基于多尺度注意力和空间通道重构卷积的冲击回波频谱图像分类
作者:
作者单位:

1.华北理工大学 人工智能学院;2.华北理工大学 冶金与能源学院

基金项目:

河北省自然科学(E2012209025) ;河北省创新能力提升计划项目(20557605D);教育部产学合作协同育人项目(230806528022738);华北理工大学研究生创新项目(2023047),


Impact echo spectral image classification based on multi-scale attention and spatial channel reconstruction convolution
Author:
Affiliation:

1.College of Artificial Intelligence,North China University of Science and Technology;2.College of Metallurgy and Energy,North China University of Science and Technology

Fund Project:

Natural Science Foundation of Hebei Province(E2012209025), Hebei Province Innovation Capability Improvement Project (20557605D), Ministry of Education industry-university cooperative education project(230806528022738), Graduate Innovation Project North China University of Science and Technology (2023047).

  • 摘要
  • | |
  • 访问统计
  • |
  • 参考文献
  • | |
  • 引证文献
  • | |
  • 文章评论
    摘要:

    针对传统卷积神经网络对冲击回波信号频谱图像进行分类时,面临卷积神经网络特征提取能力不足和数据集类别不平衡的问题,提出一种基于多尺度注意力和空间通道重构卷积的神经网络模型(Multi-scale Hybrid Attention And Spatial Channel Reconstruction Convolutional Neural Networks,MHA-SCConvNet)。首先设计了多尺度混合注意力(Multi-scale Hybrid Attention,MHA)模块,用于提取不同尺度的频谱图像特征并增强模型对频谱波形关键信息的关注力度。其次,引入空间通道重构卷积(Spatial and Channel reconstruction Convolution,SCConv)模块,通过优化图像特征的表示来降低特征冗余。最后,提出了新的混合损失函数-梯度与分布协调边距损失(Gradient and Distribution Harmonized Margin Loss,GDHM Loss) ,该损失函数在动态情况下同时考虑难分类样本和少数类样本。在自建的数据集上进行了训练与测试,并与AlexNet、VGGNet、GoogLeNet等分类模型对比,MHA-SCConvNet准确率达到94.58%。实验结果表明,MHA-SCConvNet模型能够有效提高冲击回波信号频谱图像分类的准确率和效率。

    Abstract:

    Abstract: When traditional convolutional neural networks are used to classify the spectrum images of impact echo signals, they are faced with the problems of insufficient feature extraction ability, unbalanced data sets and small scale. This paper presents the Multi-scale Hybrid Attention and Spatial Channel Reconstruction Convolutional Neural Network (MHA-SCConvNet). The model initiates with a multi-scale hybrid attention module that extracts spectral features at diverse scales, significantly sharpening the focus on essential waveform information. This is followed by the integration of a Spatial and Channel Reconstruction Convolution (SCConv) module, designed to optimize the representation of image features and reduce redundancy effectively. Additionally, the paper introduces the Gradient and Distribution Harmonized Margin Loss (GDHM Loss), a dynamic hybrid loss function tailored to address the challenges of classifying both hard-to-classify and minority class samples. The MHA-SCConvNet was rigorously evaluated on a proprietary dataset, where it demonstrated a notable accuracy of 94.58%, outperforming established models such as AlexNet, VGGNet, and GoogLeNet. Experimental results validate the MHA-SCConvNet's superior capability in enhancing the accuracy and efficiency of impact echo spectral image classification.

    参考文献
    [1] Commandeur C, Sprik R, Stolk C, et al. Observations and improvements on the impact-echo method used for blast furnace hearth refractory measurement[J]. Ironmaking & Steelmaking, 2023, 50(1): 30-43.
    [2] Xu J, Yu X. Detection of concrete structural defects using impact echo based on deep networks [J]. Journal of Testing and Evaluation, 2021, 49(1): 109-120.
    [3] Dorafshan S, Azari H. Deep learning models for bridge deck evaluation using impact echo [J]. Construction and Building Materials, 2020, 263: 120109.
    [4] Sengupta A, Ilgin Guler S, Shokouhi P. Interpreting impact echo data to predict condition rating of concrete bridge decks: A machine-learning approach [J]. Journal of Bridge Engineering, 2021, 26(8): 04021044.
    [5] 罗炜,薛亚东,贾非,等. 基于深度学习的无砟轨道砂浆层脱空病害识别 [J]. 现代隧道技术, 2021, 58 (S1): 129-136.
    [6] 耿豪劼,刘荣桂,蔡东升,等. 冲击回波法检测混凝土构件内部缺陷大小研究 [J]. 混凝土, 2021, (11): 150-154+160
    [7] Ida Y, Fujita E, Hirose T. Classification of volcano-seismic events using waveforms in the method of k-means clustering and dynamic time warping [J]. Journal of Volcanology and Geothermal Research, 2022, 429: 107616.
    [8] Tehseen R, Farooq M S, Abid A. Earthquake prediction using expert systems: a systematic mapping study [J]. Sustainability, 2020, 12(6): 2420.
    [9] 孙波,杨磊,郭秀梅,等. 基于CNN和SVM混合模型的心电信号识别方法 [J]. 山东农业大学学报(自然科学版), 2020, 51 (02): 283-288.
    [10] Lin B, Wei X, Junjie Z. Automatic recognition and classification of multi-channel microseismic waveform based on DCNN and SVM[J]. Computers & geosciences, 2019, 123: 111-120.
    [11] Li J, Tang S, Li K, et al. Automatic recognition and classification of microseismic waveforms based on computer vision[J]. Tunnelling and Underground Space Technology, 2022, 121: 104327.
    [12] 余刃,谢旭阳,王天舒等.基于深度学习波形图像识别的轴承故障诊断方法[J].海军工程大学学报,2021,33(04):76-82.
    [13] Ullah A, Anwar S M, Bilal M, et al. Classification of arrhythmia by using deep learning with 2-D ECG spectral image representation[J]. Remote Sensing, 2020, 12(10): 1685.
    [14] Junyi J I A, Mingli W U, Qi W. A Recognition Method for Time-domain Waveform Images of Electric Traction Network Overvoltage Based on Deep Learning[C]//IOP Conference Series: Earth and Environmental Science. IOP Publishing, 2021, 645(1): 012073.
    [15] Peng G, Tuo X, Shen T, et al. Recognition of rock micro-fracture signal based on deep convolution neural network inception algorithm[J]. Ieee Access, 2021, 9: 89390-89399.
    [16] 夏景明,邢露萍,谈玲,等. 基于MDM-ResNet的脑肿瘤分类方法 [J]. 南京信息工程大学学报(自然科学版), 2022, 14 (02): 212-219.
    [17] 张凯琳,阎庆,夏懿,等. 基于焦点损失的半监督高光谱图像分类 [J]. 计算机应用, 2020, 40 (04): 1030-1037.
    [18] Li B, Liu Y, Wang X. Gradient harmonized single-stage detector[C]//Proceedings of the AAAI conference on artificial intelligence. 2019, 33(01): 8577-8584.
    [19] Cao K, Wei C, Gaidon A, et al. Learning imbalanced datasets with label-distribution-aware margin loss[J]. Advances in neural information processing systems, 2019, 32.
    [20] 刘忠洋,周杰,陆加新,等. 基于注意力机制的多尺度特征融合图像去雨方法 [J]. 南京信息工程大学学报(自然科学版), 2023, 15 (05): 505-513.
    [21] Li J, Wen Y, He L. Scconv: spatial and channel reconstruction convolution for feature redundancy[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023: 6153-6162.
    [22] 宰清鹏,徐杨. 融合空间与通道重构卷积和注意力的轻量型动物姿态估计 [J/OL]. 计算机工程与应用, 1-14[2024-05-17].
    [23] Woo S, Park J, Lee J Y, et al. Cbam: Convolutional block attention module[C]//Proceedings of the European conference on computer vision (ECCV). 2018: 3-19.
    [24] Park J, Woo S, Lee J Y, et al. Bam: Bottleneck attention module[J]. arXiv preprint arXiv:1807.06514, 2018.
    [25] Zhang H, Zu K, Lu J, et al. EPSANet: An efficient pyramid squeeze attention block on convolutional neural network[C]//Proceedings of the Asian Conference on Computer Vision. 2022: 1161-1177.
    [26] 季长清,高志勇,秦静,等. 基于卷积神经网络的图像分类算法综述 [J]. 计算机应用, 2022, 42 (04): 1044-1049.
    [27] 梁雪慧,程云泽,张瑞杰,等. 基于卷积神经网络的桥梁裂缝识别和测量方法 [J]. 计算机应用, 2020, 40 (04): 1056-1061.
    [28] Zhang H, Wu C, Zhang Z, et al. Resnest: Split-attention networks[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022: 2736-2746.
    [29] Hitawala S. Evaluating resnext model architecture for image classification[J]. arXiv preprint arXiv:1805.08700, 2018.
    相似文献
    引证文献
    网友评论
    网友评论
    分享到微博
    发 布
引用本文

崔博,武冰冰,陈伟,孟庆洪,王晓,黄祺祥.基于多尺度注意力和空间通道重构卷积的冲击回波频谱图像分类[J].南京信息工程大学学报,,():

复制
分享
文章指标
  • 点击次数:22
  • 下载次数: 0
  • HTML阅读次数: 0
  • 引用次数: 0
历史
  • 收稿日期:2024-07-19
  • 最后修改日期:2024-10-05
  • 录用日期:2024-10-08

地址:江苏省南京市宁六路219号    邮编:210044

联系电话:025-58731025    E-mail:nxdxb@nuist.edu.cn

南京信息工程大学学报 ® 2025 版权所有  技术支持:北京勤云科技发展有限公司