The detection of bridge damage is crucial for ensuring public safety and social stability.H owever,traditional manual inspection methods suffer from inefficiency,misdetection,and omission issues.To address these challenges,a bridge damage detection method based on YOLO-Bridge is pro-posed.YOLO-Bridge is an improved model based on YOLOv5 with the following enhancements:1)The introduction of a lightweight up sampling operator CARAFE to enhance the model's ability to ex-tract key features of bridge diseases.2)The use of a bidirectional feature pyramid network BiFPN to improve the model's performance in detecting small targets and fusing multi-scale features.3)The a-doption of a new ECA attention mechanism and C3 module fusion method to strengthen the convolution-al layers'sensitivity to input features.Additionally,a bridge damage dataset was constructed,and data augmentation techniques were employed to improve the model's generalization capability.Experimental results demonstrate that YOLO-Bridge achieves a 6.5％increase in mAP compared to the original YOLOv5.Furthermore,compared with other popular object detection algorithms such as Faster-RC-NN,SSD,YOLOv3,and YOLOv7-tiny,YOLO-Bridge achieves higher detection accuracy while maintaining a lightweight model.
维普
万方数据