RD-Crack: A Study of Concrete Crack Detection Guided by a Residual Neural Network Improved Based on Diffusion Modeling

Automated crack detection in concrete structures is an important aspect of structural health monitoring (SHM) to ensure safety and durability. Traditional methods mainly rely on manual inspection, which suffers from subjectivity and inefficiency challenges. To address these issues, machine learning, especially deep learning techniques, has been gradually adopted to improve accuracy and reduce reliance on large amounts of labeled data. This paper introduces RD-Crack, an innovative concrete crack detection method. Our RD-Crack framework combines the encoder with ResNeXt and extrusion excitation modules for feature extraction and uses a diffusion model for parameter optimization to achieve accurate crack detection in complex engineering environments. Experimental results show that our RD-Crack outperforms other state-of-the-art methods in comprehensive performance.