基于偏差扩散与双流对比学习的工业异常检测方法

doi:10.16451/j.cnki.issn1003-6059.202511003

摘要
图/表
参考文献(0)
相关文章 (7)

全文: PDF (2452 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要

工业外观异常检测在智能制造与品质管控中具有重要意义,但受限于异常样本稀缺和分布多样化,现有方法在多类别场景中容易出现重建失真或特征混淆,导致异常定位不准确.为此,文中提出基于偏差扩散与双流对比学习的工业异常检测方法,在潜在空间中通过偏差扩散模型直接学习从含噪特征到正常特征的“偏离方向”,结合类别嵌入捕获特定类别的正常结构,并在推理中融合潜在空间域与图像域差异,生成鲁棒的异常热力图.同时,引入双流对比学习,拉大正常样本与异常样本的间隔.设计前景感知的异常合成策略,构造结构性与纹理性的互补异常,为“正常流-合成异常流”的对比学习提供高质量异常样本.在MVTec-AD、VisA数据集上的实验表明,文中方法在图像级和像素级多项指标上均较优,表现出较强的泛化能力.

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	朱圣
	张亚飞
	李华锋

关键词 ：异常检测, 工业视觉, 偏差扩散模型, 双流对比学习

Abstract：

Industrial anomaly detection is crucial to intelligent manufacturing and quality control. However, due to the scarcity and diverse distribution of anomaly samples, existing methods are prone to reconstruction distortion or feature confusion in multi-class scenarios, leading to inaccurate anomaly localization. To address these issues, an industrial anomaly detection method based on deviation diffusion and dual-stream contrastive learning is proposed. The deviation direction from noisy features to normal features in the latent space is directly learned through a deviation diffusion model. The normal structure of specific classes is captured by combining class embedding. A robust anomaly heatmap is generated through fusing the differences between the latent space domain and the image domain during inference. Simultaneously, dual-stream contrastive learning is introduced to widen the representation gap between normal samples and anomalies. Furthermore, a foreground-aware anomaly synthesis strategy is designed to construct complementary anomalies with structural and textural characteristics to provide high-quality anomaly samples for the contrastive learning of the normal flow-synthesized anomaly flow. The experimental results on MVTec-AD and VisA datasets show that the proposed method achieves excellent performance on multiple evaluation metrics and exhibits strong generalization ability.

Key words： Anomaly Detection Industrial Vision Deviation Diffusion Model Dual-Stream Contrastive Learning

收稿日期: 2025-10-18

ZTFLH:

TP311

基金资助:

云南省基础研究项目(No.202301AV070004)资助

通讯作者: 李华锋,博士,教授,主要研究方向为图像处理、计算机视觉.E-mail: hfchina99@163.com.

作者简介: 朱圣,硕士研究生,主要研究方向为计算机视觉、图像处理.E-mail:zhushxxx@163.com.
张亚飞,博士,教授,主要研究方向为图像处理、模式识别.E-mail:zyfeimail@163.com.

引用本文:

朱圣, 张亚飞, 李华锋. 基于偏差扩散与双流对比学习的工业异常检测方法[J]. 模式识别与人工智能, 2025, 38(11): 986-998. ZHU Sheng, ZHANG Yafei, LI Huafeng. Industrial Anomaly Detection Method Based on Deviation Diffusion with Two-Stream Contrastive Learning. Pattern Recognition and Artificial Intelligence, 2025, 38(11): 986-998.

链接本文:

http://manu46.magtech.com.cn/Jweb_prai/CN/10.16451/j.cnki.issn1003-6059.202511003 或 http://manu46.magtech.com.cn/Jweb_prai/CN/Y2025/V38/I11/986

[1] 吕承侃,沈飞,张正涛,等.图像异常检测研究现状综述.自动化学报, 2022, 48(6): 1402-1428.
(LÜ C K, SHEN F, ZHANG Z T, et al. Review of Image Anomaly Detection. Acta Automatica Sinica, 2022, 48(6): 1402-1428.)
[2] AN J, CHO S. Variational Autoencoder Based Anomaly Detection Using Reconstruction Probability[C/OL]. [2025-09-23]. http://dm.snu.ac.kr/static/docs/TR/SNUDM-TR-2015-03.pdf.
[3] SCHLEGL T, SEEBÖCK P, WALDSTEIN S M, et al. f-AnoGAN: Fast Unsupervised Anomaly Detection with Generative Adversarial Networks. Medical Image Analysis, 2019, 54: 30-44.
[4] BERGMAN L, HOSHEN Y. Classification-Based Anomaly Detection for General Data[C/OL]. [2025-09-23]. https://arxiv.org/pdf/2005.02359.
[5] ROTH K, PEMULA L, ZEPEDA J, et al. Towards Total Recall in Industrial Anomaly Detection // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2022: 14298-14308.
[6] HO J, JAIN A, ABBEEL P. Denoising Diffusion Probabilistic Mo-dels // Proc of the 34th International Conference on Neural Informa-tion Processing Systems. Cambridge, USA: MIT Press, 2020: 6840-6851.
[7] BEIZAEE F, LODYGENSKY G A, DESROSIERS C, et al. Correcting Deviations from Normality: A Reformulated Diffusion Model for Multi-class Unsupervised Anomaly Detection // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2025: 19088-19097.
[8] BERGMANN P, FAUSER M, SATTLEGGER D, et al. MVTec AD-A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2019: 9584-9592.
[9] ZOU Y, JEONG J, PEMULA L, et al. SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation[C/OL].[2025-09-23]. https://arxiv.org/pdf/2207.14315.
[10] DEFARD T, SETKOV A, LOESCH A, et al. PaDiM: A Patch Distribution Modeling Framework for Anomaly Detection and Loca-lization // Proc of the International Conference Workshops and Challengers on Pattern Recognition. Berlin, Germany: Springer, 2021: 475-489.
[11] WANG G D, HAN S M, DING E R, et al. Student-Teacher Fea-ture Pyramid Matching for Anomaly Detection[C/OL].[2025-09-23]. https://www.bmva-archive.org.uk/bmvc/2021/assets/papers/1273.pdf.
[12] DENG H Q, LI X Y.Anomaly Detection via Reverse Distillation from One-Class Embedding // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2022: 9727-9736.
[13] GUDOVSKIY D, ISHIZAKA S, KOZUKA K.CFLOW-AD: Real-Time Unsupervised Anomaly Detection with Localization via Conditional Normalizing Flows // Proc of the IEEE/CVF Winter Confe-rence on Applications of Computer Vision. Washington, USA: IEEE, 2022: 1819-1828.
[14] YU J W, ZHENG Y, WANG X, et al. FastFlow: Unsupervised Anomaly Detection and Localization via 2D Normalizing Flows[C/OL].[2025-09-23]. https://arxiv.org/pdf/2111.07677.
[15] LI C L, SOHN K, YOON J, et al. CutPaste: Self-Supervised Lear-ning for Anomaly Detection and Localization // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2021: 9659-9669.
[16] ZAVRTANIK V, KRISTAN M, SKOC?J D. DRAEM: A Discriminatively Trained Reconstruction Embedding for Surface Anomaly Detection // Proc of the IEEE/CVF International Conference on Computer Vision. Washington, USA: IEEE, 2021: 8310-8319.
[17] WYATT J, LEACH A, SCHMON S M, et al. AnoDDPM: Ano-maly Detection with Denoising Diffusion Probabilistic Models Using Simplex Noise // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Washington, USA: IEEE, 2022: 649-655.
[18] MOUSAKHAN A, BROX T, TAYYUB J.Anomaly Detection with Conditioned Denoising Diffusion Models // Proc of the 46th DAGM German Conference on Pattern Recognition. Berlin, Germany: Springer, 2025: 181-195.
[19] YOU Z Y, CUI L, SHEN Y J, et al. UniAD: A Unified Model for Multi-class Anomaly Detection // Proc of the 36th International Conference on Neural Information Processing Systems. Cambridge, USA: MIT Press, 2022: 4571-4584.
[20] LU R Y, WU Y J, TIAN L, et al. Hierarchical Vector Quantized Transformer for Multi-class Unsupervised Anomaly Detection // Proc of the 37th International Conference on Neural Information Processing Systems. Cambridge, USA: MIT Press, 2023: 8487-8500.
[21] JIANG X, CHEN Y, NIE Q, et al. Toward Multi-class Anomaly Detection: Exploring Class-Aware Unified Model against Inter-Class Interference[C/OL].[2025-09-23]. https://arxiv.org/pdf/2403.14213.
[22] HE H Y, ZHANG J N, CHEN H X, et al. A Diffusion-Based Framework for Multi-class Anomaly Detection. Proc of the AAAI Conference on Artificial Intelligence, 2024, 38(8): 8472-8480.
[23] HE H Y, BAI Y H, ZHANG J N, et al. MambaAD: Exploring State Space Models for Multi-class Unsupervised Anomaly Detection // Proc of the 38th International Conference on Neural Information Processing Systems. Cambridge, USA: MIT Press, 2024: 71152-71187.
[24] GAO B B.Learning to Detect Multi-class Anomalies with Just One Normal Image Prompt(OneNIP) // Proc of the 18th European Conference on Computer Vision. Berlin, Germany: Springer, 2024: 454-470.
[25] GU Z P, ZHU B K, ZHU G B, et al. UniVAD: A Training-Free Unified Model for Few-Shot Visual Anomaly Detection // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2025: 15194-15203.
[26] SONG J M, MENG C L, ERMON S. Denoising Diffusion Implicit Models[C/OL]. [2025-09-23]. https://arxiv.org/pdf/2010.02502.
[27] ACHANTA R, HEMAMI S, ESTRADA F, et al. Frequency-Tuned Salient Region Detection // Proc of the IEEE Conference on Com-puter Vision and Pattern Recognition. Washington, USA: IEEE, 2009: 1597-1604.
[28] ROMBACH R, BLATTMANN A, LORENZ D, et al. High-Resolution Image Synthesis with Latent Diffusion Models // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2022: 10674-10685.
[29] LIU Z K, ZHOU Y M, XU Y S, et al. SimpleNet: A Simple Network for Image Anomaly Detection and Localization // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2023: 20402-20411.
[30] ZHANG X, LI S Y, LI X, et al. DeSTSeg: Segmentation Guided Denoising Student-Teacher for Anomaly Detection // Proc of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington, USA: IEEE, 2023: 3914-3923.
[31] MENG S Y, MENG W C, ZHOU Q H, et al. MoEAD: A Parameter-Efficient Model for Multi-class Anomaly Detection // Proc of the 18th European Conference on Computer Vision. Berlin, Germany: Springer, 2024: 345-361.
[32] YAO H, LIU M, YIN Z C, et al. GLAD: Towards Better Reconstruction with Global and Local Adaptive Diffusion Models for Unsupervised Anomaly Detection // Proc of the 18th European Con-ference on Computer Vision. Berlin, Germany: Springer, 2024: 1-17.