China Beijing Higher Institution Engineering Research Center of Structural Engineering and New Materials,Beijing University of Civil Engineering and Architecture,Beijing 100044,China
In order to enhance intelligent analysis of bridge long-term service performance and solve insufficient accuracy of existing defect detection algorithms in complex scenarios,an improved YOLOv8s-based algorithm(YOLOv8s-RC) was proposed for bridge surface disease detection. The large kernel separable convolutional attention mechanism(LSKA) was introduced into the spatial pyramid pooling-fast module (SPPF) of the backbone network to enhance defect feature extraction capability. The weighted feature fusion principle of the bidirectional feature pyramid network(BiFPN) was applied to optimize feature fusion in the neck network,thereby enhancing fusion effectiveness. The original CIoU loss function was replaced by the SIoU loss function to improve bounding box localization accuracy. Ablation experiments on the CODEBRIM dataset demonstrate that the YOLOv8s-RC model achieves 2.3% higher precision,1.7% higher recall,2.0% higher F1-score,and 1.6% higher mAP@0.5 compared with the baseline model. The results show that this algorithm exhibits stronger detection capabilities for small target defects and weak feature defects. With 12.2×106 parameters and 107.5 FPS inference speed,the model is verified to meet real-time monitoring requirements on lightweight devices. Generalization tests on the DACL10K dataset indicate superior performance over Faster-RCNN,SSD,YOLOv5s and YOLOv8s models in various bridge defect detection scenarios. It is concluded that the YOLOv8s-RC algorithm provides an effective technical solution for identifying bridge surface defects in complex environments.
使用的评价指标有F1得分(F1-score,F1)、全类平均精确度(mean average precision,mAP)、模型参数量(Params)、每秒检测帧数(FPS)。其中以精确率(precision,记为P)和召回率(recall,记为R)计算得出的F1和mAP作为表观病害目标检测任务的最终评价指标。Params反映了在低算力移动设备中实现模型构建的难易程度,通常数值越小对模型所需的计算能力要求越低。相关计算公式为
HeShuanhai, ZhaoXiangmo, MaJian, et al. Review of highway bridge inspection and condition assessment[J]. China Journal of Highway and Transport, 2017, 30(11): 63-80.
ZhuDexiang, LiRui, YangJingjing, et al. The method for earthquake damage prediction of continuous rigid frame bridge based on the disease detection[J]. Science Technology and Engineering, 2012, 12(34): 9420-9424.
JiangShixin, TangChuncheng, YangJianxi, et al. Lightweight detection method of surface damage on concrete bridges based on improved Segformer[J]. China Journal of Highway and Transport, 2024, 37(2): 77-87.
MaYe. Infrared thermal imaging detection and recognition technology for defects in concrete structure[J]. Journal of Highway and Transportation Research and Development. 2017, 34(12): 59-65.
ZhangHairu, WangGuofu, ZhangFaquan, et al. Dielectric spectrum feature vector of ground penetrating radar signal in frequency bands for detecting expressway subgrade diseases[J]. Science Technology and Engineering, 2018, 18(4): 344-348.
[11]
LingT, ZhangL, HuangF, et al. OMHT method for weak signal processing of GPR and its application in identification of concrete micro-crack[J]. Journal of Central South University, 2019, 26(11): 3057-3065.
[12]
ChenS, LaeferD F, ManginaE, et al. UAV bridge inspection through evaluated 3D reconstructions[J]. Journal of Bridge Engineering, 2019, 24(4): 05019001.
[13]
ZhangC, ChangC, JamshidiM. Concrete bridge surface damage detection using a single-stage detector[J]. Computer-Aided Civil and Infrastructure Engineering, 2020, 35(4): 389-409.
[14]
LiuW, AnguelovD, ErhanD, et al. SSD: Single shot multibox detector[C]// Computer Vision-ECCV 2016: 14th European Conference. Amsterdam: Springer International Publishing, 2016: 21-37.
[15]
RedmonJ. You only look once: unified, real-time object detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2016: 91.
[16]
RenS, HeK, GirshickR, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 39(6): 1137-1149.
RenSong, ZhuQianwen, TuXinyue, et al. Lining disease identification of highway tunnel based on deep learning[J]. Journal of Zhejiang University: Engineering Science, 2022, 56(1): 92-99.
LuoHui, YuJunying, TuSuocheng. Pavement defect detection algorithm based on deep learning[J]. Science Technology and Engineering, 2022, 22(13): 5299-5305.
KongLingxin, ChenZiqiang, JinLiangnian, et al. A lightweight road disease detection algorithm fused with deformable convolution[J]. Science Technology and Engineering, 2025, 25(2): 683-694.
[23]
LinJ J, IbrahimA, SarwadeS, et al. Bridge inspection with aerial robots: automating the entire pipeline of visual data capture, 3D mapping, defect detection, analysis, and reporting[J]. Journal of Computing in Civil Engineering, 2021, 35(2): 04020064.
[24]
ZhangY, HuangJ, CaiF. On bridge surface crack detection based on an improved YOLOv3 algorithm[J]. IFAC-Papers Online, 2020, 53(2): 8205-8210.
ZouJunzhi, YangJianxi, LiHao, et al. Bridge apparent damage detection based on the improved YOLOv3 in complex background[J]. Journal of Railway Science and Engineering. 2021, 18(12): 3257-3266.
ZhouQingsong, DongShaojiang, LuoJiayuan, et al. Bridge apparent disease detection based on improved YOLOv3[J]. Journal of Chongqing University, 2022, 45(6): 121-130.
PengYunuo, LiuMin, WanZhi, et al. A dual deep network based on the improved YOLO for fast bridge surface defect detection[J]. Acta Automatica Sinica, 2022, 48(4): 1018-1032.
DongShaojiang, TanHao, LiuChao, et al. Apparent disease detection of bridges using improved YOLOv5s[J]. Journal of Chongqing University, 2024, 47(9): 91-100.
[33]
XiongC, ZayedT, AbdelkaderE M. A novel YOLOv8-GAM-Wise-IoU model for automated detection of bridge surface cracks[J]. Construction and Building Materials, 2024, 414: 135025.
[34]
DosovitskiyA. An image is worth 16x16 words: transformers for image recognition at scale[J]. arXiv Preprint, arXiv: 2010.11929, 2020.
[35]
WanH, GaoL, YuanZ, et al. A novel transformer model for surface damage detection and cognition of concrete bridges[J]. Expert Systems with Applications, 2023, 213: 119019.
[36]
YuG, ZhouX. An improved YOLOv5 crack detection method combined with a bottleneck transformer[J]. Mathematics, 2023, 11(10): 2377.
[37]
SrinivasA, LinT Y, ParmarN, et al. Bottleneck transformers for visual recognition[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2021: 16519-16529.
LauK W, PoL M, RehmanY A U. Large separable kernel attention: rethinking the large kernel attention design in CNN[J]. Expert Systems with Applications, 2024, 236: 121352.
[40]
TanM, PangR, LeQ V. Efficientdet: scalable and efficient object detection[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2020: 10781-10790.
[41]
GevorgyanZ. SIoU loss: More powerful learning for bounding box regression[J]. arXiv Preprint, arXiv: 2205.12740, 2022.
[42]
MundtM, MajumderS, MuraliS, et al. Meta-learning convolutional neural architectures for multi-target concrete defect classification with the concrete defect bridge image dataset[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2019: 11196-11205.
[43]
FlotzingerJ, RöschP J, BramlT. DACL10K: benchmark for semantic bridge damage segmentation[C]// Proceedings ofthe IEEE/CVF Winter Conference on Applications of Computer Vision. New York: IEEE, 2024: 8626-8635.
[44]
WangX, GirshickR, GuptaA, et al. Non-local neural networks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 7794-7803.
[45]
DingX, ZhangX, HanJ, et al. Scaling up your kernels to 31×31: Revisiting large kernel design in cnns[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 11963-11975.
[46]
WooS, ParkJ, LeeJ Y, et al. Cbam: Convolutional block attention module[C]// Proceedings ofThe European Conference on Computer Vision. New York: IEEE, 2018: 3-19.
[47]
LiuZ, MaoH, WuC Y, et al. A convnet for the 2020s[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2022: 11976-11986.
[48]
BochkovskiyA, WangC Y, LiaoH Y M. Yolov4: Optimal speed and accuracy of object detection[J]. arXiv Preprint, arXiv: 2004.10934, 2020.
[49]
ZhengZ, WangP, RenD, et al. Enhancing geometric factors in model learning and inference for object detection and instance segmentation[J]. IEEE Transactions on Cybernetics, 2021, 52(8): 8574-8586.
[50]
SelvarajuR R, CogswellM, DasA, et al. Grad-cam: Visual explanations from deep networks via gradient-based localization[C]// Proceedings of the IEEE International Conference on Computer Vision. New York: IEEE, 2017: 618-626.