改进的基于Mask R-CNN的碳纤维图像分割方法

被引:2
作者
张娟娟 [1 ]
宋圭辰 [2 ]
刘斌 [1 ]
周舒婕 [1 ]
机构
[1] 陕西科技大学电子信息与人工智能学院
[2] 宁波大学材料科学与工程学院
关键词
图像分割; 碳纤维复合材料; 深度学习; 软化非极大抑制;
D O I
暂无
中图分类号
TP391.41 []; TP183 [人工神经网络与计算]; TB332 [非金属复合材料];
学科分类号
080203 ; 081104 ; 0812 ; 0835 ; 1405 ; 0805 ; 080502 ;
摘要
为提高碳纤维增强复合材料(carbon fiber reinforced composite, CFRC)的扫描电子显微镜(scanning electron microscope, SEM)图像精度,实现纤维图像的自动化分割,提出一种改进的Mask R-CNN(Soft-Mask R-CNN),将压缩和激励网络与残差网络相结合获得一种改进的残差网络(squeeze and excitation residua network, SE-ResNet),以提高特征提取效果;采用群组归一化(group normalization, GN)替代批量归一化(batch normalization, BN),提升小批量网络的性能;利用改进的软化非极大抑制(soft non-maximum suppression, Soft-NMS)进行筛选,提高对粘连目标的检测效果。实验结果表明,与Mask R-CNN相比,Soft-Mask R-CNN能有效提高对碳纤维SEM图像的分割准确率和碳纤维区域边界的分割精度,图像分割的平均精确度、交并比分别为87.2%、90.6%,具有较好的泛化能力和较高的精确度,可为CFRC的性能参数研究提供可靠指导。
引用
收藏
页码:1189 / 1194+1208 +1208
页数:7
相关论文
共 8 条
[1]   基于深度学习的SEM纤维图像分割方法研究 [J].
王雯 ;
张芳 .
中国纤检, 2019, (02) :84-87
[2]   Variational and Deep Learning Segmentation of Very-Low-Contrast X-ray Computed Tomography Images of Carbon/Epoxy Woven Composites [J].
Sinchuk, Yuriy ;
Kibleur, Pierre ;
Aelterman, Jan ;
Boone, Matthieu N. ;
Van Paepegem, Wim .
MATERIALS, 2020, 13 (04)
[3]  
A novel method for multi-scale carbon fiber distribution characterization in cement-based composites[J] . Zheng Tong,Haoyan Guo,Jie Gao,Zhenjun Wang. Construction and Building Materials . 2019
[4]   Hybrid Deep VGG-NET Convolutional Classifier for Video Smoke Detection [J].
Matlani, Princy ;
Shrivastava, Manish .
CMES-COMPUTER MODELING IN ENGINEERING & SCIENCES, 2019, 119 (03) :427-458
[5]  
Squeeze-and-Excitation Networks[J] . Hu Jie,Shen Li,Albanie Samuel,Sun Gang,Wu Enhua. IEEE Transactions on Pattern Analysis and Machine Intelligence . 2019 (8)
[6]   ImageNet Classification with Deep Convolutional Neural Networks [J].
Krizhevsky, Alex ;
Sutskever, Ilya ;
Hinton, Geoffrey E. .
COMMUNICATIONS OF THE ACM, 2017, 60 (06) :84-90
[7]  
Optimisation of the rivet joints of the CFRP composite material and aluminium alloy[J] . W. Hufenbach,L.A. Dobrzański,M. Gude,J. Konieczny,A. Czulak. Journal of Achievements in Materials and Manufacturing Engineering . 2007 (1-2)
[8]  
Batch normalization:Accelerating deep network training by reducing internal covariate shift .2 Ioffe S,Szegedy C. 32nd International Conference on Machine Learning(ICML 2015) . 2015