimensional geological model of Yeyahu Lake

图7 训练过程损失函数值

Fig.7 Loss function value of training process

图8 地质剖面分析

Fig.8 Geological section analysis

图9 不同建模方法效果对比

Fig.9 Comparison of the effects of different modeling methods

图10 不同建模方法剖面对比

Fig.10 Sections comparison of different modeling methods

表2 不同方法建模结果地质界线吻合程度比较

Table 2 Comparison of the coincidence degree of geological boundaries based on different methods

评价指标	本文方法	钻孔数据隐式建模方法
召回率	79.07%	78.29%
精确率	50.49%	23.50%

参考文献 32

[1]	侯恩科, 吴立新. 三维地学模拟几个方面的研究现状与发展趋势[J]. 煤田地质与勘探, 2000, 28(6): 1-5.
[2]	胡金星, 吴立新, 高卫贞, 等. 三维地学模拟体视化技术的应用研究[J]. 煤炭学报, 1999, 24(4): 11-15.
[3]	GUO J, WU L, ZHOU W, et al. Towards automatic and topologically consistent 3D regional geological modeling from boundaries and attitudes[J]. ISPRS International Journal of Geo-Information, 2016, 5(2): 17. DOI URL
[4]	车德福, 吴立新, 殷作如, 等. 基于GTP的断层三维交互建模方法[J]. 东北大学学报(自然科学版), 2008, 29(3): 395-398.
[5]	郭甲腾, 刘寅贺, 韩英夫, 等. 基于机器学习的钻孔数据隐式三维地质建模方法[J]. 东北大学学报(自然科学版), 2019, 40(9): 1337-1342. DOI
[6]	牛露佳, 石成岳, 王占刚, 等. 三维复杂地质结构模型的InterfaceGrid表达方法[J]. 地学前缘, 2024, 31(4): 129-138. DOI
[7]	郭甲腾, 吴立新, 王江梅, 等. 基于隐式化Coons曲面的局部地质构造区域集成建模方法[J]. 地理与地理信息科学, 2018, 34(1): 1-6.
[8]	GUO J, WANG J, WU L, et al. Explicit-implicit-integrated 3-D geological modelling approach: a case study of the Xianyan Demolition Volcano (Fujian, China)[J]. Tectonophysics, 2020, 795: 228648. DOI URL
[9]	周良辰, 林冰仙, 王丹, 等. 平面地质图的三维地质体建模方法研究[J]. 地球信息科学学报, 2013, 15(1): 46-54.
[10]	HILLIER M J, SCHETSELAAR E M, DE KEMP E A, et al. Three-dimensional modelling of geological surfaces using generalized interpolation with radial basis functions[J]. Mathematical Geosciences, 2014, 46(8): 931-953. DOI URL
[11]	GUADAGNIN F, CORDOVA D P. Variational Gaussian processes for implicit geological modeling[J]. Computers & Geosciences, 2023, 174: 105323. DOI URL
[12]	GUO J, WANG X, WANG J, et al. Three-dimensional geological modeling and spatial analysis from geotechnical borehole data using an implicit surface and marching tetrahedra algorithm[J]. Engineering Geology, 2021, 284: 106047. DOI URL
[13]	郭甲腾, 吴立新, 周文辉. 基于径向基函数曲面的矿体隐式自动三维建模方法[J]. 煤炭学报, 2016, 41(8): 2130-2135.
[14]	HILLIER M, WELLMANN F, DE KEMP E A, et al. GeoINR 1.0: an implicit neural representation network for three-dimensional geological modelling[J]. Geoscientific Model Development, 2023, 16(23): 6987-7012. DOI URL
[15]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]// Advances in Neural Information Processing Systems. Red Hook: Curran Associates, 2017: 5998-6008.
[16]	RADFORD A, NARASIMHAN K, SALIMANS T, et al.Improving language understanding by generative pre-training[R]. San Francisco: OpenAI, 2018.
[17]	DEVLIN J, CHANG M W, LEE K, et al.BERT: pre-training of deep bidirectional transformers for language understanding[PP/OL]. arXiv (2019-05-24) [2026-01-16]. https://arxiv.org/pdf/1810.04805.
[18]	ACHIAM J, ADLER S, AGARWAL S, et al. GPT-4 technical report[PP/OL]. arXiv (2024-05-04)[2026-01-16]. https://arxiv.org/pdf/2303.08774.
[19]	EL-KISHKY A, WEI A, SARAIVA A, et al. Competitive programming with large reasoning models[PP/OL]. arXiv (2025-02-18) [2026-01-16]. https://arxiv.org/pdf/2502.06807.
[20]	GUO D, YANG D, ZHANG H, et al. DeepSeek-R1:incentivizing reasoning capability in llms via reinforcement learning[PP/OL]. arXiv (2026-01-04) [2026-01-16]. https://arxiv.org/pdf/2501.12948.
[21]	HUANG Y, GAO T, XU H, et al. PEACE: empowering geologic map holistic understanding with MLLMs[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2025: 3899-3908.
[22]	ZHANG Y, WEI C, HE Z, et al. GeoGPT: an assistant for understanding and processing geospatial tasks[J]. International Journal of Applied Earth Observation and Geoinformation, 2024, 131: 103976. DOI URL
[23]	周永章, 肖凡. 管窥人工智能与大数据地球科学研究新进展[J]. 地学前缘, 2024, 31(4): 1-6. DOI
[24]	季晓慧, 董雨航, 杨中基, 等. 基于知识图谱多跳推理的中文矿物知识问答方法与系统[J]. 地学前缘, 2024, 31(4): 37-46. DOI
[25]	叶育鑫, 刘家文, 曾婉馨, 等. 基于本体指导的矿产预测知识图谱构建研究[J]. 地学前缘, 2024, 31(4): 16-25. DOI
[26]	LONG Q, WU Y, WANG W, et al. Does in-context learning really learn? Rethinking how large language models respond and solve tasks via in-context learning[PP/OL]. arXiv (2024-07-23) [2026-01-16]. https://arxiv.org/abs/2404.07546.
[27]	LEWIS P, PEREZ E, PIKTUS A, et al. Retrieval-augmented generation for knowledge-intensive NLP tasks[C]// Advances in Neural Information Processing Systems. Red Hook: Curran Associates, 2020: 9459-9474.
[28]	KARPUKHIN V, OGUZ B, MIN S, et al. Dense passage retrieval for open-domain question answering[C]// Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). Stroudsburg: ACL, 2020: 6769-6781.
[29]	CHEN J, XIAO S, ZHANG P, et al. BGE M3-Embedding:multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation[PP/OL]. arXiv (2025-12-12) [2026-01-16]. https://arxiv.org/pdf/2402.03216.
[30]	JESSELL M, VALENTA R. Structural geophysics: integrated structural and geophysical modelling[J]. Computer Methods in The Geosciences, 1996, 15: 303-324.
[31]	杨长义, 尹崧宇, 李俊昆, 等. 基于自适应八叉树的深度学习地质建模并行显式化采样方法[J]. 地理与地理信息科学, 2025, 41(6): 1-6.
[32]	罗金海, 梁文天, 于在平. 构造地质学[M]. 1版. 北京: 高等教育出版社, 2018: 362.

基于层序智能提取的构造地质图三维地质建模方法研究

Research on 3D geological modeling method of structural geological maps based on intelligent extraction of layer sequence

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 32

相关文章 15

编辑推荐

Metrics

本文评价

[1]	于文豪, 黄兴富, 郭晓玉, 高锐. 青藏高原东北缘晚中生代构造演化[J]. 地学前缘, 2026, 33(3): 338-353.
[2]	冯乾乾, 邱楠生, 腾格尔, 龙康杰, 李萱. 磷灰石裂变径迹和(U-Th)/He约束下雪峰山隆起北缘中—新生代构造抬升过程[J]. 地学前缘, 2026, 33(3): 354-362.
[3]	韩飞, 李建江, 秦宇龙, 赵海波, 杨学俊, 叶春林, 李峥, 贾小川, 黄柏鑫, 张彤, 熊昌利, 黄永高, 罗改. 西藏冈底斯南缘新生代隆升剥露历史及地质意义[J]. 地学前缘, 2026, 33(2): 440-450.
[4]	王朱亭, 顾承串, 鲁海峰, 王一波, 胡圣标, 姜光政, 张超. 两淮地区逆冲推覆构造对区域地温场分布的影响[J]. 地学前缘, 2026, 33(2): 451-461.
[5]	秦阳, 刘池洋, 彭光荣, 黄雷, 李洪博, 梁超, 吴哲, 杨丽华. 珠江口盆地云开低凸起形成演化及构造分区作用[J]. 地学前缘, 2025, 32(4): 422-443.
[6]	王国灿, 赵子豪, 申添毅, 马骋, 周亚波. 从中亚岩石冷却的时空差异性浅析天山中新生代隆升剥露的动力来源[J]. 地学前缘, 2025, 32(1): 322-342.
[7]	田述军, 温宇航, 伍文洽, 李凯. 岷江上游坡断型裂点迁移速率与构造隆升历史研究[J]. 地学前缘, 2024, 31(4): 314-325.
[8]	吴晨, 陈宣华, 丁林. 祁连造山带构造演化与新生代变形历史[J]. 地学前缘, 2023, 30(3): 262-281.
[9]	李路顺, 汪泽成, 肖安成, 胡安平, 陈友智, 王芊芊. 扬子板块北缘新元古代盆地结构与马槽园群归属研究[J]. 地学前缘, 2022, 29(6): 291-304.
[10]	周琦杰, 刘永江, 王德英, 关庆彬, 王光增, 王宇, 李遵亭, 李三忠. 渤海湾中部中、新生代构造演化与潜山的形成[J]. 地学前缘, 2022, 29(5): 147-160.
[11]	李典, 王根厚, 刘正勇, 李鹏胜, 冯翼鹏, 唐宇, 李超, 李阳. 西藏南羌塘晚三叠世陆缘俯冲增生造山带的褶皱-冲断与增生杂岩双层结构厘定[J]. 地学前缘, 2022, 29(4): 231-248.
[12]	Xiaojun WANG, Yong SONG, Baoli BIAN, Junmeng ZHAO, Heng ZHANG, Maodu YAN, Shunping PEI, Qiang XU, Shuaijun WANG, Hongbing LIU, Changhui JU. 准噶尔盆地基底构造[J]. 地学前缘, 2021, 28(6): 235-255.
[13]	李典, 王根厚, 刘正勇, 李鹏胜, 冯翼鹏, 唐宇, 李超, 李阳. 西藏南羌塘增生杂岩中俯冲反向驱动高压变质岩折返:来自猫耳山大型拆离断层的启示[J]. 地学前缘, 2021, 28(6): 205-226.
[14]	陈宇，曹淑云，邓友国，程雪梅，吕美霞，董彦龙. 哀牢山—红河剪切带中流体记录及地质意义[J]. 地学前缘, 2019, 26(2): 42-57.
[15]	李洪奎，禚传源，耿科，梁太涛. 郯庐断裂带陆内伸展构造: 以沂沭断裂带的表现特征为例[J]. 地学前缘, 2017, 24(2): 73-84.