摘要
复杂纹理瓷砖表面存在较多的低可视度小目标缺陷与严重的复杂纹理背景干扰,使应用目标检测方法时易出现较高的误检率和漏检率。为提升复杂纹理瓷砖表面缺陷检测效率,提出了基于通道与空间联合注意力的复杂纹理瓷砖表面缺陷检测方案。首先通过建模深浅层特征通道间关系设计了一种选择性特征融合方法,以提升模型对小目标缺陷的特征表达;其次,提出了通道与空间联合注意力模块,通过通道注意力和空间注意力来筛选关键特征通道和抑制纹理区域,使模型着重于学习缺陷特征以增强模型辨别缺陷与纹理的能力;最后,在复杂纹理瓷砖表面缺陷数据上进行了实验验证。实验结果表明,相较于AFF(attentional feature fusion)和CBAM(convolutional block attention module)方法,选择性特征融合方法和通道与空间联合注意力模块使模型检测性能分别提高了5.3 AP、6.32 AP。最终,实验证明了该方案分别优于现有的瓷砖检测方法YOLOv5和纹理织物缺陷检测AFAM方法1.32 AP、2.12 AP。
In the complex texture of the tile surface,there are more low-visibility small defects,and the interference from the complex textured background is serious.This results high false detection and false alarm rate using traditional object detection methods.To enhance the efficiency of defect detection,this paper proposed a defect detection approach on complex textured tile surfaces based on the joint attention mechanisms of channels and spatial.Firstly,to enhance the feature expression of small defects,it proposed a selective feature fusion method by modeling the relationship between deep and shallow feature channels.Secondly,it designed a joint channel and spatial attention module that selected key feature channels and suppressed texture regions through channel and spatial attention,enabling the model to focus on learning defect features and enhancing its ability to discriminate between defects and texture.Finally,it validated the approach on a dataset of complexly textured cera-mic tile surface defects.The experimental results demonstrate that compared to the AFF and CBAM methods,the selective feature fusion method and channel&spatial joint attention achieved improvements of 5.3 AP and 6.32 AP,respectively.In addition,this paper compared the overall approach with the existing tile detection method YOLOv5 and texture fabric defect detection method AFAM.The results show that it outperforms these methods,with respective improvements of 1.32 AP and 2.12 AP.
作者
叶旭芳
陈梅
李晖
曹阳
王喜宾
Ye Xufang;Chen Mei;Li Hui;Cao Yang;Wang Xibin(State Key Laboratory of Public Big Data,Guiyang 550000,China;School of Computer Science&Technology,Guizhou University,Guiyang 550000,China;School of Mecha-nical Engineering,Guizhou University,Guiyang 550000,China;School of Data Science,Guizhou Institute of Technology,Guiyang 550000,China)
出处
《计算机应用研究》
CSCD
北大核心
2024年第3期944-950,共7页
Application Research of Computers
基金
国家自然科学基金资助项目(62162010,72161005)
贵州省科技资助项目(黔科合支撑[2021]一般449,黔科合基础-ZK[2022]一般184,黔科合支撑[2022]一般271,黔科合成果[2023]一般010)。
关键词
表面缺陷检测
注意力机制
特征融合
目标检测
surface defect detection
attention mechanism
feature fusion
object detection
