【摘要】 针对纤维板生产中出现的多种缺陷且其检测精度低、漏检率高和定位误差大等问题，提出了一种基于改进YOLOv8算法的纤维板表面缺陷检测模型。在骨干网络中添加卷积注意力机制CBAM(Convolutional Block Attention Module),使模型从不同的维度捕捉并学习缺陷的特征，提高卷积神经网络的特征表达能力；采用加权双向特征金字塔网络(BiFPN)来取代原始的路径聚合网络(PANet)结构，增强不同尺度之间的信息流动和特征融合；针对数据集存在的小部分低质量实例，采用Wise-IoU作为损失函数，提升网络模型的边界框回归性能。结果表明，改进后的模型在精确率、召回率和mAP@0.5分别达到了94.8%、94.6%、95.5%,相较于原始模型mAP@0.5提高了3.3%,验证了该方法在纤维板表面缺陷检测方面的可靠性。更多还原

【Abstract】 In order to solve the problems of low detection accuracy, high missed detection rate and large positioning error in the production of fiberboard, a fiberboard surface defect detection model based on the improved YOLOv8 algorithm was proposed. In this study, the convolutional attention mechanism(CBAM) was added to the backbone network, so that the model could capture and learn the features of defects from different dimensions, and improve the feature expression ability of the convolutional neural network. The Weighted Bidirectional Feature Pyramid Network(BiFPN) was used to replace the original Path Aggregation Network(PANet) structure to enhance the information flow and feature fusion between different scales. For a small number of low-quality instances in the dataset, Wise-IoU is used as the loss function to improve the bounding box regression performance of the network model. The results show that the accuracy, recall and mAP@0.5 of the improved model reach 94.8%, 94.6% and 95.5%, respectively, which is 3.3% higher than that of the original model mAP@0.5, which verifies the reliability of the method in the detection of surface defects of fiberboard.更多还原