Computer vision-based automatic detection and three-dimensional modeling method for tunnel lining cracks

By combining computer vision and physical information neural networks, the problem of high-fidelity reconstruction and mechanical feature detection of tunnel lining cracks in complex environments has been solved, enabling accurate detection and risk assessment of tunnel structures.

CN122223246APending Publication Date: 2026-06-16GUIZHOU ROAD & BRIDGE GRP
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
GUIZHOU ROAD & BRIDGE GRP
Filing Date
2026-05-20
Publication Date
2026-06-16

AI Technical Summary

Technical Problem

Existing technologies for detecting cracks in tunnel linings struggle to accurately reconstruct the geometric and mechanical characteristics of cracks under harsh lighting conditions and with monotonous textures. Furthermore, traditional methods cannot invert the deep stress field of the structure, leading to inaccurate detection results.

Method used

A computer vision-based approach was adopted, combining multi-view image acquisition, neural radiation field model and physical information neural network. Crack features were identified through multilayer perceptron network and deep convolutional neural network, and a mechanical model constrained by physical information neural network was constructed to invert the internal stress field of the lining and generate a three-dimensional digital twin model.

Benefits of technology

It achieves high-fidelity reconstruction of tunnel lining cracks and accurate detection of mechanical characteristics, and can invert internal stress concentration areas and crack depth, improving the objectivity and intelligence of detection and providing scientific risk assessment support.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122223246A_ABST
    Figure CN122223246A_ABST
Patent Text Reader

Abstract

The application relates to the technical field of tunnel lining images, and particularly discloses a tunnel lining crack automatic detection and three-dimensional modeling method based on computer vision. The method comprises the following steps: collecting and pre-processing lining sequence images; constructing a neural radiation field model to reconstruct a three-dimensional geometric surface; adopting a deep convolutional neural network and an attention mechanism to segment cracks and map the cracks to a three-dimensional model; constructing a mechanical model under the constraint of a physical information neural network, introducing a solid mechanics equation into a loss function; performing internal stress field inversion and crack depth estimation; and fusing multi-source data to generate a digital twin model for safety evaluation. Through physical driving and visual feature fusion, the application eliminates geometric distortion caused by environmental interference, and improves the objectivity, accuracy and fine level of tunnel disease monitoring.
Need to check novelty before this filing date? Find Prior Art