Deep Learning-Based Feature Silencing for Accurate Concrete Crack Detection.

Area of Science:

• Civil Engineering
• Computer Science
• Artificial Intelligence

Background:

• Deteriorated concrete structures pose significant safety risks, necessitating automated inspection methods.
• Current concrete crack detection systems often struggle with accuracy and efficiency.

Purpose of the Study:

• To develop an autonomous concrete crack inspection framework for enhanced automated structural assessment.
• To improve the performance of deep convolutional neural networks in crack segmentation.

Main Methods:

• A deep convolutional neural network architecture was employed for crack segmentation.
• A feature silencing module was incorporated to eliminate non-discriminative feature maps, addressing the gradient vanishing problem.

Main Results:

• The proposed framework demonstrated improved robustness, sensitivity, and specificity in crack detection.
• The system effectively managed the trade-off between specificity and sensitivity for concrete crack analysis.
• The framework achieved higher precision rates and faster processing times compared to existing state-of-the-art methods.

Conclusions:

• The feature silencing module enhances convolutional neural network performance for concrete crack detection.
• The developed framework offers a robust and efficient solution for autonomous concrete inspection, improving infrastructure safety.