Crack detection in track slabs based on an improved YOLOv8 algorithm
Cracks in high-speed railway track slabs pose a severe threat to the safety of vehicle opera-tions.To address the issue of ineffective crack repairs in current maintenance practices,this study pro-poses a multi-class crack detection method based on the YOLOv8-DSC model.First,the Dynamic Snake Convolution(DSC)module is incorporated into the backbone network.Based on this,the Bottleneck structure in C2f is reconstructed and established as the C2f-v1 module,which replaces cer-tain C2f modules in the YOLOv8 backbone network to enhance the extraction of multi-scale detailed features related to ineffective crack repairs.Second,the CBAM attention mechanism is introduced into the neck network to improve the model's focus on critical features,enhancing the transmission of small crack features within the neural network.Third,the SIoU loss function is employed to replace CIoU,reducing the excessive penalization caused by geometric factors and minimizing training interfer-ence,thereby increasing the model's generalization capability for similar cracks.Finally,the proposed method is validated and evaluated in four dimensions:network structure,crack data,classification methods,and environmental conditions.Experimental results demonstrate that,compared to the original YOLOv8 model,the YOLOv8-DSC model significantly reduces both missed and false detections of ineffective re-paired cracks in track slabs.The model achieves a 4.6%increase in mean average precision(mAP)and a 4.0%improvement in recall,demonstrating strong robustness and adaptability under adverse environmental conditions.The method effectively enables accurate detection of ineffective crack repairs in track slabs.
万方数据
维普
