In December 2024, a paper titled “Enhanced Real-Time Detection Transformer (RT-DETR) for Robotic Inspection of Underwater Bridge Pier Cracks” was published in the international top-tier journal Automation in Construction (with an impact factor of 10.3) by Graduate Student Lv Zhenming and Professor Dong Shaojiang from Chongqing Jiaotong University.

Crack defects in underwater bridge piers are unavoidable, as they are a critical and vulnerable component of bridges. In order to address the issue of insufficient crack data for underwater bridge piers, this paper introduced a global local discriminator and pixel attention mechanism. Additionally, a crack generation method for underwater bridge piers is proposed, which is based on an improved CycleGAN. This method replaces the generator of CycleGAN with the U-Net framework and incorporates the pixel attention mechanism after each downsampling and upsampling. This method in this paper is capable of generating finer crack images and simultaneously replaces the discriminator with a global local discriminator to mitigate the artifacts and noise produced by the generator. The improved CycleGAN algorithm surpasses the original model by 16.52 %, 10.06 %, 4.65 %, and 9.37 % when comparing the PSNR, SSIM, UIQM, and UCIQE. It offers robust support for the subsequent detection and analysis of cracks.

Automation in Construction, an international top-tier journal in the field of civil engineering informatics, holds a prominent position, ranking 17th among 242 journals in the Control and Systems Engineering category, 5th among 174 journals in the Building and Construction category, and 9th among 310 journals in the Civil and Structural Engineering category.