Aiming atthe problems ofleakage and wrong detection when the YOLOv8 algorithm is applied to the surface defectdetection of strip steel, an improved YOLOv8 algorithm is proposed. For the labels of small targets in the dataset, normalized Gaussian Wasserstein distance(NWD) is added on top ofthe originallossyCIOU, which improves the model's ability to detectdefects ofsmalltargets. FocalModulation is used to replace the spatial pooling pyramid of the YOLOv8 model, which improvesthe expression abilityofmulti-scalefeatureswhile lightweighting. Mobile inverted bottleneck Conv(MBConv) is used to replacethe Conv inC2fto constructa new module C2f-MB, and atthe sametime replace the original C2f-MB with C2f-MB. MB to replace the original C2f module with C2f-MB, which enhances the feature expression ability and multi-scale feature fusion ability. the convolutional block attention module (CBAM) is added in the backbonepartto suppressthe background interference, which can bettercapturethe globalinformation and improve the feature extraction ability of the backbone part. Experiment results show that the improved YOLOv8 algorithm improves mAP@ 0.5 by 3% while decreasing the computation amount, which significantly improves the problems ofmissed detection and wrong detection.