Real-time grading method of tunnel surrounding rock based on image recognition

Area of Science:

• Geotechnical Engineering
• Artificial Intelligence in Civil Engineering

Background:

• Accurate grading of tunnel surrounding rock is crucial for construction safety and efficiency.
• Traditional methods for rock grading can be time-consuming and subjective.

Purpose of the Study:

• To develop a rapid and accurate real-time grading method for tunnel surrounding rock.
• To integrate image processing with lightweight deep learning for automated rock classification.

Main Methods:

• An automated pipeline was developed combining image processing and machine learning algorithms.
• Image preprocessing included grayscale conversion, denoising, enhancement, and normalization.
• A ShuffleNetV2 convolutional neural network model was optimized for lithology identification and rock-mass property sensing.

Main Results:

• The method achieved 85% accuracy for lithology classification.
• Rock-mass integrity was classified with 75% accuracy.
• Overall surrounding-rock grade was achieved with 80% accuracy on an engineering site.

Conclusions:

• The proposed real-time grading method demonstrates feasibility and practical value for tunnel construction.
• The integration of image processing and deep learning offers theoretical insights for scientific evaluation of tunnel surrounding rock.