多注意力机制金字塔池化金手指划痕分割方法

doi:10.3778/j.issn.1002-8331.2106-0307

摘要/Abstract

摘要： 针对传统图像处理方法和基于深度学习的分类模型对金手指表面划痕检测效果不理想的情况，提出了一种多注意力机制金字塔池化方法对金手指表面划痕进行语义分割。采用ResNet50模型获取输入图像的特征图；在金字塔的不同层中将特征图分成大小不同的子区域，然后对每个子区域进行平均池化操作；池化后的特征图加入多种注意力机制来提取关键部分的特征信息，并使用边界细化模块对边缘区域进一步精细化，提高分割准确度。通过上采样，将四种不同尺寸的特征图采用级联的方式对划痕区域进行特征融合；与带有整体信息的特征图拼接后经过卷积操作得到最后的预测结果。实验结果表明，本文采用的方法较其他常用分割模型在MIOU和MPA指标上具有明显提升，分别达到86.03%和94.35%，具有一定的应用价值。

关键词: 金手指, 语义分割, 金字塔池化, 注意力机制, 边界细化模块, MIOU指标

Abstract: Aiming at the situation that the traditional image processing method and the classification model based on deep learning are not ideal for the surface scratch detection of golden finger, a multi-attention mechanism Pyramid pooling method is proposed for semantic segmentation of scratches on the surface of gold fingers. Firstly, the ResNet50 model is used to obtain the feature map of the input image, the feature map is divided into sub-regions of different sizes in different layers of the Pyramid pooling, and then performs an average pooling operation on each sub-region. The pooled feature map adds a variety of attention mechanisms to extract the feature information of the key part, and uses the boundary refinement module to further refine the edge region, and improve segmentation accuracy. Through up-sampling, the feature maps of four different sizes are feature fused using a cascade method. After splicing with the feature maps with overall information, the final prediction result is obtained through convolution operations.The experimental results show that the method used in this paper has a significant improvement in MIOU and MPA indicators compared with other commonly used segmentation models, reaching 86.17% and 94.47% respectively, which has certain application value.

Key words: golden finger, semantic segmentation, pyramid pooling, attention mechanism, boundary refinement module, MIOU mdicator

吴良武, 周永霞, 王宇航, 朱钰萍. 多注意力机制金字塔池化金手指划痕分割方法[J]. 计算机工程与应用, 2023, 59(1): 213-220.

WU Liangwu, ZHOU Yongxia, WANG Yuhang, ZHU Yuping. Multi-Attention Mechanism Pyramid Pooling Method for Scratch Segmentation of Gold Finger[J]. Computer Engineering and Applications, 2023, 59(1): 213-220.

参考文献

[1] LONG J，SHELHAMER E，DARRELL T.Fully convolutional networks for semantic segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition，2015：3431-3440.
[2] RONNEBERGER O，FISCHER P，BROX T.U-Net：Convolutional networks for biomedical image segmentation[C]//Proceedings of International Conference on Medical Image Computing and Computer-Assisted Intervention，2015：234-241.
[3] 刘孝保，张嘉祥，阴艳超，等.主从特征融合驱动的铝型材表面缺陷检测[J/OL].计算机集成制造系统：1-14（2021-02-26）[2021-05-30].http：//kns.cnki.net/kcms/detail/11.5946.TP.20210226.1552.004.html.
LIU X B，ZHANG J X，YIN Y C，et al.Surface defect detection of aluminum profile driven by master-slave feature fusion[J/OL].Computer Integrated Manufacturing Systems：1-14（2021-02-26）[2021-05-30].http：//kns.cnki.net/kcms/detail/11.5946.TP.20210226.1552.004.html.
[4] 任秉银，李智勇，代勇.手机屏幕轻微划痕检测方法[J].哈尔滨工业大学学报，2021，53（1）：29-36.
REN B Y，LI Z Y，DAI Y.Method for detecting minor scratches on mobile phone screen[J].Journal of Harbin Institute of Technology，2021，53（1）：29-36.
[5] BADRINARAYANAN V，KENDALL A，CIPOLLA R.SegNet：A deep convolutional encoder-decoder architecture for scene segmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2015：2481-2495.
[6] ZHAO H S，SHI J，QI X J，et al.Pyramid scene parsing network[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition，2017：6230-6239.
[7] HE K，ZHANG X，REN S，et al.Deep residual learning for image recognition[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition，2016：770-778.
[8] YU F，KOLTUN V.Multi-scale context aggregation by dilated convolutions[J].arXiv：1511.07122，2015.
[9] WOO S，PARK J.CBAM：Convolutional block attention module[C]//Proceedings of European Conference on Computer Vision，2018：3-19.
[10] WANG Q，WU B，ZHU P，et al.ECA-Net：Efficient channel attention for deep convolutional neural networks[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition，2020：11531-11539.
[11] DONG H，SONG K，HE Y，et al.PGA-Net：Pyramid feature fusion and global context attention network for automated surface defect detection[J].IEEE Transactions on Industrial Informatics，2020，16（12）：7448-7458.
[12] SIMONYAN K，ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[C]//Proceedings of International Conference on Learning Representations，2015：1-14.
[13] HU J，SHEN L，SUN G.Squeeze-and-excitation networks[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition，2018：7132-7141.
[14] LI X，SUN X，MENG Y，et al.Dice loss for data-imbalanced NLP tasks[J].arXiv：1911.02855，2019.
[15] 袁伟，许文波，周甜.TopPixelLoss：类别不均衡的遥感影像语义分割损失函数[J].中国空间科学技术，2021（6）：85-90.
YUAN W，XU W B，ZHOU T.TopPixelLoss：A loss function for semantic segmentation of remote sensing image with class imbalance[J].China Space Science and Technology，2021（6）：85-90.
[16] 刘兵，杨娟，汪荣贵，等.结合记忆与迁移学习的小样本学习[J].计算机工程与应用，2022，58（19）：242-249.
LIU B，YANG J，WANG R G，et al.Memory-based transfer learning for few-shot learning[J].Computer Engineering and Applications，2022，58（19）：242-249.
[17] CHEN L，PAPANDREOU G，KOKKINOS I，et al.DeepLab：Semantic image segmentation with deep convolutional nets，atrous convolution，and fully connected CRFs[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2018，40（4）：834-848.