基于BS-YOLO11模型的混凝土桥梁裂缝识别研究  

Research on Crack Detection of Concrete Bridges Based on BS-YOLO11 Model

作　　者：周水兴[1] 罗成 周琳淇 ZHOU Shuixing;LUO Cheng;ZHOU Linqi(School of Civil Engineering,Chongqing Jiaotong University,Chongqing 400074)

机构地区：[1]重庆交通大学土木工程学院,重庆400074

出　　处：《公路交通技术》2025年第2期119-126,共8页Technology of Highway and Transport

基　　金：贵州省交通运输厅科技项目(2021-022-123)。

摘　　要：为进一步提高裂缝识别精度和识别率,提出一种新型裂缝识别BS-YOLO11模型,该模型结合了加权双向特征金字塔网络、挤压-激励注意力模块与YOLO11模型三者的优点。为验证该模型的有效性,设计了4组模块消融试验,评估了BiFPN模块与SE模块对YOLO11模型性能的影响,并与原YOLO系列模型的识别效果进行对比。结果表明:YOLO模型中结合BiFPN与SE注意力机制模块,能充分发挥各自优势所带来的性能提升,实现图像增强与目标识别的高效协同,可将识别精度提升到81.4%,BS-YOLO11模型表现出较高的精度和鲁棒性,具有良好的应用前景。To further improve the accuracy and recognition rate of crack detection is still an urgent problem to be solved,a novel crack detection model—BS-YOLO11 model,which combines all the advantages of Weighted Bidirectional Feature Pyramid Network(BiFPN),Squeeze-and-Excitation(SE)attention module,and YOLO11 was proposed.To verify the effectiveness of the model,four sets of module ablation experiments were designed to evaluate the effects of BiFPN module and SE module on the performance of YOLO11 model,and the recognition effect was compared with that of the original YOLO series model.The results indicate that the combination of BiFPN and SE attention mechanism modules can fully leverage their respective advantages and bring about performance improvements.The improved BS-YOLO11 model can achieve efficient collaboration between image enhancement and object detection,with a detection accuracy improved to 81.4%,demonstrating high precision and robustness,and demonstrating its promising application prospects.

关 键 词：混凝土桥梁 裂缝识别 BS-YOLO11 BiFPN模块 挤压-激励注意力模块 深度学习 

分 类 号：U445.71[建筑科学—桥梁与隧道工程]