An interactive framework integrating segment anything model and structure-from-motion for three-dimensional discontinuity identification in rock masses

Abstract

The identification of rock mass discontinuities is critical for rock mass characterization. While high-resolution digital outcrop models (DOMs) are widely used, current digital methods struggle to generalize across diverse geological settings. Large-scale models (LSMs), with vast parameter spaces and extensive training datasets, excel in solving complex visual problems. This study explores the potential of using one such LSM, Segment anything model (SAM), to identify facet-type discontinuities across several outcrops via interactive prompting. The findings demonstrate that SAM effectively segments two-dimensional (2D) discontinuities, with its generalization capability validated on a dataset of 2426 identified discontinuities across 170 outcrops. The model achieves 0.78 mean IoU and 0.86 average precision using 11-point prompts. To extend to three dimensions (3D), a framework integrating SAM with Structure-from-Motion (SfM) was proposed. By utilizing the inherent but often overlooked relationship between image pixels and point clouds in SfM, the identification process was simplified and generalized across photogrammetric devices. Benchmark studies showed that the framework achieved 0.91 average precision, identifying 87 discontinuities in Dataset-3D. The results confirm its high precision and efficiency, making it a valuable tool for data annotation. The proposed method offers a practical solution for geological investigations.

Keywords

Rock Mass / Discontinuity / Digital outcrop model (DOM) / Point clouds / Large-scale model (LSM) / Foundation model (FM)

Acknowledgments

The authors would like to thank Dr. Matthew J. Lato for providing raw data of the Roadcut in Canada, as well as Dr. Hamid Daghigh for his support in dataset preparation. This study was funded by National Natural Science Foundation of China (Nos. 42422704 and 52379109), Opening the fund of State Key Laboratory of Geohazard Prevention and Geoenvironment Protection (Chengdu University of Technology) (No. SKLGP2024K028), and Science and Technology Research and Design Projects of China State Construction Engineer-ing Corporation Ltd. (No. CSCEC-2024-Q-68).

Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.ijmst.2025.09.005.

References

[1] Hudson JA, Priest SD. Discontinuities and rock mass geometry. Int J Rock Mech Min Sci Geomech Abstr 1979; 16(6):339-62.

[2] Zhu HH, Azarafza M, Akgün H. Deep learning-based key-block classification framework for discontinuous rock slopes. J Rock Mech Geotech Eng 2022; 14 (4):1131-9.

[3] Zheng J, Guo JC, Wang JC, Sun HL, Deng JH, Lv Q.A universal elliptical disc (UED) model to represent natural rock fractures. Int J Min Sci Technol 2022; 32 (2):261-70.

[4] Azarafza M, Ghazifard A, Akgün H, Asghari-Kaljahi E. Development of a 2D and 3D computational algorithm for discontinuity structural geometry identification by artificial intelligence based on image processing techniques. Bull Eng Geol Environ 2019; 78(5):3371-83.

[5] Azarafza M, Koçkar MK, Faramarzi L. Spacing and block volume estimation in discontinuous rock masses using image processing technique: A case study. Environ Earth Sci 2021; 80(14):471.

[6] Zare Naghadehi M, Jimenez R, KhaloKakaie R, Jalali SE. A new open-pit mine slope instability index defined using the improved rock engineering systems approach. Int J Rock Mech Min Sci 2013; 61:1-14.

[7] Zare M, Jimenez R. On the development of a slope instability index for open-pit mines using an improved systems approach. In: Proceedings of the ISRM Regional Symposium-EUROCK 2015. Salzburg:ISRM; 2015: ISRM-EUROCK-2015-170.

[8] Palmstrom A. Measurement and Characterization of Rock Mass Jointing. In:Proceedings of the In-Situ Characterization of Rocks. London: A. A. Balkema Publishers; 2001.p.97.

[9] Xu XM, Aiken CLV, Bhattacharya JP, Corbeanu RM, Nielsen KC, McMechan GA, Abdelsalam MG. Creating virtual 3-D outcrop. Lead Edge 2000; 19(2):197-202.

[10] Bellian JA, Kerans C, Jennette DC. Digital outcrop models: Applications of terrestrial scanning lidar technology in stratigraphic modeling. J Sediment Res 2005; 75(2):166-76.

[11] Riquelme AJ, Abellán A, Tomás R, Jaboyedoff M. A new approach for semi-automatic rock mass joints recognition from 3D point clouds. Comput Geosci 2014; 68:38-52.

[12] Haneberg WC. Using close range terrestrial digital photogrammetry for 3-D rock slope modeling and discontinuity mapping in the United States. Bull Eng Geol Environ 2008; 67(4):457-69.

[13] Daghigh H, Tannant DD, Daghigh V, Lichti DD, Lindenbergh R. A critical review of discontinuity plane extraction from 3D point cloud data of rock mass surfaces. Comput Geosci 2022; 169:105241.

[14] Battulwar R, Zare-Naghadehi M, Emami E, Sattarvand J. A state-of-the-art review of automated extraction of rock mass discontinuity characteristics using three-dimensional surface models. J Rock Mech Geotech Eng 2021; 13 (4):920-36.

[15] Wang MY, Wang EZ, Liu XL, Wang CC. Scale-space effect and scale hybridization in image intelligent recognition of geological discontinuities on rock slopes. J Rock Mech Geotech Eng 2024; 16(4):1315-36.

[16] Ma J, He YT, Li FF, Han L, You CY, Wang B. Segment anything in medical images. Nat Commun 2024; 15(1):654.

[17] Osco LP, Wu QS, de Lemos EL, Gonçalves WN, Ramos APM, Li J, Junior JM. The Segment Anything Model (SAM) for remote sensing applications: from zero to one shot. Int J Appl Earth Obs Geoinf 2023; 124:103540.

[18] Awais M, Naseer M, Khan S, Anwer RM, Cholakkal H, Shah M. Foundation models defining a new era in vision: a survey and outlook. IEEE Trans Pattern Anal Mach Intell 2025; 47(4):2245-64.

[19] Kirillov A, Mintun E, Ravi N, Mao HZ, Rolland C, Gustafson L, Xiao TT, Whitehead S, Berg AC, Lo WY, Dollár P, Girshick R. Segment anything.arXiv:230402643 [Preprint]. 2023 [cited 2025 Jun 1]: [30 p.]. Available from: https://arxiv.org/abs/2304.02643.

[20] Lee YK, Kim J, Choi CS, Song JJ. Semi-automatic calculation of joint trace length from digital images based on deep learning and data structuring techniques. Int J Rock Mech Min Sci 2022; 149:104981.

[21] Bommasani R, Hudson DA, Adeli E, Altman R, Arora S, von Arx S, Bernstein MS, Bohg J, Bosselut A, Brunskill E, Brynjolfsson E, Buch S, Card D, Castellon R, Chatterji N.On the opportunities and risks of foundation models. arXiv:210807258 [Preprint]. 2022 [cited 2025 Jun 1]: [214 p.]. Available from: https://arxiv.org/abs/2108.07258.

[22] Jain J, Li JC, Chiu MT, Hassani A, Orlov N, Shi H. OneFormer: One transformer to rule universal image segmentation. arXiv:221106220 [Preprint]. 2022 [cited 2025 Jun 1]: [19 p.]. Available from: https://arxiv.org/abs/2211.06220.

[23] Li H, Zhu JG, Jiang XH, Zhu XZ, Li HS, Yuan C, Wang XH, Qiao Y, Wang XG, Wang WH, Dai JF.Uni-Perceiver v2:A generalist model for large-scale vision and vision-language tasks. arXiv:221109808 [Preprint]. 2022 [cited 2025 Jun 1]:[14 p.]. Available from: https://arxiv.org/abs/2211.09808.

[24] Wang XL, Zhang XS, Cao Y, Wang W, Shen CH, Huang TJ. SegGPT: Segmenting everything in context. arXiv:230403284 [Preprint]. 2023 [cited 2025 Jun 1]:[12 p.]. Available from: https://arxiv.org/abs/2304.03284.

[25] Zou XY, Yang JW, Zhang H, Li F, Li LJ, Wang JF, Wang LJ, Gao JF, Lee JY.Segment everything everywhere all at once. arXiv:230406718 [Preprint]. 2023 [cited 2025 Jun 1]: [13 p.]. Available from: https://arxiv.org/abs/2304.06718.

[26] Moulon P, Monasse P, Marlet R. Adaptive structure from motion with a contrario model estimation. Computer Vision - ACCV 2012. Berlin, Heidelberg: Springer, 2013:257-70.

[27] Microsoft. Image Composite Editor [software]. [software].2015 Dec 02[2025 Jun 1]

[28] Gledhill D, Tian GY, Taylor D, Clarke D. Panoramic imaging: A review. Comput Graph 2003; 27(3):435-45.

[29] Bemis SP, Micklethwaite S, Turner D, James MR, Akciz S, Thiele ST, Bangash HA. Ground-based and UAV-Based photogrammetry: A multi-scale, high-resolution mapping tool for structural geology and paleoseismology. J Struct Geol 2014; 69:163-78.

[30] Griwodz C, Gasparini S, Calvet L, Gurdjos P, Castan F, Maujean B, De Lillo G, Lanthony Y. In: AliceVision Meshroom: An open-source 3D reconstruction pipeline. Istanbul Turkey: ACM; 2021:. p. 241-7.

[31] Agisoft LLC. 1.2.6 softwarecited Agisoft Metashape [software]. 2025 Jun 1]. Available from: https://www.agisoft.com/.

[32] Rey OI.Anatomy of the SIFT method. Doctoral dissertation Paris: Ecole Normale Supérieure de Cachan 2015.

[33] Alcantarilla PF, Nuevo J, Bartoli A. Fast explicit diffusion for accelerated features in nonlinear scale spaces 2013; 13:.1-13.11.

[34] Sweeney C, Hollerer T, Theia TM. In: A fast and scalable structure-from-motion library. Brisbane Australia: ACM; 2015. p. 693-6.

[35] Russell BC, Torralba A, Murphy KP, Freeman WT. LabelMe: A database and web-based tool for image annotation. Int J Comput Vis 2008; 77(1):157-73.

[36] Xu N, Price B, Cohen S, Yang JM, Huang T. In: Deep interactive object selection. Las Vegas: IEEE; 2016. p. 373-81.

[37] CloudCompare. CloudCompare [software]. Version 2.13.2 [software].2024 Jul 11 [2025 Jun 1]. Available from: https://www.cloudcompare.org/.

[38] Lato M, Kemeny J, Harrap RM, Bevan G. Rock bench: establishing a common repository and standards for assessing rockmass characteristics using LiDAR and photogrammetry. Comput Geosci 2013; 50:106-14.

[39] Zhang BH, Li WK, Zheng J, Kulatilake PHSW, Qing L, Duan YS. A numerical simulation procedure for evaluating the accuracy of 3-dimensional photogrammetric models and its application to geometric parameters of discontinuities in rock masses. Phys Chem Earth Parts A/B/C 2023; 129:103364.

[40] Battulwar R, Emami E, Naghadehi MZ, Sattarvand J. Automatic extraction of joint orientations in rock mass using PointNet and DBSCAN. advances in Visual Computing. Cham: Springer International Publishing; 2020. p. 718-27.

[41] Xu QS, Tao WB. In: Multi-scale geometric consistency guided multi-view stereo. Long Beach: IEEE; 2019. p. 5478-87.

[42] Paixão A, Muralha J, Resende R, Fortunato E. Close-range photogrammetry for 3D rock joint roughness evaluation. Rock Mech Rock Eng 2022; 55(6):3213-33.

[43] Xiang X, Wang ZY, Lao SS, Zhang BC. Pruning multi-view stereo net for efficient 3D reconstruction. ISPRS J Photogramm Remote Sens 2020; 168:17-27.

[44] Yang WQ, Chen GY, Chen CF, Chen ZF, Wong KK. PS-NeRF:Neural inverse rendering for multi-view photometric stereo. In: Computer Vision - ECCV 2022. Cham: Springer Nature Switzerland; 2022:266-84.

[45] Wang FJH, Zhu QT, Chang D, Gao QK, Han JL, Zhang T, Hartley R,Pollefeys M. Learning-based multi-view stereo: a survey. arXiv preprint 2024: arXiv:2408.15235.

[46] Mehrishal S, Kim J, Song JJ, Sainoki A. A semi-automatic approach for joint orientation recognition using 3D trace network analysis. Eng Geol 2024; 332:107462.

[47] Wang JW, Zheng J, Hu J, Jimenez R, Zhu AL, Lü Q, Wang JC. In-situ block intelligent identification: Large-scale artificial intelligence (AI) models enable efficient geological surveys via image segmentation. Eng Geol 2025; 356:108286.