注意力引导特征融合的织物表面缺陷检测

Fabric Surface Defect Detection based on Attention-guided Feature Fusion

在复杂图案织物表面缺陷检测任务中,传统分割算法在特征融合时忽略了不同尺度特征的语义差异,从而造成了分割精度的下降。针对这个问题,提出了一种注意力引导特征融合的缺陷检测方法。使用空间注意力模块来抑制纹理背景的干扰,突出缺陷信息;使用自注意力金字塔池化模块整合局部特征和全局特征,进一步增强特征表示;使用通道注意力模块引导网络选择有用特征,实现深层特征与浅层特征更好地进行信息融合。实验结果显示,相较于传统分割算法,本文方法在像素精度、平均像素精度以及均交并比三个指标上均有所提升,证明了算法的有效性。

In the task of detecting defects on the surface of fabrics with complex patterns,traditional segmentation algorithms ignore the semantic differences of different scale features in feature fusion,resulting in a decrease in segmentation accuracy.To solve this problem,this paper proposes a defect detection method based on attention-guided feature fusion.Spatial attention module is used to suppress the interference of the texture background and highlight defect information.Self-attention pyramid pooling module is used to integrate local and global features to further enhance feature representation.Channel attention module is used to guide the network to select useful features,so to achieve better information fusion of deep features and shallow features.The experimental results show that compared with the traditional segmentation algorithm,the proposed method has improved pixel accuracy,average pixel accuracy,and mean intersection over union,which shows the effectiveness of the algorithm.