An Unsupervised-Learning-Based Approach for Automated Defect Inspection on Textured Surfaces
The DOI matches its Crossref record; 50 of the 51 references it cites were found in a registry or an index.
Abstract
Automated defect inspection has long been a challenging task especially in industrial applications, where collecting and labeling large amounts of defective samples are usually harsh and impracticable. In this paper, we propose an approach to detect and localize defects with only defect-free samples for model training. This approach is carried out by reconstructing image patches with convolutional denoising autoencoder networks at different Gaussian pyramid levels, and synthesizing detection results from these different resolution channels. Reconstruction residuals of the training patches are used as the indicator for direct pixelwise defect prediction, and the reconstruction residual map generated in each channel is combined to generate the final inspection result. This novel method has two prominent characteristics, which benefit the implementation of automatic defect inspection in practice. First, it is absolutely unsupervised that no human intervention is needed throughout the inspection process. Second, multimodal strategy is utilized in this method to synthesize results from multiple pyramid levels. This strategy is capable of improving the robustness and accuracy of the method. To evaluate this approach, experiments on convergence, noise immunity, and defect inspection accuracy are conducted. Furthermore, comparative tests with some excellent algorithms on actual and simulated data sets are performed. Experimental results demonstrated the effectiveness and superiority of the proposed method on homogeneous and nonregular textured surfaces.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Moreover, AE-based algorithms also demonstrate strong competitiveness in steel surface defect detection, which are reported to be fairly noise-robust.”
From Automated Visual Defect Detection for Flat Steel Surface: A Survey (2020) · cites this paper as [114]
In the citing paper
“Mei et al. [114] utilized convolutional denoising AE network to reconstruct image patches, combined with the reconstruction residual maps, this scheme can reliably learning final detection results, where no manual intervention is needed throughout all the detection process.”
From Automated Visual Defect Detection for Flat Steel Surface: A Survey (2020) · cites this paper as [114]
Citations
Cites
Scikit-learn: Machine Learning in Python
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Müller, A., Nothman, J., Louppe, G., et al. (2012) · arXiv · doi:10.48550/arxiv.1201.0490 · Cited by 71,016 · cited 1×
“The packages involved in the implementation mainly include Keras [47], Theano [48], Scikit-Learn [49], Numpy [50], and Scipy [51].”
MethodThe NumPy Array: A Structure for Efficient Numerical Computation
van der Walt, S., Colbert, S. C., & Varoquaux, G. (2011) · Computing in Science & Engineering · doi:10.1109/mcse.2011.37 · Cited by 9,381 · cited 1×
“The packages involved in the implementation mainly include Keras [47], Theano [48], Scikit-Learn [49], Numpy [50], and Scipy [51].”
Method(2010) · Journal of Machine Learning Research · doi:10.5555/1756006.1953039 · Cited by 4,051 · cited 1×
“This is done by first corrupting the initial input x into x̃ by means of a stochastic mapping x̃ ∼ q(x̃|x) [29].”
MethodPractical Recommendations for Gradient-Based Training of Deep Architectures
Bengio, Y. (2012) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-35289-8_26 · Cited by 2,259 · cited 1×
“As for the learning rate α, a heuristic principle is applied by choosing a value close to the largest learning rate that does not cause divergence [42].”
MethodStatic Analysis of Shape in TensorFlow Programs
The Theano Development Team, Al‐Rfou, R., Alain, G., Almahairi, A., Angermueller, C., Bahdanau, D., Ballas, N., Bastien, F., Bayer, J., Belikov, A., et al. (2016) · arXiv (Cornell University) · doi:10.4230/lipics.ecoop.2020.15 · Cited by 1,589 · cited 1×
“The packages involved in the implementation mainly include Keras [47], Theano [48], Scikit-Learn [49], Numpy [50], and Scipy [51].”
MethodRefaeilzadeh, P., Tang, L., & Liu, H. (2009) · Encyclopedia of Database Systems · doi:10.1007/978-0-387-39940-9_565 · Cited by 1,549
Adaptive deconvolutional networks for mid and high level feature learning
Zeiler, M. D., Taylor, G. W., & Fergus, R. (2011) · 2011 International Conference on Computer Vision · doi:10.1109/iccv.2011.6126474 · Cited by 1,217
Salt-and-pepper noise removal by median-type noise detectors and detail-preserving regularization
Chan, R. H., Chung-Wa, & Nikolova, M. (2005) · IEEE Transactions on Image Processing · doi:10.1109/tip.2005.852196 · Cited by 1,152 · cited 1×
“c) Noise corruption: As stated, the salt and pepper noise [34] is utilized for data corruption in the MSCDAE”
MethodWLD: A Robust Local Image Descriptor
Chen, J., Shan, S., He, C., Zhao, G., Pietikäinen, M., Chen, X., & Gao, W. (2010) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2009.155 · Cited by 1,024
Weimer, D., Scholz-Reiter, B., & Shpitalni, M. (2016) · CIRP Annals · doi:10.1016/j.cirp.2016.04.072 · Cited by 577 · cited 1×
“Weimer et al. [23] and Park et al. [24] also proposed novel deep CNN architectures, which obviously improve the accuracy of automated defect detection.”
Background
Research works printed but not linked: 4
- mario fritz, 2017, The KTH-TIPS and KTH-TIPS2 Image Databases
- jones, 2014, SciPy Open source scientific tools for Python
- al-rfou, 2016, Theano A Python framework for fast computation of mathematical expressions
- chollet, 2017, Keras
Cited by
Items 301–323 of 323
An Adaptive Margin Semi-supervisedLearning for Industrial Surface InspectionPreprint
Zheng, X., Zhou, L., Liu, Y., Xue, A., Kong, Y., & Wang, Y. (2024) · Research Square · doi:10.21203/rs.3.rs-3980581/v1
A Novel Two-Stage Method for Texture Defect Detection
Liu, L., Jiang, M., & Xu, L. (2024) · International Conference on Consumer Electronics and Computer Engineering (ICCECE) · doi:10.1109/iccece61317.2024.10504178
“Mei et al. [9] introduced a CAE-based method for fabric defect detection.”
MethodWang, T. (2024) · Civil-comp conferences · doi:10.4203/ccc.7.8.4
New Energy Vehicle Tire Defect Detection Algorithm Based on Improved YOLOv8
Yang, H. (2024) · International Journal of Computer Science and Information Technology · doi:10.62051/ijcsit.v4n2.15
“Mei et al [3]. constructed a Gaussian pyramid-based CDAE architecture to realize the reconstruction of images with different resolutions.”
BackgroundSelf-citationNeuronal Networks for Visual Inspection of Assembly Completeness and Correctness in Manufacturing
Baykar, A. Ö., Kural, A., & Lambrecht, J. (2024) · Lecture notes in networks and systems · doi:10.1007/978-3-031-76197-3_4
Deep Learning-Based Method for Detecting Surface Defects in Air Filters
Yuan, Y., Lin, Y., & Zhang, T. (2024) · International Conference on Computer, Vision and Intelligent Technology (ICCVIT) · doi:10.1109/iccvit63928.2024.10872413
“GAN methods typically generate defect-free images, known as golden images, and compare them with unknown samples for pixel-level defect localization [7].”
Generative AI for Improved Defect Detection in Semiconductor Wafers
Huynh, T., Truong, D., & Bandaragoda, T. (2024) · Annual Conference of the IEEE Industrial Electronics Society (IECON) · doi:10.1109/iecon55916.2024.10905182
“The second, based on unsupervised learning [8] , focuses on spotting defects patterns in wafer images during inspection without human guidance.”
Pan, Q., Liu, T., Hou, Y., Chang, J., & Yang, H. S. (2025) · Neural Computing and Applications · doi:10.1007/s00521-025-11124-6
Self-citationUnsupervised representation learning for wafer defect recognition
Geng, S., Wang, F., & Liu, H. (2025) · Robotic Intelligence and Automation · doi:10.1108/ria-04-2023-0043
Optical Model of Mechanical Surface Inspection Based on Machine Vision
(2025) · International journal for housing science and its applications. · doi:10.70517/ijhsa463755
Defect Detection in Patterned Wafers Using Dynamic Local Adaptive Sensitivity Adjustment
Cai, X., Chen, Y., Yao, Z., & Wang, Y. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2025.3604128
“Conventional defect detection in industrial applications relies heavily on manual inspection by skilled personnel [4].”
FRD: Feature Reconstruction and Discrimination for Unsupervised Anomaly Localization
Chen, Y., Gong, X., & Tao, X. (2025) · International Conference on Computer Science and Application Engineering (CSAE) · doi:10.1145/3772886.3772888
“To address this limitation, convolutional auto-encoders (CAEs) have been widely adopted [7]-[9].”
InfluentialDual‐task teacher–student networks for unsupervised anomaly detection
Wu, T., Wang, Z., Zhang, H., Lu, S., & Zhang, H. (2025) · Coloration Technology · doi:10.1111/cote.70051
Li, Y., Li, J., Zhao, J., Hu, Y., Yi, C., Cai, M., Qiang, Y., & Zhao, J. (2026) · Lecture notes in computer science · doi:10.1007/978-981-95-5699-1_3
Ma, Y., Hou, W. S., Zhou, Y., Huang, Y., Xie, Y., Ma, H., & Zhang, J. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2026.3654934
“The comparative methods include global threshold method [34], local LBP feature method [35], and LMI feature method [36].”
Manigrasso, F., Calandra, D., Morra, L., & Lamberti, F. (2026) · Engineering Reports · doi:10.1002/eng2.70652
Dynamic visual saliency in image sequences
Maczyta, L. (2020) · theses.fr (ABES) · doi:10.70675/44a50fa3zc71fz4d80za512zcc3021e0fbee
Jiang, D., Li, S., & Sun, Y. (2026) · IEEE Transactions on Semiconductor Manufacturing · doi:10.1109/tsm.2026.3714028
“Conversely, reference-free approaches utilize techniques such as Gabor filters, the Hough transform, or morphological thresholding to directly extract linear features [8], [9].”
Pseudo defect guided meta calibration for few shot industrial defect segmentation
Zhang, F., Yang, W., Yin, Y., Wang, C., & Chen, D. (2026) · Scientific Reports · doi:10.1038/s41598-026-63902-4
Invariant Texture Features to Gray Level Discretization
Yan, Y., Lefcourt, S., Vedula, S. S., Akshintala, V. S., Canto, M. I., Initiative, A. D. N., & Jones, C. K. (2026) · Journal of Imaging Informatics in Medicine · doi:10.1007/s10278-026-02297-0
Ferreira, L. (2025) · Journal of Computer Science and Software Applications
Chen, A., He, T., Zhou, J., Fu, Y., Sun, C., Lu, Y., Wang, X., & Chen, D. (2026) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2026.116509
Seeing the unseen: unsupervised steel surface defect detection via image inpainting autoencoders
Bekiryazıcı, T., Aydemır, G., Avcı, A., Gürkan, H., Hanilçi, C., Akbaş, Y., & Öper, M. (2026) · Signal Image and Video Processing · doi:10.1007/s11760-026-05807-3
Metadata
- Authors
- Mei, S., Yang, H., & Yin, Z.
- Affiliations
- State Key Laboratory of Digital Manufacturing Equipment and Technology (CN)
- Venue
- IEEE Transactions on Instrumentation and Measurement
- Published
- 2018-06
- Volume
- 67
- Issue
- 6
- Pages
- 1266-1277
- ISSN
- 0018-9456 (ISSN-L)
- 1557-9662
- OpenAlex
- W2791709739
- Semantic Scholar
- 38ec86557f805ec27e18e5517a34ca43b2bfa60f
- Cited by
- 308
- References
- 51
- Citations by year
- 2026: 16 · 2025: 41 · 2024: 36 · 2023: 59 · 2022: 63 · 2021: 44 · 2020: 36 · 2019: 11 · 2018: 2
- FWCI
- 19.21
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- IEEE Transactions on Instrumentation and Measurement (publishedVersion)
- Funders
- National Science Foundation of China (51327801, 51475193)
- Major Project Foundation of Hubei Province (2016AAA009)
- Topics
- Physical Sciences › Engineering › Industrial and Manufacturing Engineering › Industrial Vision Systems and Defect Detection
- Physical Sciences › Engineering › Mechanical Engineering › Non-Destructive Testing Techniques
- Physical Sciences › Engineering › Civil and Structural Engineering › Infrastructure Maintenance and Monitoring
- Keywords
- automated defect detection, unsupervised learning, textured surfaces, convolutional denoising autoencoder, Gaussian pyramid, reconstruction residuals, defect localization, multimodal fusion, industrial inspection
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.