LLM enhances tunnel lining inspection accuracy from 3D point clouds
R4Tun: LLM-guided adaptive segmental tunnel lining segmentation in point clouds
Computer Vision and Pattern Recognition
Summary
Inspecting tunnels is important but tricky because conditions change and automatic methods often struggle. The authors developed R4Tun, a system that uses large language models (LLMs) to adjust the inspection process automatically based on the tunnel's situation. This makes the system much better at identifying parts of the tunnel lining without needing extra labeled data. Their tests showed clear improvements in accuracy across different tunnel types and LLM types.
What this means in practice
- •For infrastructure inspection teams: Automatically improve 3D tunnel lining segmentation accuracy by adapting parameters with language model guidance without extra labeling.
- •For civil engineering consultants: Use LLM-driven tuning to maintain reliable tunnel condition assessments under varying real-world environments and complexities.
Authors
Xinghui Tao, Zehao Ye, Guangming Wang, Jelena Ninić, Brian Sheil
Abstract
Automated inspection of segmental tunnel linings requires adaptive segmentation from 3D point clouds, yet expert-tuned pipelines often degrade when tunnel conditions vary. This paper presents R4Tun, a large language model (LLM)-driven adaptation framework that extends an expert-designed pipeline (SAM4Tun) with bounded parameter tuning informed by structured context: memory ($m$), state ($s$), and knowledge ($k$). Evaluated on 30 selected Seg2Tunnel subsets (13 regular, 17 complex) across three LLMs, the full $m+s+k$ design raised mean Intersection-over-Union (mIoU) from 0.18 to 0.43--0.48 and overall accuracy (OA) from 0.42 to 0.59--0.65 relative to the static SAM4Tun baseline, with the near-reference regular (staggered) subsets reaching mIoU 0.784--0.796 across LLMs. Across 270 (30 tunnels $\times$ 3 different LLMs $\times$ 3 context settings) runs, the LLMs showed similar parameter-adjustment trends (with overlapping 95\% CIs on mean gains) and consistently adjusted a shared set of critical parameters. These results support R4Tun as a controlled, label-free, cross-LLM adaptation mechanism in the tested SAM4Tun--Seg2Tunnel setting, demonstrating consistent accuracy gains; we position R4Tun as a mechanism contribution rather than a deployable final-inspection system, in which each bounded parameter change is auditable via logged rationales.