【摘要】 针对高速公路长期运行过程中出现的道路病害，如层间不良、层间松散和裂缝等问题，笔者提出一种基于注意力机制改进的Pix2Pix探地雷达（GPR）道路病害检测方法。该方法通过在Pix2Pix网络中集成卷积块注意力模块（CBAM）和空间注意力机制（SAM），并使用LSGAN损失函数优化网络结构和训练过程。在仅100张图像的小样本数据集中，提出的方法检测精度达到85.45%，显著优于传统的YOLOv5和Faster RCNN方法。实验结果表明，该方法在小样本情景下具备较高的检测精度与泛化能力，为道路病害的自动化检测提供了新的解决方案。更多还原

【Abstract】 In this study, a road damage detection method based on attention mechanism-improved Pix2Pix ground-penetrating radar(GPR) is proposed to address road defects such as interlayer bonding issues, interlayer looseness, and cracks that occur during longterm highway operation. The method integrates the Convolutional Block Attention Module(CBAM) and Spatial Attention Mechanism(SAM) into the Pix2Pix network, utilizing the LSGAN loss function to optimize the network structure and training process. In a small sample dataset of only 100 images, the proposed method achieves a detection accuracy of 85.45%, significantly outperforming traditional YOLOv5 and Faster R-CNN methods. Experimental results demonstrate that the proposed method exhibits high detection accuracy and generalization ability in small sample scenarios, offering a novel solution for the automated detection of road damage.更多还原