Visual inspection and identification method for surface defects of refractory material

Abstract

The invention discloses a refractory material surface defect visual detection and identification method, which belongs to the technical field of refractory material defect detection, and comprises the following steps: S1, image preprocessing: collecting crack image information of a refractory brick, and filtering the image; s2, watershed segmentation: adopting a watershed algorithm to extract aggregate and cracks in the image; s3, crack detection and classification: screening crack areas, and classifying the crack areas; s4, statistical analysis: calculating the crack length and width, and performing performance evaluation; accurate segmentation of the aggregate and the crack is achieved through the watershed algorithm and morphological operation, the crack type and parameters are quantified in combination with the classification criterion, the detection efficiency and accuracy are remarkably improved, and a reliable technical means is provided for refractory material performance evaluation.

Description

Technical Field

[0001] The present invention belongs to the technical field of refractory defect detection, and particularly relates to a visual detection and recognition method for surface defects of refractories. Background Art

[0002] Refractories refer to a class of inorganic non-metallic materials with a refractoriness of not less than 1580 °C. Refractoriness refers to the Celsius temperature at which a refractory cone specimen does not soften and melt down under the action of high temperature without load. However, defining refractories only by refractoriness can no longer comprehensively describe them, and 1580 °C is not absolute. It is now defined that any material whose physical and chemical properties allow it to be used in a high-temperature environment is called a refractory. Refractories are widely used in industrial fields such as metallurgy, chemical industry, petroleum, machinery manufacturing, silicate, and power. They are used in the largest amount in the metallurgical industry, accounting for 50% - 60% of the total output.

[0003] During the working process of refractories, they must meet the actual requirements of the high-temperature thermo-mechanical properties of refractories in extreme working environments. During use, due to repeated rapid cooling and heating changes, thermal stress will continuously generate inside the refractory products. When the stress value generated during high-temperature service is greater than the structural strength of the refractory products, cracking, spalling, or even direct damage will occur.

[0004] In recent years, computer vision technology has been introduced into crack detection, but there are still the following problems:

[0005] Limitations of manual interpretation: The identification of crack types completely depends on the visual interpretation of operators, with technical bottlenecks such as large subjective judgment errors and poor repeatability. It is particularly difficult to accurately classify complex intertwined crack networks.

[0006] Insufficient accuracy of quantitative analysis: Traditional methods use manual measurement tools to count crack lengths, which not only have low efficiency but also cannot obtain in-depth characteristic parameters such as crack topological structures and fractal dimensions.

[0007] Complex crack morphologies: The diversity of crack widths, lengths, and paths makes it difficult for traditional algorithms to accurately extract them.

[0008] Lack of unified classification standards: There is a lack of automated classification methods for the propagation paths of cracks in the matrix, aggregates, or interfaces. Summary of the Invention

[0009] In view of the problems mentioned in the background art, the present invention proposes a visual detection and recognition method for surface defects of refractories, which improves the accuracy of crack detection and classification of refractories based on image processing algorithms.

[0010] Technical Solution: To solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0011] A visual detection and recognition method for surface defects of refractory materials, comprising the following steps:

[0012] S1: Image preprocessing: Collect the crack image information of refractory bricks and filter the images.

[0013] S2: Watershed segmentation: Use the watershed algorithm to extract the aggregates and cracks in the images.

[0014] S3: Crack detection and classification: Screen the crack areas and classify them.

[0015] S 4: Statistical analysis: Calculate the crack length and width and conduct performance evaluation.

[0016] Preferably, the specific process of S1 is as follows:

[0017] S11: Data collection;

[0018] Place the refractory brick to be detected on a clean and tidy plane, place the industrial camera directly above the refractory brick to be detected, take pictures of the refractory brick to be detected, and obtain the surface crack image of the refractory brick to be measured.

[0019] S12: Use a Gaussian filter to smooth the crack image noise and retain the gray distribution characteristics.

[0020] Preferably, the specific content of S12 is as follows:

[0021] Let G(x, y) represent the two-dimensional Gaussian filter in the frequency domain, and the specific calculation formula is:

[0022]

[0023] Among them, (x, y) represents the plane coordinates, and σ represents the standard deviation.

[0024] Preferably, in S2, use the watershed algorithm and use the gradient image as the input image to obtain the edge information of the image, specifically:

[0025]

[0026] Among them, (x, y) represents the coordinates of the pixel, g(x, y) represents the output image, f(x, y) represents the original image, and grad{.} represents the gradient operation.

[0027] Preferably, the specific process of S3 is as follows:

[0028] S31: Contour fitting: Use the minimum circumscribed circle factor to screen the crack areas;

[0029] S32: Crack skeleton extraction: Refine the crack area through morphological operations to obtain the crack skeleton, and remove small branches to fit the main propagation path;

[0030] S33: Crack classification: Classify the cracks according to the positional relationship between the cracks and the aggregates.

[0031] Preferably, in S31, it is determined whether it is a crack through the minimum circumscribed crack circle, and the calculation formula of the minimum circumscribed circle is:

[0032]

[0033] where ω represents the circle factor, S represents the area of the extracted crack, and l represents the diameter of the minimum circumscribed circle.

[0034] Preferably, in S33, the cracks are classified into cracks within the matrix, cracks within the aggregates, and cracks between the aggregates and the matrix.

[0035] Preferably, in S1, the Euclidean distance is used to calculate the crack width and length, specifically:

[0036]

[0037] where x1, y1, x2, and y2 represent the plane coordinates (x1, y1) and (x2, y2).

[0038] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0039] (1) The present invention uses the watershed algorithm and classification criteria to analyze the crack propagation behavior of refractory materials and evaluate the cracks in the image. Compared with manual techniques, the stability and accuracy of automatic calibration are demonstrated by examining previously collected crack data.

[0040] (2) The algorithm of the present invention can determine the type of crack propagation and its length and width, providing an effective method for evaluating the performance and properties of refractory materials. In addition, the proposed algorithm realizes automatic crack classification and calibration, benefiting from efficient and consistent classification criteria. When the proposed algorithm of the present invention combines indexes such as the crack propagation path ratio and crack width, it can be used to quantitatively evaluate the fracture behavior of other types of refractory materials. Description of the Drawings

[0041] Figure 1 is a flowchart of the visual detection and recognition method for surface defects of refractory materials of the present invention. Detailed Embodiments

[0042] The following further clarifies the present invention in conjunction with specific embodiments. The embodiments are implemented on the premise of the technical solution of the present invention. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention.

[0043] In this embodiment, taking refractory bricks as an example, a visual detection and recognition method for surface defects of refractory materials is proposed, which mainly includes the following steps:

[0044] S1: Image preprocessing: Collect the crack image information of refractory bricks and filter the image;

[0045] S11: Data acquisition;

[0046] Place the refractory brick to be detected on a clean and tidy plane, place the industrial camera directly above the refractory brick to be detected, take a picture of the refractory brick to be detected, and obtain the surface crack image of the refractory brick to be detected. Since the surface image of the refractory brick to be detected may be affected by external environmental factors during the acquisition process, there may be noise in the surface image of the refractory brick to be detected. To reduce the impact of noise on the detection of surface defects of refractory bricks, the image is filtered.

[0047] S12: Gaussian filtering: Use a Gaussian filter to smooth the crack image noise and retain the gray distribution characteristics;

[0048] Image filtering is a basic operation in image processing and computer vision. Noise will affect the accurate extraction of key and valuable information from the image, making it difficult to detect the object of interest. To solve this problem, various filtering techniques are used to remove image noise; this operation is called image smoothing. Among them, Gaussian filtering retains more gray distribution characteristics of the image, while median filtering effectively removes noise from various sources, such as salt and pepper noise.

[0049] In this embodiment, 3×3 Gaussian filtering is used to remove most of the noise in the crack image. When adjacent pixels are closer to the central pixel, Gaussian filtering adds more "weight" to the average value. Let G(x, y) represent the two-dimensional Gaussian filter in the frequency domain, and the specific calculation formula is:

[0050]

[0051] where (x, y) represents the plane coordinates and σ represents the standard deviation.

[0052] S2: Watershed segmentation: Use the watershed algorithm to extract the aggregates and cracks in the image;

[0053] This algorithm classifies the image as a topological landform in geography. The gray level of each pixel point in the image can be regarded as the altitude in geology, where the receiving basin is formed by local minima and the influence area. The watershed defines the basin. Usually, the gradient image is used as the input image to obtain the edge information of the image, specifically:

[0054]

[0055] Among them, (x, y) represents the coordinates of a pixel, g(x, y) represents the output image, f(x, y) represents the original image, and grad{.} represents the gradient operation.

[0056] According to the principle of the watershed segmentation algorithm, the bottoms of the model are connected. When the bottom is completely submerged in water, the bottom with a lower gray value is submerged; when a certain amount of water is poured in, the peaks and valleys with higher gray values will remain submerged throughout the model diagram. This process is attributed to the aggregation segmentation of the watershed algorithm.

[0057] S3: Crack detection and classification: Screen the crack area and classify it;

[0058] S31: Contour fitting: Use the minimum circumscribed circle factor to screen the crack area;

[0059] The minimum circumscribed crack circle can be used to determine whether it is a crack. The calculation formula for the minimum circumscribed circle is:

[0060]

[0061] Among them, ω represents the circle factor, S represents the area of the extracted crack, and l represents the diameter of the minimum circumscribed circle. If ω is close to zero, the candidate crack has a high aspect ratio, while a value of ω close to 1 indicates a circle. According to this criterion, the value of ω can be determined in advance. If the value of ω is less than the predetermined value, it can be regarded as a crack. In this embodiment, ω = 0.3 is selected to represent the cracks in the ramming material. Subsequently, image segmentation is performed, and possible image cracks are screened according to their roundness.

[0062] S32: Crack skeleton extraction: Refine the crack area through morphological operations to obtain the crack skeleton, and remove small branches to fit the main propagation path;

[0063] Cracks are usually irregular curves with a certain width, which makes it difficult to directly calculate crack parameters from the crack contour. The skeletonization of the crack area helps to describe crack propagation by accurately depicting the morphology of the crack propagation path, directly calculate the crack propagation length, and greatly simplify the calculation of the crack width. In this embodiment, this morphological operation is used to reconstruct the crack area. However, since there are small branches in the skeleton that affect the fitting of the main direction of the skeleton, the main principle for removing small branches is to remove small branches with the number of pixels between the end point and the main branch less than a certain threshold. In addition, the least squares method is also used in this algorithm to fit the crack skeleton as a curve.

[0064] S33: Crack classification: Classify the cracks according to the positional relationship between the cracks and the aggregates;

[0065] Generally speaking, cracks in refractories can be classified into those propagating within the matrix, within the aggregate, or between the aggregate and the matrix. The crack classification criteria are proposed based on the relationship between the aggregate and the crack, and can be specifically divided into: matrix cracks, interface cracks, and aggregate cracks. Specifically:

[0066] Input layer: Crack contour coordinate input: Obtain a set of discrete coordinate points (x i , i , i , i , i , i , y i ), i ∈ [1, n];

[0067] Primary judgment layer (aggregate distribution detection):

[0068] Judgment condition: Whether the coordinate point (x i , y i ) is located within the aggregate area. If not, it is the matrix propagation path; the crack type is marked as matrix crack. If so, it is the secondary judgment layer (adjacent point aggregate detection);

[0069] Secondary judgment layer (neighborhood aggregate correlation verification);

[0070] Verification object: The aggregate distribution of adjacent points (x i , y i ) under the same x coordinate; if not, it is the interface propagation path, and the crack type is marked as interface crack (crack between the aggregate and the matrix); if so, it is the aggregate propagation path, and the crack type is marked as aggregate crack.

[0071] S4: Statistical analysis: Calculate the crack length and width, and conduct performance evaluation.

[0072] Use the Euclidean distance to calculate the crack width and length, specifically:

[0073]

[0074] Among them, x1, y1, x2, and y2 represent the plane coordinates (x1, y1) and (x2, y2).

[0075] Calculate the crack length by adding the Euclidean distance between adjacent pixels in the crack skeleton. Determine the crack direction by fitting a straight line through every certain adjacent pixel points in the skeleton. Finally, determine the crack width by fitting a straight line in the direction perpendicular to the crack direction and intersecting two points on the crack contour.

[0076] The present invention uses the watershed algorithm and classification criteria to analyze the crack propagation behavior of refractory materials and evaluate the cracks in the images. Compared with manual techniques, the stability and accuracy of automatic calibration are demonstrated by examining the previously collected crack data. The results show that the algorithm can determine the type of crack propagation and its length and width, providing an effective method for evaluating the performance and properties of refractory materials. In addition, the proposed algorithm realizes automatic crack classification and calibration, benefiting from efficient and consistent classification criteria. The results of this study show that the proposed algorithm can be used to quantitatively evaluate the fracture behavior of other types of refractory materials when combined with indicators such as the crack propagation path ratio and crack width.

[0077] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A visual detection and recognition method for surface defects of refractory materials, characterized in that: It includes the following steps: S1: Image preprocessing: Collect the crack image information of the refractory brick and filter the image; S2: Watershed segmentation: Use the watershed algorithm to extract the aggregates and cracks in the image; S3: Crack detection and classification: Screen the crack areas and classify them; S4: Statistical analysis: Calculate the crack length and width and conduct performance evaluation.

2. The visual inspection and recognition method for surface defects of refractory materials according to claim 1, characterized in that: The specific process of S1 is: S11: Data collection; Place the refractory brick to be detected on a clean and tidy plane, place the industrial camera directly above the refractory brick to be detected, and take pictures of the refractory brick to be detected to obtain the surface crack image of the refractory brick to be measured; S12: Use a Gaussian filter to smooth the crack image noise and retain the gray distribution characteristics.

3. The visual inspection and recognition method for surface defects of refractory materials according to claim 2, characterized in that: The specific content of S12 is: Let G(x, y) represent the two-dimensional Gaussian filter in the frequency domain. The specific calculation formula is: where (x, y) represents the plane coordinates and σ represents the standard deviation.

4. The visual inspection and recognition method for surface defects of refractory materials according to claim 1, characterized in that: In S2, use the watershed algorithm and use the gradient image as the input image to obtain the edge information of the image. Specifically: where (x, y) represents the coordinates of the pixel, g(x, y) represents the output image, f(x, y) represents the original image, and grad{.} represents the gradient operation.

5. The visual inspection and recognition method for surface defects of refractory materials according to claim 1, characterized in that: The specific process of S3 is: S31: Contour fitting: Use the minimum circumscribed circle factor to screen the crack areas; S32: Crack skeleton extraction: Refine the crack areas through morphological operations to obtain the crack skeleton, and remove the small branches to fit the main propagation path; S33: Crack classification: Classify the cracks according to the positional relationship between the cracks and the aggregates.

6. The visual inspection and recognition method for surface defects of refractory materials according to claim 1, characterized in that: In S31, determine whether it is a crack through the minimum circumscribed crack circle. The calculation formula of the minimum circumscribed circle is: where ω represents the circle factor, S represents the area of the extracted crack, and l represents the diameter of the minimum circumscribed circle.

7. The visual inspection and recognition method for surface defects of refractory materials according to claim 1, characterized in that: In S33, the cracks are divided into cracks in the matrix, cracks in the aggregates, and cracks between the aggregates and the matrix.

8. The visual inspection and recognition method for surface defects of refractory materials according to claim 1, characterized in that: In S1, the Euclidean distance is used to calculate the crack width and length. Specifically: where x1, y1, x2, and y2 represent the plane coordinates (x1, y1) and (x2, y2).

Images