Traditional metallographic analysis of steel materials is highly dependent on manual experience, and it is difficult to accurately quantify grain boundaries and fine secondary phases. Therefore, a method for intelligent recognition and quantitative characterization of microstructure based on a multi-scale convolutional attention network (MSCA-Net) was proposed. This method integrated multi-scale convolution and attention mechanisms, which significantly improved the accuracy of microstructure recognition. Compared with those of the fully convolutional network (FCN), the pixel accuracy (PA) and frequency weighted intersection over union (FWIoU) of MSCA-Net on the test set were improved by 2.71% and 5.07%, respectively. In addition, OpenCV was applied to quantitatively analyze microstructure features. The results demonstrate that MSCA-Net significantly reduces measurement errors, achieves high-precision automated characterization of microstructure features, and provides a new intelligent method for the analysis of material microstructure.
钢铁材料的力学性能依赖于其显微组织的形态与空间分布,因此,实现显微组织的精准识别与定量表征,对于调控材料组织演化与指导高质量钢铁产品的可控化生产具有重要意义[1-3].目前,钢铁材料的显微组织表征多基于Image J, Image-Pro Plus 6.0等图像处理软件,该过程需要具备专业知识的研究者手动标定晶粒并计算特征参数[4-6].这种依赖人工统计的方法不仅耗时耗力,且测量结果受研究者经验的影响,导致数据可重复性差.
此外,为进一步评估MSCA-Net模型在显微组织识别中的准确性,本文对测试集的显微组织特征进行了精细化统计与对比分析.所统计的精细化显微组织特征包括铁素体面积分数、珠光体面积分数、晶界面积分数、铁素体相平均尺寸、珠光体相平均尺寸、单位面积相界长度和单位面积晶界长度.图16显示了基于FCN,DANet,DeepLabV3+与MSCA-Net模型识别结果所提取的特征数据与真实值之间的均方根误差(root mean square error,RMSE).
YangD P, DuP J, WuD, et al. The microstructure evolution and tensile properties of medium-Mn steel heat-treated by a two-step annealing process[J]. Journal of Materials Science & Technology, 2021, 75: 205-215.
LiHao, LiuYan-hui, WangMiao, et al. Effect of microstructure on mechanical properties of TC11 alloy [J]. Journal of Shaanxi University of Science & Technology, 2025, 43(2): 163-169, 195.
TianHong-lei, WuGuang-yu, WangTian, et al. Influence of microstructure characteristics on the mechanical properties of 6N01-T6 aluminum alloy materials in service [J]. Light Metals, 2022(4): 49-54.
[6]
LiuJ, ArugueteD M, JinschekJ R, et al. The non-oxidative dissolution of galena nanocrystals: insights into mineral dissolution rates as a function of grain size, shape, and aggregation state[J]. Geochimica et Cosmochimica Acta, 2008, 72(24): 5984-5996.
ZhuJi-fei, ZhangLi, XuTao, et al. Quantificational characterization of microstructural parameters of cemented carbides based on Image J software [J]. Materials Science and Engineering of Powder Metallurgy, 2015, 20(1): 26-31.
HuangYe-en, AnJian-hong, ZhangSi-si, et al. Image-Pro Plus and Image J: comparison and application in image analysis of biological tissues [J]. Chinese Journal of Stereology and Image Analysis, 2015, 20(2): 185-196.
[11]
ChowdhuryS A, TaufiqueM F N, WangJ, et al. Automated grain boundary (GB) segmentation and microstructural analysis in 347H stainless steel using deep learning and multimodal microscopy[J]. Integrating Materials and Manufacturing Innovation, 2024, 13(1): 244-256.
ZhangYong-zhi, XinQuan-zhong, WangYong-liang, et al. Research on classification and recognition method of steel metallographic structure based on transfer learning [J]. Materials Reports, 2021, 35(24): 24152-24157.
[14]
MedghalchiS, KortmannJ, LeeS H, et al. Automated segmentation of large image datasets using artificial intelligence for microstructure characterisation and damage analysis[J]. Materials & Design, 2024, 243: 113031.
[15]
HuangZ L, WangX G, HuangL C, et al. CCNet: criss-cross attention for semantic segmentation[C]//2019 IEEE/CVF International Conference on Computer Vision (ICVV). Seoul, 2019: 603-612.
[16]
HuangG, LiuZ, Van Der MaatenL, et al. Densely connected convolutional networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, 2017: 2261-2269.
[17]
ChenL C, ZhuY K, PapandreouG, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//the European Conference on Computer Vision (ECCV). Cham: Springer, 2018: 833-851.
[18]
JangJ, VanD, JangH, et al. Residual neural network-based fully convolutional network for microstructure segmentation[J]. Science and Technology of Welding and Joining, 2020, 25(4): 282-289.
[19]
BanerjeeS, GhoshS K, DattaS, et al. Segmentation of dual phase steel micrograph: an automated approach[J]. Measurement, 2013, 46(8): 2435-2440.
[20]
AzimiS M, BritzD, EngstlerM, et al. Advanced steel microstructural classification by deep learning methods[J]. Scientific Reports, 2018, 8: 2128.
[21]
LiX, CuiL Y, LiJ K, et al. Automation of intercept method for grain size measurement: a topological skeleton approach[J]. Materials & Design, 2022, 224: 111358.
[22]
BowlesA, NogitaK, DarguschM, et al. Grain size measurements in Mg-Al high pressure die castings using electron back-scattered diffraction (EBSD)[J]. Materials Transactions, 2004, 45(11): 3114-3119.
[23]
GajalakshmiK, PalanivelS, NaliniN J, et al. Grain size measurement in optical microstructure using support vector regression[J]. Optik, 2017, 138: 320-327.
[24]
BanerjeeS, ChakrabortiP C, SahaS K. An automated methodology for grain segmentation and grain size measurement from optical micrographs[J]. Measurement, 2019, 140: 142-150.
[25]
FliponB, GrandV, MurgasB, et al. Grain size characterization in metallic alloys using different microscopy and post-processing techniques[J]. Materials Characterization, 2021, 174: 110977.
[26]
GuoM H, LuC Z, HouQ, et al. SegNeXt: rethinking convolutional attention design for semantic segmentation[J]. Advances in Neural Information Processing Systems, 2022, 35: 1140-1156.
[27]
CholletF. Xception: Deep learning with depthwise separable convolutions[C]//2017 IEEE conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, 2017: 1800-1807.
PaszkeA, GrossS, MassaF, et al. PyTorch: an imperative style, high-performance deep learning library[J]. Advances in Neural Information Processing Systems, 2019, 32: 8024-8035.
[30]
PedregosaF, VaroquauxG, GramfortA, et al. Scikit-learn: machine learning in Python[J]. Journal of Machine Learning Research, 2011, 12: 2825-2830.
[31]
BradskiG. The OpenCV library[J]. Software Tools for the Professional Programmer, 2000, 25(11): 120-123.
[32]
LongJ, ShelhamerE, DarrellT. Fully convolutional networks for semantic segmentation[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition. Boston, 2015: 3431-3440.
[33]
FuJ, LiuJ, TianH J, et al. Dual attention network for scene segmentation[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, 2019: 3146-3154.
[34]
ChenL C, ZhuY K, PapandreouG, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//the European Conference on Computer Vision (ECCV). Cham: Springer, 2018: 833-851.