基于改进YOLOv8的钢轨表面缺陷检测算法

Surface Defect Detection Algorithm for Steel Rails Based on Improved YOLOv8

钢轨是铁路轨道组成的主要部件,钢轨的表面缺陷问题严重影响了铁路系统的运行质量和安全。针对实际运用中铁轨表面损伤检测精度和检测效率较低的问题,提出了一种基于改进YOLOv8算法的铁轨表面损伤检测算法。将SE注意力机制添加在主干网络的末端,建立卷积特征通道之间的相互依赖性,提高网络的表示能力;引入空间深度卷积(SPD-Conv)替换YOLOv8网络中的传统卷积模块,通过对每个特征映射进行卷积操作,并保留通道维度中的全部信息,从而提高模型在低分辨率图像和小物体检测方面的性能。结果表明,与YOLOv8m模型相比,改进模型的mAP@0.5和mAP@0.5:0.95分别提高了约6.2%和10.8%,召回率上升了约6.8%,准确率提高了约4.4%,有效提高了检测精度和速度。

Steel rails are the main components of railway tracks,and the surface defects of steel rails seriously affect the quality and safety of railway system operation.In response to the low accuracy and efficiency of rail surface damage detection in practical applications,a rail surface damage detection algorithm based on improved YOLOv8 algorithm is proposed.Add SE attention mechanism at the end of the backbone network,establish the interdependence between convolutional feature channels,and improve the network's representation ability.Introducing Spatial Deep Convolution(SPD Conv)to replace the traditional convolution module in YOLOv8 network,by convolving each feature map and preserving all information in the channel dimension,the performance of the model in low resolution image and small object detection is improved.The results indicate that compared with the YOLOv8m model,the improved model mAP@0.5 and mAP@0.5:0.95 They increased by about 6.2%and 10.8%respectively,with a recall rate of about 6.8%and an accuracy rate of about 4.4%,effectively improving detection accuracy and speed.