基于扩散生成的两阶段工业异常检测方法

曾炜峰; 程良伦; 黄国恒

doi:10.12052/gdutxb.230204

欢迎来到能源电力期刊网！

English Version 期刊登录

您当前的位置：

首页 >

文章列表页 >

基于扩散生成的两阶段工业异常检测方法

计算机科学与技术 | 更新时间：2026-05-27

- 基于扩散生成的两阶段工业异常检测方法
- A Two-stage Industrial Anomaly Detection Method Based on Diffusion Generative Model
- 广东工业大学学报 2025年42卷第2期页码：11-19
- 作者机构：
  
  广东工业大学计算机学院,广东,广州,510006
- 作者简介：
  
  曾炜峰(1998–)，男，硕士研究生，主要研究方向为计算机视觉、图像生成，E-mail：961231790@qq.com
  黄国恒(1985–)，男，副教授，博士，主要研究方向为计算机视觉、模式识别和人工智能，E-mail：kevinwong@gdut.edu.cn
- 基金信息：
  
  广东省重点领域研发计划项目(2018B010109007);佛山市重点领域科技攻关项目(2020001006832)
- DOI：10.12052/gdutxb.230204
  中图分类号： TP391.4
- 纸质出版：2025
- 稿件说明：
移动端阅览
曾炜峰, 程良伦, 黄国恒. 基于扩散生成的两阶段工业异常检测方法[J]. 广东工业大学学报, 2025, 42(2): 11–19.

Zeng Weifeng, Cheng Lianglun, Huang Guoheng. A Two-stage Industrial Anomaly Detection Method Based on Diffusion Generative Model[J]. 2025, 42(2): 11-19.
曾炜峰, 程良伦, 黄国恒. 基于扩散生成的两阶段工业异常检测方法[J]. 广东工业大学学报, 2025, 42(2): 11–19. DOI： 10.12052/gdutxb.230204

Zeng Weifeng, Cheng Lianglun, Huang Guoheng. A Two-stage Industrial Anomaly Detection Method Based on Diffusion Generative Model[J]. 2025, 42(2): 11-19. DOI： 10.12052/gdutxb.230204.

摘要

现有的工业异常检测方法大多是采用图像修复的思路，在训练阶段利用人工合成伪异常样本的方式，将人工缺陷图像和模型修复后的重建图像进行判别，并计算判别的偏差，得到异常区域。然而，人工合成的缺陷图像与实际缺陷语义关联度不高，不能准确覆盖实际缺陷类型，导致模型鲁棒性不足。为了解决这个问题，本文提出了基于扩散生成的两阶段工业异常检测模型——DADNet(Diffusion Anomaly Detection Network)。首先，利用语义引导的异常生成模块合成已知缺陷，并以此作为异常检测的先验信息。第一阶段采用图像重建模型对工件异常区域进行修复。第二阶段则训练基于修复图像与初始图像的判别模型，用来检测异常区域。此外，本文通过联合关注机制聚合特征，进一步强化重建模型的性能。DADNet在公开的不同材质工件数据集上的性能表现都优于现有模型，并在工业异常缺陷检测中更有应用前景。

Abstract

Most existing industrial anomaly detection methods adopt the idea of image restoration

which utilizes artificial synthesis of pseudo anomaly samples in the training stage to discriminate the artificial defect image from the reconstructed images after model restoration

and calculates the deviation of the discrimination to obtain the anomalous region. However

the artificial defect images do not correlate well with the actual defect semantics and cannot accurately cover the actual defect types

resulting in less robustness of the model. In order to solve this problem

this paper proposes a two-stage industrial Diffusion Anomaly Detection network (DADNet). Firstly

a semantically-guided anomaly generation module is utilized to synthesize the known defects

and this is used as the a priori information for anomaly detection. In the first stage

an image reconstruction model is used to repair the abnormal region of the workpiece. In the second stage

a discriminative model based on the repaired image and the initial image is trained and used to detect the anomalous regions. In addition

this paper further enhances the performance of the reconstruction model by aggregating features through the joint attention mechanism. DADNet outperforms the existing models on the publicly available datasets of workpieces with different materials

promising prospect for industrial anomaly defect detection.

关键词

Keywords

references

LI Z H, BAI Q Y, WANG F M, et al . Real-time detection system of rail surface defects based on semantic segmentation[J ] . Computer Engineering and Applications, 2021, 57(12): 248-256.

ZAVRTANIK V, KRISTAN M, SKOČAJ D. DRAEM—A discriminatively trained reconstruction embedding for surface anomaly detection[C] //2021 IEEE/CVF International Conference on Computer Vision (ICCV). Online: IEEE Computer Society, 2021: 8310-8319.

YOUKACHEN S, RUCHANURUCKS M, PHATRAP-OMNANT T, et al . Defect segmentation of hot-rolled steel strip surface by using convolutional auto-encoder and conventional image processing[C ] //2019 10th International Conference of Information and Communication Technology for Embedded Systems (IC-ICTES) . New York: IEEE, 2019: 1-5.

KANG G, GAO S, YU L, et al . Deep architecture for high-speed railway insulator surface defect detection: denoising autoencoder with multitask learning[J ] . IEEE Transactions on Instrumentation and Measurement, 2018, 68(8): 2679-2690.

HUANG J H, WANG Z Y. A research on deep learning object detection algorithm based on feature fusion[J]. Journal of Guangdong University of Technology, 2021, 38(4): 52-58.

CHOW J K, SU Z, WU J, et al . Anomaly detection of defects on concrete structures with the convolutional autoencoder[J ] . Advanced Engineering Informatics, 2020, 45: 101105.

NAKAZAWA T, KULKARNI D V. Anomaly detection and segmentation for wafer defect patterns using deep convolutional encoder–decoder neural network architectures in semi-conductor manufacturing[J]. IEEE Transactions on Semi-conductor Manufacturing, 2019, 32(2): 250-256.

BALZATEGUI J, ECIOLAZA L, MAESTRO-WATSON D. Anomaly detection and automatic labeling for solar cell quality inspection based on generative adversarial network[J]. Sensors, 2021, 21(13): 4361.

TAYEH T, ABURAKHIA S, MYERS R. Distance-based anomaly detection for industrial surfaces using triplet networks[C]//2020 11th IEEE Annual Information Technology, Electronics and Mobile Communication Conference (IEMCON). Online: IEEE, 2020: 0372-0377.

ZAVRTANIK V, KRISTAN M, SKOČAJ D. Reconstruction by inpainting for visual anomaly detection[J]. Pattern Recognition, 2021, 112: 107706.

SCHLÜTER H M, TAN J, HOU B, et al . Natural synthetic anomalies for self-supervised anomaly detection and localization[C ] //European Conference on Computer Vision. Cham: Springer, 2022: 474-489.

ZHAO Z, LI B, DONG R, et al . A surface defect detection method based on positive samples[C ] //PRICAI 2018: Trends in Artificial Intelligence: 15th Pacific Rim International Conference on Artificial Intelligence. Switzerland: Springer International Publishing, 2018: 473-481.

ZHANG D, GAO S, YU L, et al . DefGAN: defect detection GANs with latent space pitting for high-speed railway insulator[J ] . IEEE Transactions on Instrumentation and Measurement, 2020, 70: 1-10.

YU J, KIM D Y, Lee Y, et al . Unsupervised pixel-level road defect detection via adversarial image-to-frequency transform[C ] //2020 IEEE Intelligent Vehicles Symposium (IV) . Stuttgart: IEEE, 2020: 1708-1713.

DEFARD T, SETKOV A, LOESCH A, et al . PaDiM: a patch distribution modeling framework for anomaly detection and localization[C ] //International Conference on Pattern Recognition. Online: Springer: 2021: 475-489.

DENG H, LI X. Anomaly detection via reverse distillation from one-class embedding[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . New Orleans: IEEE, 2022: 9737-9746.

RADFORD A, KIM J, HALLACY C, et al . Learning transferable visual models from natural language supervision [C ] //International Conference on Machine Learning. Online: PMLR: 2021: 8748-8763.

RAUNIYAR A, RAJ A, KUMAR A, et al . Text to image generator with latent diffusion models[C ] //2023 International Conference on Computational Intelligence, Communication Technology and Networking (CICTN) . Ghaziabad: IEEE, 2023: 144-148.

VASWANI A, SHAZEER N, PARMAR N, et al . Attention is all you need[C ] //Neural Information Processing Systems. Long Beach: Curran Associates, 2017: 5998-6008.

WANG Z, BOVIK A C, SHEIKH H R, et al . Image quality assessment: from error visibility to structural similarity[J ] . IEEE Transactions on Image Processing, 2004, 13(4): 600-612.

HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 7132-7141.

BERGMANN P, FAUSER M, SATTLEGGER D, et al . MVTec A D — A comprehensive real-world dataset for unsupervised anomaly detection[C ] //2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Long Beach: IEEE, 2019: 9592-9600.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于生成式样本合成的工件缺陷样本数据增强

多尺度对比嵌入增强的自适应入侵检测模型

基于PSI技术的福州市地面形变监测与影响因素分析

多套餐共享储能模式下考虑功率互济的社区联盟优化运行策略

人工噪声辅助无人机太赫兹安全通信的鲁棒设计