Metal surface corrosion detection method based on improved otsu algorithm

Abstract

The present application relates to a kind of metal surface corrosion detection method based on improved Otsu algorithm.For the corrosion defect detection of metal surface, the present application proposes an improved Otsu algorithm.Based on the traditional Otsu algorithm segmentation corrosion area is prone to cause part of corrosion area loss, by adding weight k adjustment maximum between-class variance value, obtain corresponding threshold value from the relationship between threshold value and maximum between-class variance value, so as to realize the best segmentation of corrosion area.Comparison of mean square error to traditional Otsu algorithm and improved Otsu algorithm for segmentation, analyze corrosion effect, and according to corrosion condition, introduce corrosion degree, and divide corrosion grade.Experimental results show that the algorithm is simple for metal surface corrosion feature extraction, detection error is small, adaptive range is wide, and detection precision is high.

Description

Technical Field

[0001] This invention provides a metal corrosion detection method based on an improved Otsu algorithm, specifically involving improvements to the segmentation method, evaluation of segmentation effect, and classification of corrosion degree levels. Background Technology

[0002] With the rapid development of the times, metals have gradually become important manufacturing materials used by mankind. More commonly used metals such as iron, aluminum, and copper are widely used in industrial fields such as automobiles, ships, construction, aerospace, and manufacturing.

[0003] At various stages of metal processing, factors such as impure raw materials, improper equipment installation, errors in process design, and improper worker operation can easily lead to deviations in metal products, such as burrs, holes, and cracks. Prolonged use without proper maintenance can also cause corrosion. If these surface defects are not detected before being used in production and processing, they can cause quality problems and shorten product lifespan, or even lead to equipment damage, resulting in significant property losses and, in severe cases, threatening the lives of workers. Therefore, defect detection of metal surfaces is essential in related manufacturing processes to ensure the quality of metal products. The significance of accurately detecting surface defects in metal products lies in: timely rejection of substandard products to ensure the quality of products leaving the factory; analysis of rejected products to determine the causes of defects, providing relevant experience for subsequent defect detection; and laying the groundwork for improvements in subsequent processing techniques and technological upgrades.

[0004] Currently, based on the characteristics of various metallic materials, my country has developed corresponding metal testing equipment, such as ultrasonic, X-ray, and penetrant testing methods. However, considering the potential interference from radiation and noise that these devices may cause, potentially harming the physical and mental health of workers, their large-scale use is not advisable. In routine metal inspection, traditional manual visual inspection is the most common method. However, this method is heavily influenced by subjectivity and is prone to false positives and false negatives.

[0005] In recent years, machine vision has developed rapidly and has been applied in automated production, traffic detection, medical testing, and agricultural production. It has not only improved the efficiency of detection but also greatly improved the accuracy of detection. The traditional Ostu algorithm is generally used for corrosion analysis, but for some corrosion areas where the difference between the corrosion area and the background is not obvious or where there is serious noise interference, there are accuracy problems, which lead to analysis errors. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a metal surface corrosion detection method based on an improved Otsu algorithm.

[0007] The objective of this invention is achieved through the following technical solution: a metal surface corrosion detection method based on an improved Otsu algorithm, characterized by comprising the following steps:

[0008] Step S1: Perform image straightening processing on the acquired corrosion image to simplify the image, reduce the amount of data, and highlight key areas;

[0009] Step S2: Image enhancement of the metal corrosion area improves the contrast between the corrosion area and the metal surface, facilitating subsequent operations;

[0010] The specific process is as follows: Based on the characteristics of the eroded image, the image enhancement is performed using the emphasize() operator in the spatial domain method; a low-pass filter (mean_image) is used for filtering, and the obtained gray value (mean) and the original gray value (orig) are calculated to obtain the gray value (res) according to the following formula: res = round((orig-mean)*Factor) + orig; where: Factor is the contrast intensity, which controls the image enhancement effect by working together with the mean filter template.

[0011] Step S3: Threshold segmentation is performed on the enhanced image based on the improved Otsu algorithm to make the eroded region more consistent with the actual situation.

[0012] The improved Otsu algorithm is based on the original Otsu algorithm with added weights. A weight k is introduced before the inter-class variance value, thereby changing the correspondence between the inter-class variance value and the threshold and obtaining the desired threshold.

[0013] Improved maximum inter-class variance σ′ B 2 The relationship between the original parameters and the parameter can be expressed as: σ′ B 2 =kσ B 2 =kP0P1(μ0-μ1) 2 The derivation of the weight k:

[0014] Establish a dataset of erosion images;

[0015] The image is processed using the Otsu algorithm to obtain the corresponding Otsu threshold and P0.

[0016] Analyze the image to obtain the actual expected threshold, and compare it with the Otsu threshold to distinguish the difference between the Otsu threshold and the expected threshold;

[0017] The maximum inter-class variance σ is obtained. B 2 And obtain the corresponding inter-class variance value σ′ through the expected threshold T'. B 2;

[0018] Comparison σ′ B 2 With σ B 2 The value of k is used to obtain the corresponding weight k, and there is a certain relationship between P0 corresponding to different k values;

[0019] The weight k obtained from the acquired image is compared and analyzed with the corresponding P0 to obtain the corresponding relationship;

[0020] Based on the above analysis, the relationship between k and P0 can be obtained as follows:

[0021] Step S4: Perform feature filtering on the segmented image, filter out noise, and select the contours of the eroded regions.

[0022] Step S5, Morphological Processing: Using the circular closing operation in morphology, isolated eroded regions are connected together to obtain the overall eroded region. The closing operation process connects the voids and isolated points within the region into a unified whole; its principle is to first expand... After corrosion "Θ", the expansion effect is greater than the corrosion effect. Using the closing operation of B on A, represented as A●B, defined as... After the closing operation, the overall eroded area can be obtained by filling and filtering the area.

[0023] For areas with rivets, proceed to step S6; for areas without rivets, proceed directly to step S8.

[0024] Step S6, Rivet Region Extraction: Multiple thresholds are obtained through automatic threshold segmentation to quickly select the rivet outline. Then, the rivet region is completed by morphological closing operations and region filling, and finally the rivet region is obtained through the maximum inscribed circle outline;

[0025] Step S7, Extraction of corrosion area on metal surface around rivet: Subtract the overall rivet corrosion area from the rivet area using difference() to obtain the skin corrosion area around the rivet.

[0026] Step S8, Segmentation Performance Evaluation: To measure the quality of segmentation, mean squared error (MSE) is introduced for evaluation. The segmentation performance is measured by objectively evaluating the probability that the eroded region is incorrectly segmented after thresholding, as shown below: In the formula: W×H represents the total number of pixels in the image; X(i,j) represents the actual number of pixels in the target region; Y(i,j) represents the number of pixels in the region obtained after thresholding. The better the segmentation effect, the closer the MSE value is to 0; the worse the segmentation, the closer the MSE value is to 1.

[0027] The specific process of corrosion degree analysis in step S9 is as follows: The corrosion degree is introduced to divide the corrosion levels. For the part with rivets, let the total corrosion area be A1 and the rivet area be A2. Then the corrosion degree can be expressed as: K = (A1 - A2) / A2 × 100%; The corrosion conditions are divided into three levels. Level I is mild damage, where the rivet area is not larger than the detection area, that is, K is not larger than 1, and simple operations can be carried out for corrosion repair; Level II is moderate damage, where the corrosion area is larger than the detection area but not larger than twice the rivet area, that is, 1 < K ≤ 2. The structural corrosion damage is relatively serious. After removing the corresponding corrosion products, corresponding reinforcement operations are also required; Level III is severe damage, where the corrosion area is too large, that is, K > 2. The repair methods need to be carefully considered, and the rivets need to be considered for removal and replacement.

[0028] The specific process of corrosion degree analysis in step S9 is as follows: The corrosion degree is introduced to divide the corrosion levels. For the part without rivets, let the total corrosion area be A1', and the area of the region with a rivet head diameter of 8.5 mm used as a reference is defined as A2'. Then the corrosion degree can be expressed as: K1 = A1' / A2'; The corrosion conditions are divided into three levels. Level I is mild damage, that is, K1 is not larger than 1, and simple operations can be carried out for corrosion repair; Level II is moderate damage, where the corrosion area is slightly larger than the detection area, that is, 1 < K1 ≤ 2. The structural corrosion damage is relatively serious. After removing the corresponding corrosion products, corresponding reinforcement operations are also required; Level III is severe damage, where the corrosion area is too large, that is, K1 > 2. The repair methods need to be carefully considered.

[0029] The beneficial effects of the present invention: Based on the fact that the traditional Otsu algorithm for segmenting the corrosion area is prone to losing some corrosion areas, by adding the weight k to adjust the maximum between-class variance value, and obtaining the corresponding threshold from the relationship between the threshold and the maximum between-class variance value, the optimal segmentation of the corrosion area is achieved. The segmentation effects of the traditional Otsu algorithm and the improved Otsu algorithm are compared with the mean square error, the corrosion effect is analyzed, and according to the corrosion situation, the corrosion degree is introduced to divide the corrosion levels. The experimental results show that the algorithm has simple extraction of metal surface corrosion features, small detection errors, a wide range of adaptability, and high detection accuracy. Description of the Drawings

[0030] Figure 1 It is a graph of the relationship between the weight k and the foreground area ratio P0.

[0031] Figure 2 It is a flowchart of image processing.

[0032] Figure 3 It is a comparison graph of the corrosion image and image enhancement.

[0033] Figure 4 The segmentation effect diagrams of the traditional Otsu threshold and the improved Otsu threshold.

[0034] Figure 5 This represents the maximum inter-class variance corresponding to the traditional Otsu threshold and the improved Otsu threshold.

[0035] Figure 6 This is the overall corroded area.

[0036] Figure 7 This refers to the extracted rivet area.

[0037] Figure 8 To remove the corroded area after removing the rivets. Detailed Implementation

[0038] A method for detecting corrosion on metal surfaces based on an improved Otsu algorithm, the detection process of which is as follows: Figure 2 As shown, it includes the following steps:

[0039] Step S1: Perform image straightening processing on the acquired corrosion image to simplify the image, reduce the amount of data, and highlight key areas;

[0040] Step S2: Image enhancement of the metal corrosion area improves the contrast between the corrosion area and the metal surface, facilitating subsequent operations. (See attached image) Figure 3 As shown, the image enhancement of the metal corrosion area can improve the details in the corrosion area and increase the image contrast. This invention uses the `emphasize()` operator in the spatial domain method for image enhancement based on the characteristics of the corrosion image. The principle is to use a low-pass filter (mean_image) and calculate the gray value (res) by combining the obtained gray value (mean) and the original gray value (orig) using the following formula: `res = round((orig - mean) * Factor) + orig`. Where: `Factor` is the contrast intensity, which, in conjunction with the mean filter template, controls the image enhancement effect.

[0041] Step S3: Threshold segmentation is performed on the enhanced image based on the improved Otsu algorithm to obtain eroded regions that better reflect the actual situation. (See Appendix) Figure 4 As shown in the figure; the relationship between the traditional Otsu algorithm, the improved Otsu algorithm, and the maximum inter-class variance is shown in the appendix. Figure 5 As shown.

[0042] Threshold segmentation involves dividing an image into different grayscale levels and segmenting the target by setting corresponding grayscale thresholds. This invention improves upon the traditional Otsu algorithm. The traditional Otsu algorithm will be introduced first, followed by the improvement process.

[0043] The traditional Otsu algorithm is derived based on the least squares method, and it is computationally simple and stable. The threshold selected by this algorithm maximizes the difference between the average gray level of the entire image and the average gray level of the foreground and background regions. The specific implementation process is as follows:

[0044] Let ni be the number of pixels with gray value i in the image, and L be the number of gray levels. Then the total number of pixels N in the range [0, L-1] is: The probability p of different gray values ​​appearing is: And pi satisfies: If image pixels are segmented into two classes, C0 and C1, with a threshold T, where C0∈[0,T-1] and C1∈[T,L-1], then the probabilities P0 and P1 of regions C0 and C1 are respectively... The average gray levels μ0 and μ1 of the corresponding regions C0 and C1 are respectively Therefore, the average gray level μ of the entire image can be obtained as follows: The total variance of the two regions is σ B 2 =P0(μ0-μ) 2 +P1(μ1-μ) 2 =P0P1(μ0-μ1) 2 Here, when σ is satisfied... B 2 When the threshold is maximized, the corresponding T∈[0,L-1] is the optimal region segmentation threshold.

[0045] When the contrast between the eroded area and the background area in an eroded image is high, the grayscale histogram often exhibits a bimodal pattern. In this case, the peak ranges of the bimodal patterns represent the eroded area and the background area, respectively. The greater the contrast difference between the two, the more pronounced the peak display effect. The threshold obtained by the Otsu algorithm achieves ideal results in identifying the eroded area. However, when the contrast difference between the eroded area and the background area is not significant, the histogram exhibits a unimodal pattern, with the peak range representing the background grayscale range. The eroded area is not clearly displayed, and the threshold obtained by the Otsu method is to the right of the desired threshold, resulting in a smaller segmented eroded area. Therefore, this invention proposes an improved Otsu algorithm to address this issue by introducing a weight k before the inter-class variance value, thereby changing the correspondence between the inter-class variance value and the threshold to obtain the desired threshold.

[0046] Improved maximum inter-class variance σ′ B 2 The relationship between the original parameters and the parameter can be expressed as: σ′ B 2 =kσ B 2 =kP0P1(μ0-μ1) 2 To obtain the weight k, the following steps are performed:

[0047] Thirty-two images of the corrosion area around the rivets were obtained for analysis.

[0048] The image is processed using the Otsu algorithm to obtain the corresponding Otsu threshold and P0.

[0049] The image is analyzed to obtain the actual expected threshold, and compared with the Otsu threshold to distinguish the difference between the Otsu threshold and the expected threshold.

[0050] Analyze the image to obtain the actual expected threshold, and compare it with the Otsu threshold to distinguish the difference between the Otsu threshold and the expected threshold;

[0051] The maximum inter-class variance σ is obtained. B 2 And obtain the corresponding inter-class variance value σ′ through the expected threshold T'. B 2 ;

[0052] Comparison σ′ B 2 With σ B 2 The value of k is used to obtain the corresponding weight k, and there is a certain relationship between P0 corresponding to different k values;

[0053] By comparing and analyzing the weights k obtained from the 32 sets of images with the corresponding P0, the relationship between k and P0 can be obtained as follows: See the diagram showing the relationship between the two. Figure 1 As shown.

[0054] Step S4: Perform feature filtering on the segmented image, filter out noise, and select the contours of the eroded regions.

[0055] Step S5, Morphological Processing: Using the circular closing operation in morphology, isolated eroded regions are connected together, and unnecessary noise is further removed; see appendix. Figure 6 As shown. For areas with rivets, proceed to step S6; for areas without rivets, skip directly to step S8.

[0056] To eliminate some scattered noise, shape feature filtering is performed on the segmented image obtained by the improved Otsu algorithm. Since the area of ​​the eroded region is larger than that of the noise region, this invention chooses to extract the eroded region through area features. After area feature filtering, the structure of the eroded region is not complete and further mathematical morphological processing is required. The basic idea is to measure the eroded region in the image with a specific region shape, and then perform certain morphological transformations on the eroded region to achieve structural analysis and feature extraction. The most common morphological algorithms are erosion and dilation, and other morphological algorithms are formed by combining erosion and dilation. Directly using dilation or erosion in this paper would interfere with the selection of the target region. Therefore, the closing operation formed by combining dilation and erosion is used to process the eroded region. Through the closing operation, the holes and isolated points inside the region are connected into one, with little impact on the appearance and area of ​​the region. The principle is to first dilate... After corrosion "Θ", the expansion effect is greater than the corrosion effect. Using the closing operation of B on A, represented as A●B, defined as... After the closing operation, the overall eroded area can be obtained by filling and filtering the area.

[0057] For areas with rivets, proceed to step S6; for areas without rivets, proceed directly to step S8.

[0058] Step S6, Rivet Region Extraction: Multiple thresholding processes are performed through automatic threshold segmentation to quickly select the rivet region. Then, morphological closing operations and region filling are used to complete the rivet region. Finally, the rivet region is obtained through the maximum inscribed circle contour (see appendix). Figure 7 As shown;

[0059] Step S7, Extraction of the corrosion area on the metal surface around the rivet: Subtract the overall rivet corrosion area from the rivet area using difference() to obtain the corrosion area of ​​the skin around the rivet, see appendix. Figure 8 As shown;

[0060] Step S8, Segmentation Performance Evaluation: To measure the quality of the segmentation, mean squared error (MSE) is introduced for evaluation. MSE is a measure that reflects the degree of difference between the estimator and the estimated quantity. In this invention, it is mainly used to objectively evaluate the probability that the eroded region is incorrectly segmented after threshold segmentation. Specifically, it is expressed as follows: In the formula: W×H represents the total number of pixels in the image; X(i,j) represents the actual number of pixels in the target region; Y(i,j) represents the number of pixels in the region obtained after thresholding. The better the segmentation effect, the closer the MSE value is to 0; the worse the segmentation, the closer the MSE value is to 1. Ten sets of data were randomly selected to compare the segmentation effects of the traditional Otsu algorithm and the improved Otsu algorithm using the MSE value as a reference. The comparison results are shown in Table 1.

[0061] Table 1

[0062]

[0063] Step S9, Corrosion Degree Analysis: To confirm the corrosion condition of the skin around the rivets, the corrosion degree is introduced to classify the corrosion levels. The area of the overall corrosion region is A1, and the area of the rivet region is A2. Then the corrosion degree can be expressed as: K = (A1 - A2) / A2 × 100%. Through the corrosion degree, the corrosion condition of the present invention is classified into three levels. Level I is mild damage, where the corrosion region is not larger than the rivet region, that is, K is not larger than 1, and simple operations can be carried out for corrosion repair. Level II is moderate damage, where the corrosion region is larger than the rivet region but not larger than twice the rivet region, that is, 1 < K ≤ 2. The structural corrosion damage is relatively serious. After removing the corresponding corrosion products, corresponding reinforcement operations are also required. Level III is severe damage, where the corrosion region is too large, that is, K > 2. At this time, the repair means need to be carefully considered, and if necessary, operations such as removing the rivets and replacing them with new ones will be carried out. The corrosion degree conditions of the skin around the rivets in 10 groups of improved Otsu algorithms are shown in Table 2.

[0064] Table 2

[0065]

[0066] For the part without rivets, let the area of the overall corrosion region be A1', and the area of the region with a rivet head diameter of 8.5 mm used as a reference is defined as A2'. Then the corrosion degree can be expressed as: K1 = A1' / A2'; The corrosion condition is classified into three levels. Level I is mild damage, that is, K1 is not larger than 1, and simple operations can be carried out for corrosion repair; Level II is moderate damage, where the corrosion region is slightly larger than the detection region, that is, 1 < K1 ≤ 2. The structural corrosion damage is relatively serious. After removing the corresponding corrosion products, corresponding reinforcement operations are also required; Level III is severe damage, where the corrosion region is too large, that is, K1 > 2, and the repair means need to be carefully considered.

Claims

1. A method for detecting corrosion on metal surfaces based on an improved Otsu algorithm, characterized in that, It includes the following steps: Step S1: Perform binaryzation processing on the collected corrosion images to simplify the images, reduce the data volume, and highlight the key areas; Step S2: Enhance the image of the metal corrosion area to improve the contrast between the corrosion area and the metal surface for subsequent operations; Step S3: Perform threshold segmentation on the enhanced image based on the improved Otsu algorithm to make the obtained corrosion area more in line with the actual situation; Step S4: Screen the features of the segmented image, filter out the noise, and screen out the corrosion area contour; Step S5: Morphological processing: Use circular closing operation in morphology to connect isolated corrosion areas and further remove redundant noise; for the area with rivets, execute Step S6, for the area without rivets, directly jump to Step S8; Step S6: Rivet area extraction: Obtain multiple thresholds by automatic threshold segmentation to quickly select the rivet contour; then complete the rivet area through closing operation and region filling in morphology, and finally obtain the rivet area through the maximum inscribed circle contour; Step S7: Extraction of the metal surface corrosion area around the rivet: Subtract the overall rivet corrosion area from the rivet area using difference() to obtain the corrosion area of the skin around the rivet; Step S8: Evaluation of the segmentation effect: To measure the quality of the segmentation effect, the mean square error MSE is introduced for evaluation; Step S9: Corrosion degree analysis: Divide the corrosion grade by the corrosion degree to facilitate corrosion repair by workers.

2. The metal surface corrosion detection method based on the improved Otsu algorithm according to claim 1, characterized in that, The process of image enhancement in the metal corrosion region in step S2 is as follows: based on the characteristics of the corrosion image, the emphasize() operator in the spatial domain method is used for image enhancement; low-pass filtering is used, and the obtained gray value mean and the original gray value orig are calculated to obtain the gray value res according to the following formula; In the formula: Factor is the contrast intensity, which, together with the mean filter template, controls the image enhancement effect.

3. The metal surface corrosion detection method based on the improved Otsu algorithm according to claim 1, characterized in that, In Step S3, for the threshold segmentation of the enhanced image based on the improved Otsu algorithm, the implementation process of the threshold segmentation of the enhanced image is as follows: The improved Otsu algorithm is weighted based on the original Otsu algorithm, and a weight k is introduced before the between-class variance value, thereby changing the correspondence between the between-class variance value and the threshold to obtain the expected threshold; Improved maximum inter-class variance The relationship between the original parameters and the given parameters can be expressed as: The derivation of the weight k: Establish a corrosion image dataset; Process the image with the Otsu algorithm to obtain the corresponding Otsu threshold and P0; Analyze the image to obtain the actual expected threshold, compare it with the Otsu threshold, and distinguish the difference between the Otsu threshold and the expected threshold; Obtain the maximum inter-class variance value And obtain the corresponding inter-class variance value through the expected threshold T'. ; contrast and The value of k is used to obtain the corresponding weight k, and there is a certain relationship between P0 corresponding to different k values; Compare and analyze the weight k obtained from the corrosion image dataset with the corresponding P0 to obtain the corresponding relationship.

4. The metal surface corrosion detection method based on the improved Otsu algorithm according to claim 1, characterized in that, Step S5, morphological processing, utilizes the circular closing operation in morphology to connect isolated eroded regions together to obtain the overall eroded region. The closing operation process connects internal cavities and isolated points into a unified whole; the principle is to first expand "⊕" and then erode "Θ," with the expansion effect being greater than the erosion effect. The closing operation is performed on A using B, denoted as A●B, and defined as... After the closing operation, the overall eroded area can be obtained by filling and filtering the area.

5. The metal surface corrosion detection method based on the improved Otsu algorithm according to claim 1, characterized in that, Step S8, segmentation effect evaluation: To measure the quality of the segmentation effect, mean squared error (MSE) is introduced for evaluation. The segmentation effect is measured by objectively evaluating the probability that the eroded region is incorrectly segmented after threshold segmentation, as specifically expressed below: In the formula: W×H is the total number of pixels in the image; X(i,j) is the actual number of pixels in the target region; Y(i,j) is the number of pixels in the region obtained after thresholding; the better the segmentation effect, the closer the MSE value is to 0; the worse the segmentation, the closer the MSE value is to 1.

6. The metal surface corrosion detection method based on the improved Otsu algorithm according to claim 1, characterized in that, The specific process of the corrosion degree analysis in step S9 is as follows: The corrosion degree is introduced to divide the corrosion level. For the part with rivets, if the overall corrosion area is A1 and the rivet area is A2, the corrosion degree can be expressed as: ; The corrosion condition is divided into three levels. Level I is slight damage, where the rivet area is not larger than the detection area, that is, K is not larger than 1, and simple operations can be carried out for corrosion repair; Level II is moderate damage, where the corrosion area is larger than the detection area but not larger than twice the rivet area, that is, 1 < K ≤ 2. The structural corrosion damage is relatively serious. After removing the corresponding corrosion products, corresponding reinforcement operations are also required; Level III is severe damage, where the corrosion area is too large, that is, K > 2. The repair method needs to be carefully considered, and replacing the rivets with new ones after disassembly should also be considered.

7. The metal surface corrosion detection method based on the improved Otsu algorithm according to claim 1, characterized in that, The specific process of the corrosion degree analysis in Step S9 is as follows: Introduce the corrosion degree to divide the corrosion grade. For the part without rivets, let the area of the overall corrosion area be A1', and the area of the area with a rivet head diameter of 8.5 mm used as a reference be defined as A2'. Then the corrosion degree can be expressed as: K1 = A1' / A2'; Divide the corrosion condition into three grades. Grade I is mild damage, that is, K1 is not greater than 1, and simple operations can be carried out for corrosion repair; Grade II is moderate damage, and the corrosion area is slightly larger than the detection area, that is, 1 < K1 ≤ 2. The structural corrosion damage is relatively serious. After removing the corresponding corrosion products, corresponding strengthening operations are also required; Grade III is severe damage, and the corrosion area is too large, that is, K1 > 2, and the repair method needs to be considered carefully.

Images