基于注意力机制的酒瓶裂纹敲击异常声音检测系统

doi:10.3778/j.issn.1002-8331.2408-0401

摘要/Abstract

摘要： 酒类罐装生产线中，酒瓶缺陷的自动准确检测可以减少生产线的成本并降低安全隐患。传统的基于机器视觉酒瓶缺陷检测方法由于玻璃瓶缺陷细微且易反光等原因，检测准确率和效率仍有待提升。针对上述问题，提出一种基于注意力机制的酒瓶裂纹敲击异常声音检测系统。该方法引入了Squeeze-and-Excitation和自注意力的两种注意力机制模块构建网络，强化特征通道间的动态、非线性关系及提高模型从声学特征中同步捕捉到全局和局部信息的能力。同时，引入一种基于合成少数过采样技术的酒瓶敲击声数据扩增方法来合成样本，用于解决训练样本不足且类别不平衡的问题。提出的方法在iFLYTEK A.I.开发者大赛——酒瓶瓶体裂纹敲击检测挑战赛取得了0.989?88的加权得分，并排名第一。

关键词: 卷积神经网络, 注意力机制, 数据过采样, 异常声音检测

Abstract: Automatic and accurate detection of wine bottle defects in the liquor canning production line can reduce production costs and mitigate safety hazards. Due to the subtle nature and reflective properties of glass bottle defects, there is significant room for improving the accuracy and efficiency of machine vision-based wine bottle defect detection methods. To address these issues, this paper proposes an attention and data over-sampling-based system for detecting anomalous knocking sound of wine bottle cracks. This approach introduces two attention mechanism modules: Squeeze-and-Excitation and Conformer, to construct the network structure. These modules enhance the dynamic and non-linear relationships between feature channels and improve the ability to capture global and local information from acoustic features. Additionally, this paper introduces a data augmentation method based on synthetic minority over-sampling technique (SMOTE) to address the issues of insufficient training samples and class imbalance. The proposed method achieves a score of 0.989?88 and ranked first on the public test set of the iFLYTEK A.I. developer competition-wine bottle crack detection challenge.

Key words: convolutional neural network, attention mechanism, data over-sampling, anomalous sound detection

王谋, 白吉生, 黄思维, 李茁, 刘鑫, 杨飞然, 王子腾. 基于注意力机制的酒瓶裂纹敲击异常声音检测系统[J]. 计算机工程与应用, 2025, 61(21): 234-241.

WANG Mou, BAI Jisheng, HUANG Siwei, LI Zhuo, LIU Xin, YANG Feiran, WANG Ziteng. Attention Mechanism Based Anomalous Knocking Sound Detection System for Wine Bottle Crack[J]. Computer Engineering and Applications, 2025, 61(21): 234-241.

参考文献

[1] 罗东亮, 蔡雨萱, 杨子豪, 等. 工业缺陷检测深度学习方法综述[J]. 中国科学: 信息科学, 2022, 52(6): 1002-1039.
LUO D L, CAI Y X, YANG Z H, et al. Survey on industrial defect detection with deep learning[J]. Scientia Sinica (Informationis), 2022, 52(6): 1002-1039.
[2] 孙海丽, 龙翔, 韩兰胜, 等. 工业物联网异常检测技术综述[J]. 通信学报, 2022, 43(3): 196-210.
SUN H L, LONG X, HAN L S, et al. Overview of anomaly detection techniques for industrial Internet of Things[J]. Journal on Communications, 2022, 43(3): 196-210.
[3] 马天雨, 刘思亚, 刘金平, 等. 基于多电流特征形态组合模式挖掘的层冷辊电机故障诊断[J]. 电子学报, 2023, 51(4): 870-878.
MA T Y, LIU S Y, LIU J P, et al. Fault diagnosis of laminar cooling roller motor based on morphological combination patterns mining of multi-current features[J]. Acta Electronica Sinica, 2023, 51(4): 870-878.
[4] LIANG Q, XIANG S, LONG J, et al. Automatic defect inspection system for beer bottles based on deep residual learning[J]. International Journal of Computational Vision and Robotics, 2021, 11(3): 299-314.
[5] ZHANG T, DING B, ZHAO X, et al. LearningADD: machine learning based acoustic defect detection in factory automation[J]. Journal of Manufacturing Systems, 2021, 60: 48-58.
[6] HUANG B, MA S L, WANG P, et al. Research and implementation of machine vision technologies for empty bottle inspection systems[J]. Engineering Science and Technology, an International Journal, 2018, 21(1): 159-169.
[7] 唐伟. 声学缺陷检测中关键技术的研究[D]. 天津: 天津大学, 2018.
TANG W. Research on key technology of acoustic defect detection[D]. Tianjin: Tianjin University, 2018.
[8] 黎牧星, 黄志鸿. 基于极限学习机的玻璃瓶口缺陷检测方法研究[J]. 计算技术与自动化, 2016, 35(4): 117-120.
LI M X, HUANG Z H. Research on beer bottle defect detection method based on extreme learning machine[J]. Computing Technology and Automation, 2016, 35(4): 117-120.
[9] ZHAO X, CAO Y H, ZHANG T, et al. An improve feature selection algorithm for defect detection of glass bottles[J]. Applied Acoustics, 2021, 174: 107794.
[10] WANG Y W, FENG L Z, ZHU J M. Novel artificial bee colony based feature selection method for filtering redundant information[J]. Applied Intelligence, 2018, 48(4): 868-885.
[11] MNASRI Z, ROVETTA S, MASULLI F. Anomalous sound event detection: a survey of machine learning based methods and applications[J]. Multimedia Tools and Applications, 2022, 81(4): 5537-5586.
[12] 李伟红, 王伟冰, 龚卫国. 低信噪比下公共场所异常声音声学特征提取[J]. 声学学报, 2019, 44(5): 934-944.
LI W H, WANG W B, GONG W G. Acoustic features extraction of abnormal sounds in public places with low signal-to-noise ratio[J]. Acta Acustica, 2019, 44(5): 934-944.
[13] 陈秋菊, 徐建国. 优化正交匹配追踪和短时谱估计用于声音识别[J]. 计算机工程与应用, 2020, 56(7): 162-169.
CHEN Q J, XU J G. Optimized orthogonal matching pursuit and short-time spectrum estimation for sound recognition[J]. Computer Engineering and Applications, 2020, 56(7): 162-169.
[14] 许春冬, 刘昊, 闵源, 等. 基于双重注意力的声音事件定位与检测[J]. 计算机工程与应用, 2023, 59(19): 99-105.
XU C D, LIU H, MIN Y, et al. Sound event localization and detection based on dual attention[J]. Computer Engineering and Applications, 2023, 59(19): 99-105.
[15] BAI J S, CHEN J F, WANG M, et al. SSDPT: self-supervised dual-path transformer for anomalous sound detection[J]. Digital Signal Processing, 2023, 135: 103939.
[16] BAI J S, WANG M, CHEN J F. Dual-path transformer for machine condition monitoring[C]//Proceedings of the 2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference. Piscataway: IEEE, 2021: 1144-1148.
[17] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 7132-7141.
[18] GULATI A, QIN J, CHIU C C, et al. Conformer: convolution-augmented transformer for speech recognition[C]//Proceedings of the Interspeech 2020, 2020: 5036-5040.
[19] CHAWLA N V, BOWYER K W, HALL L O, et al. SMOTE: synthetic minority over-sampling technique[J]. Journal of Artificial Intelligence Research, 2002, 16: 321-357.
[20] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[J]. arXiv:1706.03762, 2017.
[21] XU J J, SUN X, ZHANG Z Y, et al. Understanding and improving layer normalization[C]//Proceedings of the Neural Information Processing Systems, 2019.
[22] FITZGERALD D. Harmonic/percussive separation using Median filtering[C]//Proceedings of the 13th International Conference on Digital Audio Effects, 2010.
[23] PARK D S, CHAN W, ZHANG Y, et al. SpecAugment: a simple data augmentation method for automatic speech recognition[C]//Proceedings of the Interspeech 2019, 2019: 2613-2617.
[24] ZHANG H Y, CISSE M, DAUPHIN Y N, et al. Mixup: beyond empirical risk minimization[J]. arXiv:1710.09412,2017.