CoChat
Sign Up
Cited by 23Industrial Vision Systems and Defect DetectionNon-Destructive Testing TechniquesInfrastructure Maintenance and Monitoring

Real-time system for automatic cold strip surface defect detection

Open access
Kostenetskiy, Pavel, Alkapov, Rustem, Vetoshkin, Nikita, Chulkevich, Roman, Napolskikh, Ilya, Poponin, Ostap·2019·FME Transactions
DOI 10.5937/fmet1904765k
Open this paper in CoChat — free accountOpen PDF

Abstract

Real-Time System for Automatic Cold Strip Surface Defect DetectionDetection and classification of surface defects of the rolled metal are one of the main tasks for correctly assessing product quality.Historically, these tasks were performed by a human.However, due to a multitude of production factors, such as high rolling rate and temperature of the metal, the results of such human work are rather low.Replacing a human controller with an artificial intelligence system has been relevant for a long time; it is merely necessary within the concept of Industry 4.0.This paper is devoted to the development of the prototype system automatic detection and classification of defects for one of the Iron-and-Steel Works of the Chelyabinsk region in the Russian Federation.The prototype consists of the Preprocessor, Classifier, Server, Database, and User interface.The main focus is on achieving high classification accuracy, which is planned to be obtained through the use of convolutional neural networks.

AI snapshot

Written from the paper’s full text.

TL;DR

The paper describes a prototype Automatic Defect Recognition System (ADRS) for rolled-metal surface defects, composed of preprocessing, classification, server, database, and operator-interface modules. Its CNN classifier reports 98.1% accuracy on six defect classes from the Northeastern University surface defect database.

Key findings

  • The prototype’s current model achieved 98.1% accuracy on test data with 6 classes of defects, with an execution time of 2.1 ms and input dimensions 200x200x3.
  • The classifier was based on an improved LeNet-5 topology using ReLU, dropout, Adam, and categorical cross-entropy; training used the Northeastern University surface defect database with 1800 marked images, 200x200 each, and 300 images per class across six defect types.
  • For hardware sizing, the paper assumes a minimum visible defect size of 0.5 mm, minimum defect pixels of 3, strip width 1650 mm, strip speed 7 m/s, and 5% image overlap.
  • In a comparison table, the current model’s 98.1% accuracy is reported alongside 99% for Yi et al. (7 classes), 88% for Guifang Wu (4 classes), and 90% for Masci (7 classes).

Methods

  • The ADRS prototype includes a Preprocessor, Classifier, Server, Database, and User interface; the preprocessor acquires images, splits them into parts, finds regions of interest, and sends defect-containing parts to classification.
  • Defect detection uses grayscale images, Sobel filtering along both axes, binary thresholding, morphological opening and closing, and contour filtering by area relative to the minimum possible defect area.
  • Classification uses a Keras/TensorFlow CNN with LeNet-5 as a basis, data augmentation via Keras ImageDataGenerator, Z-score normalization, dropout 0.5, Adam optimizer, and categorical cross-entropy; experiments were run on an NVIDIA GeForce GTX 1080 Ti.

Limitations

  • Stated: the dataset of 1800 marked images is described as insufficient for high prediction accuracy, so augmentation was used to prevent overfitting; the paper also says work is underway to collect and augment a training set to further improve classification.
  • Evident: the computational experiments use the public Northeastern University surface defect database and a laboratory test-bed with two static cameras, not the target Chelyabinsk-region production line.
  • Evident: the paper reports classification accuracy and execution time but does not report false-positive/false-negative rates, real-time throughput, or deployment validation on the actual rolling line; the title refers to cold strip, while the experiments are described as using hot rolled metal surfaces.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.5937/fmet1904765k

Cites

25 of 25 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 21 cited sources checked · 0 supported · 0 flagged

  • Findingown data
    § 6. CONCLUSIONS AND FUTURE WORK

    Preliminary experiments on the prototype showed stability of the detection of defects.

    “Preliminary experiments that were carried out on the prototype showed the stability of the detection of defects”
    Quote in the paperSubstantial claim
  • Findingnone given
    § 6. CONCLUSIONS AND FUTURE WORK

    An ADRS system prototype was developed, consisting of the Preprocessor, Classifier, Server, Database, and User interface.

    “ADRS system prototype was developed.It consists of the Preprocessor, Classifier, Server, Database, and User interface.”
    Quote in the paperSubstantial claim
  • Findingown data
    § 4.6 Programming implementation

    The current model's execution time is 2.1 ms.

    “Current model 200x200x3 2.1 ms 98,1% 6”
    Quote in the paperSubstantial claim
  • Findingown data
    § 4.6 Programming implementation

    The current model achieved 98.1% accuracy on 6 classes of defects.

    “Current model 200x200x3 2.1 ms 98,1% 6”
    • Also stated in: § 6. CONCLUSIONS AND FUTURE WORK · plain
      “Preliminary experiments that were carried out on the prototype showed the stability of the detection of defects and rather high accuracy of their classification -98.1% on test data with six classes of defects.”
    Quote in the paperSubstantial claim
  • Premisecitation
    § 1. INTRODUCTION

    Even small changes in the production process can cause the emergence of new types of defects.

    “Even small changes in the production process can cause the emergence of new types of defects [4].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [4] Vetoshkin, N., Valkevich, N. and Kostenetskiy P.: Razrabotka polzovatelskogo interfeysa dlya sistemy avtomaticheskogo kontrolya kachestva poverhnosty listovoy staly s pokrytiem, in: Parallel Computational Technologies (PCT'2018), 2018, p. 400.;

      Not checked in the free run.

  • Classified as probably not a substantial claim. 20
  • May combine two claims. 10

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § 2.11 Computational experiments

    After conduction of experiments, there were found coefficients of the algorithm that allows locating the highest number of defects.

    “After conduction of experiments, there were found coefficients of the algorithm that allows locating the highest number of defects.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data
    § 4.5 Neural network programming representation

    During computational experiments, the Adam optimiser showed the best score.

    “During computational experiments, Adam optimiser [30] showed the best score.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] Ruder, S.: An overview of gradient descent optimization algorithms, -Biometrics, Vol. 55, No. 2, pp. 591-596, Sep. 2016.;

      Not checked in the free run.

  • May combine two claims.

    Premisecitation
    § 1. INTRODUCTION

    Steel surfaces rolled on different rolling mills even of the same plant often differ in appearance, which complicates algorithms for detecting and classifying surface defects.

    “The challenge is since steel surfaces rolled on different rolling mills even of the same plant often differ in appearance.It leads to the complication of the algorithms for detecting and classifying surface defects [1].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [1] Baindorf, I., Anstots, T., Eberle, A., Ernenpuch, L. Holzhauser J.: Optimizatsiya protsessa i kachestva korrozionnostoykoy polosy na osnove sistemy kontrolya poverhnostey - Chernye Met., Vol. 3, pp. 45-56, 2005.;

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationreasoning
    § 2.10 System architecture implementation

    A defect might be distinguished completely after finding its borders by Sobel filter along both axes.

    “A defect then might be distinguished completely after finding its borders by Sobel filer along both axes.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Visible defects have bounds on which pixel brightness of a grayscale image changes dramatically.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 1. INTRODUCTION

    The process of detecting and classifying defects of the surface of metal strips usually occurs in three stages: finding regions of interest by segmentation, extracting features from the regions of interest, and classifying defects based on the detected features.

    “The process of detecting and classifying defects of the surface of metal strips usually occurs in three stages [9]:”
    • Also stated in: § 1. INTRODUCTION · plain
      “• finding regions of interest (ROIs) by segmentation; • extracting features from the regions of interest;”
    • Also stated in: § 1. INTRODUCTION · plain
      “• defects classification based on the detected features.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [9] Neogi, N., Mohanta, D. K. and Dutta, P. K.: Review of vision-based steel surface inspection systems, EURASIP - J. Image Video Process., vol. 2014, no. 1, p. 50, Dec. 2014.; source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given
    § 2.4 Image stitching and blending

    Because of the width of a metal sheet, one camera cannot get a picture of a surface with sufficient resolution for spotting defects.

    “Because of the width of a metal sheet, one camera cannot get a picture of a surface with sufficient resolution for spotting defects.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given
    § 2.3 ADRS installation diagram

    Due to the installation of several separate LED modules, the system has high maintainability.

    “Due to the installation of several separate LED modules, the system has high maintainability.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationcitation
    § 2.4 Image stitching and blending

    A gradient-mixing method was chosen for stitching because of its simplicity in implementation and more hidden stitch than in alpha-channel approach.

    “A gradient-mixing [12] method was chosen for stitching because of its simplicity in implementation and more hidden stitch than in alpha-channel approach.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [12] Rankov, V., Locke, R. J., Edens, R. J., Barber, P. R. and Vojnovic, B.: An algorithm for image stitching and blending - SPIE Newsroom, Vol. 5701, No. March, pp. 190-199, 2005.; source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Interpretationnone given
    § 2.2 Camera Requirements Calculation

    Monochrome cameras reduce the amount of data transmitted from cameras and speed up preprocessing algorithms, as these algorithms typically use monochrome images.

    “Monochrome cameras have been used to reduce the amount of data transmitted from cameras and to speed up preprocessing algorithms, as these algorithms typically use monochrome images.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 4.3 Neural network topology

    ReLU significantly increases the rate of convergence of stochastic gradient descent in comparison with the sigmoid.

    “ReLU significantly increases the rate of con-vergence of stochastic gradient descent in comparison with the sigmoid [26].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [26] Krizhevsky, A., Sutskever, I. and Hinton, G. E.: ImageNet classification with deep convolutional neural networks, - Commun. ACM, Vol. 60, No. 6, pp. 84-90, May 2017.; source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation
    § 1. INTRODUCTION

    Characteristics and classification of defects vary for each specific production and rolling mill and are not regulated by a generally accepted standard.

    “Characteristics and classification of defects vary for each specific production and rolling mill and are not regulated by the generally accepted standard.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [4] Vetoshkin, N., Valkevich, N. and Kostenetskiy P.: Razrabotka polzovatelskogo interfeysa dlya sistemy avtomaticheskogo kontrolya kachestva poverhnosty listovoy staly s pokrytiem, in: Parallel Computational Technologies (PCT'2018), 2018, p. 400.;

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 1. INTRODUCTION

    According to the concept of Industry 4.0, metallurgical production will be represented by a cyber-physical system.

    “According to the concept of Industry 4.0, metallurgical production will be represented by a cyber-physical system [10].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [10] Navickas V., Kuznetsova S. V. Gruzauskas: Cyber-Physical Systems Expression in Industry 4.0 Context - Financial and credit activity: problems of theory and practice, vol. 5, no. 23, pp. 188-197, 2017.;

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 4.1 Classification approaches

    The dataset has six different classes of most typical defects in metal manufacturing: rolled-in scale, patches, crazing, pitted surface, inclusion, and scratches.

    “There are six different classes of most typical defects in metal manufacturing: rolled-in scale, patches, crazing, pitted surface, inclusion, and scratches.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [24] Song, K. and Yunhui, Y.: NEU surface defect database. [Online]. Available: http://faculty.neu.edu.cn/yunhyan/NEU_surface_defect_database.html. [Accessed: 01-Feb-2018].;

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 4.1 Classification approaches

    The method described in [21] has a significant con: resource consumption in terms of hardware requirements.

    “The method is described in [21], but there are some cons, the most significant of which is resource consumption in terms of hardware requirements.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [21] Batsuuri, S., Ahn, J. and Ko, J.: Steel surface defects detection and classification using SIFT and voting strategy, - Int. J. Softw. Eng. Its Appl., Vol. 6, 2012.;

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 4.1 Classification approaches

    Neural networks are suitable for a large class of problems, including defects classification.

    “Its properties are flexible and suitable for the solution of a large class of problems [22], which includes defects classification [5], [23].”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [22] Bettine, F., Ameddah, H. and Manaa, R.: A neural network approach for predicting kinematic errors solutions for trochoidal machining in the matsuura MX-330 five-axis Machine, - FME Trans., Vol. 46, pp. 453-462, 2018.; source

      Not checked in the free run.

    • cites [5] Caleb, P. and Steuer, M.: Classification of surface defects on hot rolled steel using adaptive learning methods, in: KES'2000. Fourth International Conference on Knowledge-Based Intelligent Engineering Systems and Allied Technologies. Proceedings, (Cat. No.00TH8516), 2000, Vol. 1, pp. 103-108.;

      Not checked in the free run.

    • cites [23] Yazdchi, M. R., Mahyari, A. G. and Nazeri, A.: Detection and Classification of Surface Defects of Cold Rolling Mill Steel Using Morphology and Neural Network, in: 2008 International Conference on Computational Intelligence for Modelling Control & Automation, 2008, pp. 1071-1076.; source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 4.1 Classification approaches

    Application of neural network in terms of defect on a metal surface is popular because of its capability to distinguish a lot of features.

    “Application of neural network in terms of defect on a metal surface is popular because of its capability to distinguish a lot of features [5], [23].”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [5] Caleb, P. and Steuer, M.: Classification of surface defects on hot rolled steel using adaptive learning methods, in: KES'2000. Fourth International Conference on Knowledge-Based Intelligent Engineering Systems and Allied Technologies. Proceedings, (Cat. No.00TH8516), 2000, Vol. 1, pp. 103-108.;

      Not checked in the free run.

    • cites [23] Yazdchi, M. R., Mahyari, A. G. and Nazeri, A.: Detection and Classification of Surface Defects of Cold Rolling Mill Steel Using Morphology and Neural Network, in: 2008 International Conference on Computational Intelligence for Modelling Control & Automation, 2008, pp. 1071-1076.; source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation
    § 4.6 Programming implementation

    Yi.L., Li.G., Jiang M. achieved 99% accuracy on 7 classes with 114x114 input and 1.3 ms execution time.

    “Yi.L., Li.G., Jiang M. [31] 114x114 1.3 ms 99% 7”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [31] Yi, L., Li, G. and Jiang, M.: An End-to-End Steel Strip Surface Defects Recognition System Based on Convolutional Neural Networks, - Steel Res. Int., Vol. 88, No. 2, pp. 176-187, Feb. 2017.; source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Interpretationreasoning
    § 4.4 Dataset processing

    Data should be augmented to prevent overfitting.

    “It means that data should be augmented to prevent overfitting.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § 4.6 Programming implementation

    Masci J. achieved 90% accuracy on 7 classes with 150x150 input and 6.2 ms execution time.

    “Masci J. [7] 150x150 6.2 ms 90% 7”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [7] Masci, J., Meier, U., Ciresan, D., Schmidhuber, J. and Fricout, G.: Steel defect classification with Max-Pooling Convolutional Neural Networks, in: The 2012 International Joint Conference on Neural Networks (IJCNN), 2012, pp. 1-6.; source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 4.1 Classification approaches

    Northeastern University provides a dataset for defects classification and detection problems.

    “Northeastern University provides a dataset for defects classification and detection problems [24].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [24] Song, K. and Yunhui, Y.: NEU surface defect database. [Online]. Available: http://faculty.neu.edu.cn/yunhyan/NEU_surface_defect_database.html. [Accessed: 01-Feb-2018].;

      Not checked in the free run.

  • May combine two claims.

    Premisecitation
    § 4.4 Dataset processing

    There are 1800 marked images in the Northeastern University dataset divided into several classes of defects.

    “There are 1800 marked images in dataset divided into several classes of defects.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [24] Song, K. and Yunhui, Y.: NEU surface defect database. [Online]. Available: http://faculty.neu.edu.cn/yunhyan/NEU_surface_defect_database.html. [Accessed: 01-Feb-2018].;

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § 4.1 Classification approaches

    Each image in the dataset is represented in black and white, has the size of 200x200, and every class has 300 images.

    “Each image is represented in black and white, has the size of 200x200, and every class has 300 images.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [24] Song, K. and Yunhui, Y.: NEU surface defect database. [Online]. Available: http://faculty.neu.edu.cn/yunhyan/NEU_surface_defect_database.html. [Accessed: 01-Feb-2018].;

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation
    § 4.6 Programming implementation

    Guifang Wu achieved 88% accuracy on 4 classes with -40 ms execution time.

    “Guifang Wu. [6] -40 ms 88% 4”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [6] Guifang Wu, Hoonsung Kwak, Seyoung Jang, Ke Xu, and Jinwu Xu: Design of online surface inspection system of hot rolled strips, in: 2008 IEEE International Conference on Automation and Logistics, 2008, pp. 2291-2295.; source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given
    § 4.3 Neural network topology

    Many topologies with different parameters were compared according to accuracy, and the best were selected.

    “Many topologies with different parameters were compared (according to the accuracy), and the best are selected.”
    Quote in the paperNot a substantial claim
24 of 24

  • The paper relies on this without support.

    Existing commercial steel surface inspection systems perform defect detection well but have difficulty with classification.

    “At the same time, existing systems perform defect detection well, but still, have difficulty with their classification.”

  • The paper relies on this without support.

    Commercial steel surface inspection systems of high resolution, working with small defects (0.5 mm square) and high speed of the strip (7 m/s), are built on area scan cameras.

    “The developed prototype is focused on the use of area scan cameras, as, according to the authors, commercial steel surface inspection systems of high resolution, working with small defects (0.5 mm square) and high speed of the strip (7 m/s), are built on their basis.”

  • The paper relies on this without support.

    Line scan cameras are usually used only as an auxiliary part of such systems to search for specific types of defects.

    “Line scan cameras are usually used only as an auxiliary part of such systems to search for specific types of defects.”

  • The paper relies on this without support.

    Critical points that are not defects are challenging to find on a metal sheet.

    “This stage can be skipped because critical points that are not defects are challenging to find on a metal sheet.”

  • The paper relies on this without support.

    Visible defects have bounds on which pixel brightness of a grayscale image changes dramatically.

    “The meaning of the defect detection algorithm is visible defects has bounds, on which pixel brightness of a grayscale image changes dramatically.”

  • The paper relies on this without support.

    The architecture and detailed specifications of commercial vision-based steel surface inspection systems are not publicly available.

    “The architecture and detailed specifications of commercial vision-based steel surface inspection systems are not publicly available.”

  • The paper relies on this without support.

    There are very few developers of such systems (Codnex, EES, Matra, Parsytec, Siemens-VAI, and Sipar).

    “However, despite this high need, there are very few developers of such systems (Codnex, EES, Matra, Parsytec, Siemens-VAI, and Sipar).”

  • The paper relies on this without support.

    The need for metallurgical enterprises for automation systems for detecting and classifying surface defects is very high.

    “Thus, the need for metallurgical enterprises for automation systems for detecting and classifying surface defects is very high.”

  • The paper relies on this without support.

    Surface control is usually carried out visually (by a human) at the final stages of production.

    “Surface control is usually carried out visually (by a human) at the final stages of production.”

  • The paper relies on this without support.

    The rolling rate is very high, and there is minimal time to carry out control.

    “However, even here, the rolling rate is very high, and there is minimal time to carry out control.”

  • The paper relies on this without support.

    Often this leads to shipment of low-quality metal to the consumer, which further entails significant financial losses for the manufacturer.

    “Often this leads to shipment of low-quality metal to the consumer, which further entails significant financial losses for the manufacturer.”

  • The paper relies on this without support.

    The backlight modules use up their resource with time, and their brightness decreases.

    “The backlight modules use up their resource with time, and their brightness decreases.”

  • The paper relies on this without support.

    The most-known computer vision library nowadays is OpenCV.

    “The most-known computer vision library nowadays is OpenCVError!Reference source not found..”

  • The paper relies on this without support.

    An algorithm for creating a panorama usually includes a stage of finding special points and matching pictures with it.

    “An algorithm for creating a panorama usually includes a stage of finding special points and matching pictures with it.”

  • The paper relies on this without support.

    The cameras' offset is fixed and known.

    “Besides, the cameras' offset is fixed and known.”

  • The paper relies on this without support.

    Images from all cameras need to be united in one panoramic to ease monitor faults detection process for the operator and to get one complete image of a defect, captured from two and more cameras.

    “Images from all cameras need to be united in one panoramic to ease monitor faults detection process for the operator and to get one complete image of a defect, captured from two and more cameras.”
16 of 16

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?

Cites

Flagged first, then by citations

Research works printed but not linked: 5

  1. Baindorf, I., Anstots, T., Eberle, A., Ernenpuch, L. Holzhauser J.: Optimizatsiya protsessa i kachestva korrozionnostoykoy polosy na osnove sistemy kontrolya poverhnostey - Chernye Met., Vol. 3, pp. 45-56, 2005.;
  2. Lemon, S.: Avtomaticheskiy osmotr poverhnosty polosy dlya obespecheniya sploshnogo kontrolya kachestva v linii, Chernye Met., Vol. 4, pp. 49-54, 2003.;
  3. Stolzenberg, M., Gruber, K., Henkenmayer, H. and Joncourt, K.: Sistemy osmotra poverhnosty listovoy staly: Rezultaty issledovaniy, Chernye Met., Vol. 12, pp. 40-48, 2002.;
  4. Navickas V., Kuznetsova S. V. Gruzauskas: Cyber-Physical Systems Expression in Industry 4.0 Context - Financial and credit activity: problems of theory and practice, vol. 5, no. 23, pp. 188-197, 2017.;
  5. Batsuuri, S., Ahn, J. and Ko, J.: Steel surface defects detection and classification using SIFT and voting strategy, - Int. J. Softw. Eng. Its Appl., Vol. 6, 2012.;

Cited by

Sorted by citations

Related

Authors
Kostenetskiy, P., Alkapov, R., Vetoshkin, N., Chulkevich, R., Napolskikh, I., & Poponin, O.
Affiliations
South Ural State University (RU)
Corresponding author
Kostenetskiy, Pavel
ORCID
Venue
FME Transactions
Publisher
Centre for Evaluation in Education and Science (CEON/CEES)
Published
2019
Volume
47
Issue
4
Pages
765-774
ISSN
  • 1451-2092 (ISSN-L)
  • 2406-128X
DOI
10.5937/fmet1904765k
OpenAlex
W2979325939
Semantic Scholar
2a77bf2e8c48268384a5fef464715a90aa466093
Cited by
23
References
25
Citations by year
2025: 4 · 2024: 4 · 2023: 1 · 2022: 4 · 2021: 8 · 2020: 2
FWCI
2.58
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
FME Transaction (publishedVersion, cc-by)
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
rolled steel, defect classification, real-time systems, convolutional neural networks, industrial inspection, Industry 4.0, steel rolling, machine vision

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

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