Concrete Bridge Crack Detection Method Based on an Improved RT-DETR Model

Cracks in bridges are early indicators of potential structural damage and need quick, accurate detection to prevent accidents and extend bridge lifespan. Traditional crack detection networks often miss minor cracks, are parameter-heavy, and struggle with real-time processing. To address these issues, We have utilized the RT-DETR (Real-Time Detection Transformer) and introduced the CBAM (Convolutional Block Attention Module) attention mechanism, replacing the Transformer-encoder self-attention mechanism. This improvement enhances the attention mechanism at both the channel and spatial levels, reducing the number of model parameters and enhancing the model’s ability to recognize complex and fine crack features. The proposed improvements achieve an accuracy rate of up to 98.7%, achieving in higher detection precision and faster processing speeds. This research provides efficient and reliable support for bridge maintenance.