Automated Visual Defect Detection for Flat Steel Surface: A Survey
read more
Citations
Visual-Based Defect Detection and Classification Approaches for Industrial Applications-A SURVEY.
Triplet-Graph Reasoning Network for Few-Shot Metal Generic Surface Defect Segmentation
Automatic defect detection and segmentation of tunnel surface using modified Mask R-CNN
Image-Based Surface Defect Detection Using Deep Learning: A Review
Unsupervised Saliency Detection of Rail Surface Defects Using Stereoscopic Images
References
Textural Features for Image Classification
Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks
A comparative study of texture measures with classification based on featured distributions
Fast Discrete Curvelet Transforms
Deep Learning-Based Crack Damage Detection Using Convolutional Neural Networks
Related Papers (5)
A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects
Frequently Asked Questions (15)
Q2. What are the future works mentioned in the paper "Automated visual defect detection for flat steel surface: a survey" ?
Driven by developments of emerging machine learning and improvements of hardware computing power, algorithmic research will develop towards the urgent needs of engineering applications, and more high-quality achievements can be expected to open in the near future. More steel surface defect databases, especially raw images from real-world industrial production line, are urgently expected for enriching diversified and cumulative future research ecology, which will be sure to benefit to explore for a feasible and comparable standard of performance evaluation for distinct defect detection methodologies. Existing challenges to surface defect detection and some potential proposals are investigated from a systematic perspective as follows. Intrinsic priors of production line are suggested to assist the defect detection.
Q3. What are the traditional thresholding methods for flat steel?
The traditional thresholding methods identify defects by comparing the value of image pixels to a fix number and turn the test image into a simple binary frame, which is sensitive to random noises and non-uniform illuminations.
Q4. How did Borselli and Shi modify the Sobel operator?
In order to avoid the false detection, Borselli et al. [27] modified Sobel operator by applying thresholding to convert the grayscale image to binary matrix.
Q5. What are the methods used to detect defects on flat steel surfaces?
Thresholding methods are usually used to separate the defective regions on flat steel surfaces, which have been widely applied in online AVI systems [19, 20].
Q6. What are the three directions that are used to extract the features?
which are horizontal, vertical, and two diagonal directions, so that the features extracted by this method have better visual discrimination.
Q7. What are the common edge detection operators?
Researchers often use local image differentiation technique to obtain edge detection operator, the commonly used edge detection templates for flat steel surface are Kirsch, Sobel and Canny operator.
Q8. What is the role of morphology in image processing?
It has a huge influence on the theory and technology of image processing, especially on shape and structure analysis, which has been widely applied in noise removal [47, 48], feature extraction [49, 50] and image enhancement [51, 52].
Q9. What are the common methods used to detect defects of flat steel?
Statistical approaches are frequently used to detect defects of flat steel surface by evaluating the regular and periodic distribution of pixel intensities.
Q10. What is the way to detect defects in steel?
Using thresholding methods for defect detection directly may be ineffective in low contrast images, so it is necessary to analyze the distribution of image gray level before threshold operation.
Q11. What are the two types of statistical methods used in the article?
In summary, these methods are based on two kinds of fundamental structural properties, regularity and local orientation (anisotropy), both properties have great perceived value.
Q12. What is the significance of the defect image samples?
in Fig. 2(c), somelongitudinal crack image samples of con-casting slabs are presented (512×512 pixel), and this defect type is with high probability of occurrence on continuous casting line, which has great threat to the quality of downstream products.
Q13. What is the purpose of this paper?
To further narrow the scope of [9], that is, concentrate on the vital defect detection process on only flat steel products, this paper attempts to present a first Transactions survey on this focused topic, so as to support the AVI applications for the relevant industrial manufacturing.
Q14. What are the defects on flat steel?
Various defects on flat steel surface are generally caused by mechanical or metallurgical imperfection during the industrial manufacturing.
Q15. What is the way to detect defects?
In order to further accomplish the task of defect detection, it is promising to obtain the optimal thresholds or design smarter dynamic thresholding mechanism.