A method for characterization and assessment of marine corrosion under paint based on pulsed eddy current thermography and BEMD noise reduction

Abstract

The present invention discloses a method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method, and belongs to the technical field of marine corrosion detection. The method comprises the following steps: using pulsed eddy current thermal imaging ECPT to detect marine corrosion samples, and obtaining a corrosion image sequence of the surface temperature rise and fall of corrosion samples with different corrosion degrees; using the BEMD method to separate the corrosion image layer by layer, and remove the intrinsic mode function IMF corresponding to the noise, and reconstruct the corrosion image; comparing the corrosion area and the non-corroded area of ​​the corrosion image after noise removal, and evaluating the corrosion degree of the marine corrosion test sample. Through accurate corrosion characterization and evaluation, the present invention can provide a scientific basis for the corrosion protection design of marine structures, help engineers take protective measures in time, extend the service life of marine facilities, and reduce the economic losses caused by corrosion.

Description

[0001] The present invention belongs to the technical field of marine corrosion detection, and in particular relates to a method for characterizing and evaluating marine corrosion under a paint layer based on pulsed eddy current thermal imaging and BEMD noise reduction method. Background Art

[0002] By analyzing the corrosion image sequence obtained by the conductor material under pulsed eddy current excitation, the defects on the material surface can be characterized and evaluated. However, random interference is not fully considered and processed, which affects the original corrosion image and the reconstructed feature image. On the other hand, the change of surface emissivity will lead to the inhomogeneity of temperature value, which makes it impossible to obtain a good corrosion image, thus affecting the detection results.

[0003] Currently, the main denoising method for the eroded image sequence obtained by ECPT is principal component analysis (PCA).

[0004] The steps of principal component analysis are as follows: Figure 1 As shown, it can be divided into:

[0005] ① The corrosion image sequence captured by the infrared thermal imager is taken as the M×N raw data U(t)x×Ny, where m represents the frame number of the recorded corrosion image sequence, and Nx and Ny are the resolutions of the infrared thermal imager.

[0006] ②Through vectorization processing, the original data U(t) is arranged into a column vector matrix U V (t) of (Nx×Ny)×m is used as input data.

[0007] ③ Perform image reconstruction of the main components, such as Figure 2 From left to right are the first principal component, the second principal component, the third principal component and the fourth principal component after PCA processing.

[0008] PCA is a linear dimensionality reduction method that assumes that data is linearly separable. However, in eroded images, there are often nonlinear complex relationships, which makes it difficult for PCA to effectively extract the features of eroded images. Such nonlinear structures may be ignored, thus affecting the accurate expression of eroded image features.

[0009] Secondly, eroded images usually have high-dimensional features, and some information may be lost through PCA dimensionality reduction. Since the pixels of eroded images often correspond to complex object surfaces and their materials, dimensionality reduction may cause the features of different objects to be mixed together, reducing the differentiation and causing information loss.

[0010] In addition, PCA is more sensitive to the statistical characteristics of the data and is more sensitive to outliers. In the eroded image, due to the influence of noise and shooting conditions, there may be many outliers, which may have a greater impact on the results of PCA analysis, resulting in inaccurate dimensionality reduction results.

[0011] Finally, when processing large-scale data, PCA needs to calculate the covariance matrix and its eigenvalue decomposition, which is computationally complex and time-consuming. For high-dimensional large-scale data such as eroded images, computational complexity becomes a problem, limiting the efficiency of PCA in practical applications. Summary of the invention

[0012] The purpose of the present invention is to propose a method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermography and BEMD noise reduction method, the method comprising the following steps: using pulsed eddy current thermography (ECPT) to detect marine corrosion samples, and obtaining a corrosion image sequence of surface heating and cooling of corrosion samples with different corrosion degrees; using the BEMD method to separate the corrosion image layer by layer, and remove the intrinsic mode function IMF corresponding to the noise, and reconstruct the corrosion image; comparing the corroded area and the uncorroded area of ​​the corrosion image after noise removal, and evaluating the corrosion degree of the marine corrosion test sample; the present invention can provide a scientific basis for the corrosion protection design of marine structures through accurate corrosion characterization and evaluation, help engineers take protective measures in time, extend the service life of marine facilities, and reduce the economic losses caused by corrosion.

[0013] The technical solution adopted by the present invention is a method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method, comprising the following steps:

[0014] Step 1: Use pulsed eddy current thermal imaging to detect marine corrosion test samples.

[0015] Step 2: Obtain corrosion images of the surface of the marine corrosion test sample during the heating and cooling process.

[0016] Step 3: Use the two-dimensional empirical mode decomposition (BEMD) method to separate the different frequency and scale features in the corrosion image layer by layer, remove the intrinsic mode function (IMF) corresponding to the noise, and reconstruct the corrosion image;

[0017] Step 4: Compare the corroded area with the non-corroded area of ​​the corrosion image after noise removal to evaluate the corrosion degree of the marine corrosion test sample.

[0018] Among them, the intrinsic mode function IMF satisfies the following judgment conditions:

[0019] Condition (1): In the entire data set, the difference between the number of extreme points and the number of zero points is less than or equal to 1;

[0020] Condition (2): At any point, the average value of the envelope defined by the local maxima and minima is zero.

[0021] Furthermore, the smooth changes of spots and holes in the eroded image are processed by bilinear interpolation method.

[0022] Furthermore, the specific steps of processing the smooth changes of spots and holes in the eroded image by the bilinear interpolation method are as follows:

[0023] Interpolate in the horizontal direction:

[0024] f(x,j)=f(i,j)·(i+1-x)+f(i+1,j)·(xi)

[0025] f(x,j+1)=f(i,j+1)·(i+1-x)+f(i+1,j+1)·(xi)

[0026] Among them, f(x,j) represents the interpolation function value at the position (x,j), f(i,j) represents the value at the grid point (i,j), f(i+1,j) represents the value at the grid point (i+1,j), f(x,j+1) represents the value at the grid point (x,j+1), f(i,j+1) represents the value at the grid point (i,j+1), f(i+1,j+1) represents the value at the grid point (i+1,j+1), x represents the coordinate of the non-grid point in the horizontal axis direction, i represents the horizontal coordinate of the first point in the grid, and j represents the vertical coordinate of the first point in the grid.

[0027] Then, interpolate in the vertical direction to get the value f(x,y) at the non-grid point (x,y):

[0028] f(x,y)=f(x,j)·(j+1-y)+f(x,j+1)·(yj)

[0029] Where f(x,y) represents the interpolation function value at the non-grid point (x,y), and y represents the coordinate of the non-grid point in the vertical direction.

[0030] Thus, the upper envelope can be obtained and lower envelope

[0031]

[0032] in Represents the value at the position (x,y) on the upper envelope surface, represents the value at the grid point (x, j) in the upper envelope surface, represents the value at the grid point (x,j+1) in the upper envelope surface, Represents the value at the position (x,y) of the lower envelope surface, represents the value at the grid point (x, j) in the lower envelope surface, Represents the value at the grid point (x, j+1) in the lower envelope surface. Calculate the upper envelope and lower envelope The average value is used to remove local low-frequency trends or global structural information in the image.

[0033] Preferably, based on the corrosion characteristics in the marine atmosphere environment, the envelope mean is adjusted in combination with the corrosion depth distribution weight, so that the severely corroded areas in the image will get higher weights during the decomposition process, so that these areas are more significantly expressed in the intrinsic mode function (IMF), and by giving higher weights to areas with greater depth, BEMD can better separate these areas, making the corroded areas in the final image clearer, which is helpful for subsequent image analysis and corrosion assessment, and specifically includes the following steps:

[0034] Calculate the corrosion depth value D(x,y) at (x,y) in the corrosion image:

[0035]

[0036] Q(x,y) represents the heat source intensity at the position (x,y) of the erosion image, and α(x,y) represents the thermal diffusion coefficient at the position (x,y) of the erosion image.

[0037] Then the corrosion depth is normalized to eliminate the influence of dimension and the normalized corrosion depth D is obtained. norm (x,y), and then construct the adaptive weight function of the envelope average:

[0038] w(x,y)=1+β·D norm (x,y)

[0039] Among them, w(x,y) represents the adaptive weight function value corresponding to the position (x,y) of the eroded image, and β is the weight adjustment coefficient, which is used to control the increase of the weight.

[0040] Finally, combined with the adaptive weight function, the adjusted average envelope is calculated:

[0041]

[0042] Among them, m 11 (x,y) represents the calculated average envelope value at (x,y) of the eroded image.

[0043] Furthermore, removing the intrinsic mode function IMF corresponding to the noise and reconstructing the eroded image includes the following steps:

[0044] Calculate the intermediate variable h of IMF 11 (x,y) indicates the first intermediate variable value at (x,y) obtained when calculating the first component of the IMF:

[0045] h 11 (x,y)=f(x,y)-m 11 (x,y)

[0046] Where f(x,y) represents the interpolation function value at the eroded image (x,y).

[0047] Repeat the calculation of the adjusted average envelope k times until h 1k (x,y) represents the kth intermediate variable value at (x,y) obtained when calculating the first component of IMF.

[0048] Satisfy the judgment condition of IMF and set it as the first component c1(x,y) of IMF:

[0049] h 1k (x,y)=h 1(k-1) (x,y)-m 1k (x,y)

[0050] c1(x,y)=h 1k (x,y)

[0051] Among them, h 1(k-1) (x, y) represents the k-1th intermediate variable value at (x, y) obtained when calculating the first component of IMF, m 1k (x,y) represents the k-th average envelope value at (x,y) obtained when calculating the first component of the IMF.

[0052] The residual r1(x, y) of the eroded image is obtained by the following formula, which represents the residual value obtained by subtracting the value at the first component (x, y) of the IMF from the value at the eroded image (x, y):

[0053] r1(x,y)=f(x,y)-c1(x,y)

[0054] Use r1(x,y) as the eroded image to remove noise.

[0055] Preferably, the first component of the IMF is Gaussian filtered, and the eroded image is smoothed by weighted averaging pixels in the neighborhood using the Gaussian function.

[0056] Furthermore, the degree of corrosion is evaluated by the grayscale contrast between the corroded area and the uncorroded area.

[0057] The grayscale contrast ratio is expressed by the following formula:

[0058]

[0059] Among them, L M and L N They represent the average grayscale values ​​of the corroded area and the non-corroded area in the corrosion image, L represents the average grayscale value of the corrosion image, M represents the corroded area, N represents the non-corroded area, and δ M and δ N represents the standard deviation of the average gray value of the corroded area and the non-corroded area, respectively, l(b) is the gray value of the b-th pixel, C represents the gray contrast value, and g is the number of pixels in the corresponding corroded area or non-corroded area.

[0060] Alternatively, the extent of corrosion is assessed by differential peak shape deviation.

[0061] The beneficial effects of the characterization and evaluation method of marine corrosion under paint layer based on pulsed eddy current thermal imaging and BEMD noise reduction method of the present invention are:

[0062] 1. The present invention reduces noise interference in the eroded image by BEMD. The eroded image sequence is considered to be stable, but the noise interference changes the number and position of the extreme points of the eroded image sequence, thereby changing the amplitude and frequency of the eroded image sequence. BEMD is a nonlinear decomposition method based on local extreme points, which can perform adaptive decomposition according to the characteristics of the eroded image itself and can effectively reduce noise interference in the eroded image.

[0063] 2. The present invention uses the peak deviation from normal distribution to characterize the temperature change process of each pixel point, which contains richer information and is conducive to improving the accuracy and sensitivity of the characterization in terms of method.

[0064] 3. Since the color value of the corrosion image sequence corresponds to the temperature value, the present invention uses the contrast between the corrosion area and the uncorroded area as an indicator for evaluating the BEMD denoising effect, which can intuitively reflect the denoising effect.

[0065] 4. A single eigenvalue has limitations in distinguishing early corrosion from mid- to late-stage corrosion. The present invention can comprehensively evaluate corrosion samples by combining differential peak deviation with contrast, and can effectively distinguish the degree of corrosion at each stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is the principle of PCA processing eroded images in the prior art;

[0067] Figure 2 It is each principal component graph after PCA processing in the prior art;

[0068] Figure 3 It is a process framework diagram of the present invention;

[0069] Figure 4 is a flow chart of eroded image reconstruction of the present invention;

[0070] Figure 5 This is the corrosion degree assessment algorithm of Example 3 of the present invention. DETAILED DESCRIPTION

[0071] The present invention is specifically aimed at processing corrosion image sequences in eddy current pulse thermal imaging, with the main purpose of noise reduction and corrosion image enhancement, especially to preserve corrosion feature information. This application is very critical for detecting marine corrosion, because such corrosion information is usually very weak and easily submerged by noise.

[0072] The present invention proposes a method for characterizing and evaluating marine corrosion under paint layer based on pulsed eddy current thermal imaging and BEMD noise reduction method, such as Figure 3 As shown, the following steps are included:

[0073] Step 1: Use pulsed eddy current thermal imaging to detect marine corrosion test samples.

[0074] Step 2: Obtain corrosion images of the surface of the marine corrosion test sample during the heating and cooling process.

[0075] Step 3: Use the two-dimensional empirical mode decomposition (BEMD) method to separate the different frequency and scale features in the corrosion image layer by layer, remove the intrinsic mode function (IMF) corresponding to the noise, and reconstruct the corrosion image;

[0076] Step 4: Compare the corroded area with the non-corroded area of ​​the corrosion image after noise removal to evaluate the corrosion degree of the marine corrosion test sample.

[0077] The following is a specific process designed by the present invention.

[0078] First, in the ECPT experiment, the infrared camera uses an infrared detector to receive the thermal radiation generated on the surface of the sample and obtains the corresponding corrosion image through signal processing. However, other radiation sources in the environment are a challenge to the imaging quality. For stable, periodic radiation sources, such as indoor light sources and excitation coils, the method of subtracting background frames is effective. However, when there are irregular, non-periodic interference sources, such as surrounding human bodies, equipment, and instruments, the differential method has limitations.

[0079] Moreover, the influence of the marine environment on the carbon steel corrosion specimens is complex and diverse, including not only uniform corrosion, but also pitting, crevice corrosion, stress corrosion cracking and other corrosion forms, which are complex and unevenly distributed. Therefore, the present invention adopts the two-dimensional empirical mode decomposition (BEMD) method, the essence of which is to identify the inherent oscillation mode through the characteristic time scale in the data, and then decompose the data into several intrinsic mode functions (IMFs). An intrinsic mode function is a function that satisfies two conditions: (1) in the entire data set, the difference between the number of extreme points and the number of zero points is less than or equal to 1; (2) at any point, the average value of the envelope defined by the local maximum and minimum values ​​is zero.

[0080] The steps of this method are as follows:

[0081] (1) Since noise interference often appears as extreme points in corrosion images, finding the extreme points of the two-dimensional corrosion image I(x, y) is a key step, where x and y represent the horizontal and vertical coordinates of any pixel in the corrosion image, and I(x, y) represents the temperature value of the pixel.

[0082] (2) Obtain the upper envelope by bilinear interpolation method and lower envelope The bilinear interpolation method has good smoothness and is suitable for processing smooth changes of spots and holes in eroded images. The calculation steps of the bilinear interpolation are as follows:

[0083] Interpolate in the horizontal direction:

[0084] f(x,j)=f(i,j)·(i+1-x)+f(i+1,j)·(xi)

[0085] f(x,j+1)=f(i,j+1)·(i+1-x)+f(i+1,j+1)·(xi)

[0086] Among them, f(x,j) represents the interpolation function value at the position (x,j), f(i,j) represents the value at the grid point (i,j), f(i+1,j) represents the value at the grid point (i+1,j), f(x,j+1) represents the value at the grid point (x,j+1), f(i,j+1) represents the value at the grid point (i,j+1), f(i+1,j+1) represents the value at the grid point (i+1,j+1), x represents the coordinate of the non-grid point in the horizontal axis direction, i represents the horizontal coordinate of the first point in the grid, and j represents the vertical coordinate of the first point in the grid.

[0087] Then, interpolate in the vertical direction to get the value f(x,y) at the non-grid point (x,y):

[0088] f(x,y)=f(x,j)·(j+1-y)+f(x,j+1)·(yj)

[0089] Where f(x,y) represents the interpolation function value at the non-grid point (x,y), and y represents the coordinate of the non-grid point in the vertical direction.

[0090] Thus, the upper envelope can be obtained and lower envelope

[0091]

[0092] in Represents the value at the position (x,y) on the upper envelope surface, represents the value at the grid point (x, j) in the upper envelope surface, represents the value at the grid point (x,j+1) in the upper envelope surface, Represents the value at the position (x,y) of the lower envelope surface, represents the value at the grid point (x, j) in the lower envelope surface, Represents the value at the grid point (x, j+1) in the lower envelope surface. Calculate the upper envelope and lower envelope The average value is used to remove local low-frequency trends or global structural information in the image.

[0093] (3) Calculate the average value of the upper and lower envelopes. In order to target the corrosion characteristics in the marine atmosphere environment, this paper innovatively proposes a method to adjust the envelope average value calculation by combining the corrosion depth distribution weight. First, it is necessary to obtain the corrosion depth value D(x, y) at each pixel in the corrosion image, and use the one-dimensional heat diffusion model to describe the temperature change process. Assuming that the initial temperature is T0, the temperature change in the corrosion area can be expressed as:

[0094]

[0095] Among them, T(t) is the temperature value at a certain moment, t is time, Q is the heat source intensity, and α is the thermal diffusion coefficient. Among them, high corrosion depth means that the thermal resistance on the surface and inside of the material increases, causing more heat to accumulate in the corrosion area. Therefore, a larger Q reflects deeper corrosion. Low corrosion depth means that the material's thermal diffusion ability is weaker. Corrosion makes the material structure loose, reduces thermal conductivity, and causes α to decrease. Therefore, a lower α also reflects deeper corrosion. Therefore, based on the thermal diffusion model parameters, the corrosion depth is estimated using the following formula, and the corrosion depth is normalized to eliminate the impact of dimension:

[0096]

[0097]

[0098] Among them, Q(x,y) represents the heat source intensity at the position (x,y) of the corrosion image, α(x,y) represents the thermal diffusion coefficient at the position (x,y) of the corrosion image, and D norm (x, y) represents the normalized erosion depth at the position (x, y) of the eroded image, D max and D min are the maximum and minimum values ​​of the corrosion depth, respectively. Next, an adaptive weight function w(x,y) is defined, which represents the adaptive weight function value corresponding to the corrosion image (x,y) position. The weight function is based on the normalized value D of the corrosion depth distribution. norm (x,y). Considering that the greater the corrosion depth, the greater its impact on the envelope average, the weight function can be defined as:

[0099] w(x,y)=1+β·D norm (x,y)

[0100] Among them, β is the weight adjustment coefficient, which is used to control the increase of weight. Combined with the adaptive weight function, the adjusted average envelope is calculated:

[0101]

[0102] (4) Calculate the intermediate variable h by the following formula 11 (x,y) indicates the first intermediate variable value at (x,y) obtained when calculating the first component of the IMF:

[0103] h 11 (x,y)=I(x,y)-m 11 (x,y)

[0104] (5) Repeat steps (2)-(4) k times until h 1k (x, y), indicating that the k-th intermediate variable value at (x, y) obtained when calculating the first component of IMF meets the judgment condition of IMF and is set as the first component of IMF c1(x, y):

[0105] h 1k (x,y)=h 1(k-1) (x,y)-m 1k (x,y)

[0106] c1(x,y)=h 1k (x,y)

[0107] (6) The residual r1(x, y) of the eroded image is obtained by the following formula, which represents the residual value obtained by subtracting the value at the first component (x, y) of the IMF from the value at the eroded image (x, y):

[0108] r1(x,y)=f(x,y)-c1(x,y)

[0109] (7) Using r1(x, y) as the eroded image to remove noise, repeat the above steps until the number of extreme points in one residual component is less than 2. The eroded image I(x, y) can be expressed as the superposition of all IMFS components, as shown in the following formula:

[0110]

[0111] Among them, c a (x,y) represents the ath component of IMF, r h (x, y) represents the hth residual, and h indicates that the BEMD decomposition is completed until the hth residual is obtained.

[0112] BEMD has multi-scale adaptive characteristics and can separate different frequency and scale features in the eroded image layer by layer. The higher-frequency IMF often contains detailed noise information, while the lower-frequency IMF retains the main structural information and erosion characteristics of the eroded image. The first IMF usually contains the highest frequency component, which is often mainly noise. However, simply removing the first IMF may lose some real information about the corrosion. Therefore, based on BEMD, this paper Gaussian filters the first IMF. The Gaussian filter smoothes the eroded image by weighted averaging the pixels in the neighborhood through the Gaussian function. The specific steps are as follows:

[0113] (1) First, define the filter window with a size of q×q, where q represents the length of the filter window.

[0114] (2) Calculate the Gaussian weight by the following formula:

[0115]

[0116] Among them, G(u,v) represents the Gaussian weight value, u represents the horizontal offset within the filter window, v represents the horizontal offset within the filter window, and σ is the standard deviation of the Gaussian function, which determines the smoothness of the filter.

[0117] (3) Apply Gaussian filtering to the first IMF component c1(x,y) and normalize it:

[0118]

[0119] c1(x+u,y+v) represents the value of the first component of IMF (x+u,y+v), c′1(x,y) represents the first component of IMF after Gaussian filtering and normalization, and W is the sum of Gaussian weights, which can be obtained by the following formula:

[0120]

[0121] The first IMF component obtained after Gaussian filtering is reconstructed with the remaining IMF and the residual term to generate a denoised eroded image. In the reconstructed eroded image, the noise is significantly reduced, while the meaningful erosion features are retained and enhanced.

[0122] Example 1

[0123] The present invention can use the degree of deviation of the temperature response curve from the normal distribution to characterize the temperature change characteristics of each pixel in the corrosion image, as shown in the following formula. This characteristic parameter is essentially the third-order central moment of the random signal, which is used to measure the degree of deviation of the random signal from the normal distribution.

[0124]

[0125] Among them, sk represents the degree to which the temperature response curve at the position (x, y) in the corrosion image deviates from the normal distribution, and f s (x, y) represents the temperature sampling value of the sth frame at the eroded image (x, y), s represents the frame number of the current pixel, is the mean of the temperature sampling values ​​of the sth frame at the eroded image (x, y), and p is the total number of frames of the eroded image.

[0126] The present invention uses it to characterize the degree of deviation of the temperature curve from the normal state, because changes in the magnetic permeability, electrical conductivity and thermal conductivity of the material will affect its heating and cooling process under external stimulation, and will be reflected as the deviation of the temperature curve, left or right. According to the above analysis, the degree of deviation of the temperature response curve of the pixel points in the corroded area and the uncorroded area of ​​the material surface from the normal distribution is different, so it can be used to reconstruct the corrosion image. The calculation steps are as follows: Figure 4 As shown, firstly, a corrosion image sequence of the corrosion sample is obtained, and BEMD denoising processing is performed on each frame of the image. Then, the temperature change curve of each pixel point is calculated and normalized to obtain the peak normal distribution of the data. The peak normal distribution of each pixel point is used as the value of the pixel point in the reconstructed image to obtain the reconstructed image.

[0127] Example 2

[0128] Since the color value of the corrosion image sequence corresponds to the temperature value, the present invention uses the contrast between the corroded area and the uncorroded area as an indicator for evaluating the BEMD denoising effect. The contrast measures the degree of light and dark changes or color differences in the corrosion image. The greater the contrast, the greater the brightness or color difference between different areas, and the clearer the detail features. For the corrosion image of the corroded sample, when the contrast value between the corroded area and the uncorroded area is greater, it indicates that the two are more distinguishable and have a more obvious boundary, which is more conducive to the detection and evaluation of corrosion. The contrast index of the corrosion image before and after BEMD denoising is compared to judge the denoising effect of BEMD. The local contrast calculation formula is:

[0129]

[0130] Among them, L M and L N They represent the average grayscale values ​​of the corroded area and the non-corroded area in the corrosion image, L represents the average grayscale value of the corrosion image, M represents the corroded area, N represents the non-corroded area, and δ M and δ N represents the standard deviation of the average grayscale value of the corroded area and the non-corroded area, respectively, l(b) is the grayscale value of the b-th pixel, C represents the grayscale contrast value, and g is the number of pixels in the corresponding corroded area or non-corroded area. The standard deviation can help detect outliers or extreme data points, which is particularly important in corrosion image processing. By considering the standard deviation in the contrast calculation, the interference of outliers on the results can be eliminated during corrosion identification, and the stability and accuracy of contrast evaluation can be improved. And it can measure the degree of dispersion of data points relative to the mean, so combining the mean with the standard deviation in corrosion detection can help determine the data distribution range of the corrosion area, thereby more accurately distinguishing the characteristics of different areas.

[0131] Example 3

[0132] The present invention fuses the differential peak deviation and contrast of the corroded area and the uncorroded area of ​​the denoised reconstructed corrosion image as an indicator for comprehensively evaluating the degree of corrosion, distinguishes corrosion samples with different degrees of corrosion, and improves the accuracy of the evaluation.

[0133] Evaluation process Figure 5As shown, firstly, the temperature change curve of each pixel is calculated and normalized to obtain the data peak normal distribution amount, and the peak normal distribution amount of each pixel is used as the value of the pixel in the reconstructed image to obtain the reconstructed image and the skewness is used to obtain a reconstructed corrosion image, and the average skewness value corresponding to each pixel in the corroded area of ​​the reconstructed image is subtracted from the average skewness value of the uncorroded area to obtain the differential skewness eigenvalue. At the same time, the contrast value of the corroded and uncorroded areas is calculated. Finally, based on the two eigenvalues, multidimensional data visualization analysis is performed, and the differential skewness eigenvalue is used as the horizontal coordinate and the contrast eigenvalue is used as the vertical coordinate. The corrosion degree of each sample is displayed in the form of coordinate points. If the corrosion degree is evaluated only by a single eigenvalue, the corrosion degree will not be accurately evaluated due to the non-monotonic change of the corrosion degree over time. Therefore, after adding the eigenvalue, different degrees of corrosion can be distinguished.

[0134] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction, characterized in that: The steps include: Step 1: Use pulsed eddy current thermal imaging to detect marine corrosion test samples. Step 2: Obtain corrosion images of the surface of the marine corrosion test sample during the heating and cooling process. Step 3: Use the two-dimensional empirical mode decomposition (BEMD) method to separate the different frequency and scale features in the corrosion image layer by layer, remove the intrinsic mode function (IMF) corresponding to the noise, and reconstruct the corrosion image; Step 4: Compare the corroded area with the non-corroded area of ​​the corrosion image after noise removal to evaluate the corrosion degree of the marine corrosion test sample; Based on the corrosion characteristics in the marine atmosphere environment, the envelope average is adjusted in combination with the corrosion depth distribution weight, which specifically includes the following steps: Calculate the corrosion depth value D(x,y) at (x,y) in the corrosion image: Q(x,y) represents the heat source intensity at the position (x,y) of the corrosion image, α(x,y) represents the thermal diffusion coefficient at the position (x,y) of the corrosion image, Then the corrosion depth is normalized to eliminate the influence of dimension and the normalized corrosion depth D is obtained. norm (x,y), and then construct the adaptive weight function of the envelope average: w(x,y)=1+β·D norm (x,y) Among them, w(x,y) represents the adaptive weight function value corresponding to the position (x,y) of the eroded image, and β is the weight adjustment coefficient, which is used to control the increase of the weight. Finally, combined with the adaptive weight function, the adjusted average envelope is calculated: Among them, m 11 (x,y) represents the average envelope value at (x,y) of the eroded image. Represents the value at the position (x,y) on the upper envelope surface, Represents the value at the position (x,y) of the lower envelope surface.

2. The method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method according to claim 1, characterized in that: The intrinsic mode function IMF satisfies the following judgment conditions: Condition (1): In the entire data set, the difference between the number of extreme points and the number of zero points is less than or equal to 1; Condition (2): At any point, the average value of the envelope defined by the local maxima and minima is zero.

3. The method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method according to claim 2, characterized in that: The smooth changes of spots and holes in the eroded image are processed by bilinear interpolation method.

4. The method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method according to claim 3, characterized in that: The specific steps of processing the smooth changes of spots and holes in the eroded image by bilinear interpolation method are as follows: Interpolate in the horizontal direction: f(x,j)=f(i,j)·(i+1-x)+f(i+1,j)·(xi) f(x,j+1)=f(i,j+1)·(i+1-x)+f(i+1,j+1)·(xi) Wherein, f(x,j) represents the interpolation function value at the position (x,j), f(i,j) represents the value at the grid point (i,j), f(i+1,j) represents the value at the grid point (i+1,j), f(x,j+1) represents the value at the grid point (x,j+1), f(i,j+1) represents the value at the grid point (i,j+1), f(i+1,j+1) represents the value at the grid point (i+1,j+1), x represents the coordinate of the non-grid point in the horizontal axis direction, i represents the horizontal coordinate of the first point in the grid, and j represents the vertical coordinate of the first point in the grid; Then, interpolate in the vertical direction to get the value f(x,y) at the non-grid point (x,y): f(x,y)=f(x,j)·(j+1-y)+f(x,j+1)·(yj) Where f(x,y) represents the interpolation function value at the non-grid point (x,y), and y represents the coordinate of the non-grid point in the vertical direction. Thus, the upper envelope can be obtained and lower envelope in Represents the value at the position (x,y) on the upper envelope surface, represents the value at the grid point (x, j) in the upper envelope surface, represents the value at the grid point (x,j+1) in the upper envelope surface, Represents the value at the position (x,y) of the lower envelope surface, represents the value at the grid point (x, j) in the lower envelope surface, Represents the value at the grid point (x, j+1) in the lower envelope surface, and calculates the upper envelope and lower envelope The average value is used to remove local low-frequency trends or global structural information in the image.

5. A method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method according to claim 4, characterized in that: Removing the intrinsic mode function IMF corresponding to the noise and reconstructing the eroded image includes the following steps: Calculate the intermediate variable h of IMF 11 (x,y) indicates the first intermediate variable value at (x,y) obtained when calculating the first component of the IMF: h 11 (x,y)=f(x,y)-m 11 (x,y) Repeat the calculation of the adjusted average envelope k times until h 1k (x,y) represents the kth intermediate variable value at (x,y) obtained when calculating the first component of IMF. Satisfy the judgment condition of IMF and set it as the first component c1(x,y) of IMF: h 1k (x,y)=h 1(k-1) (x,y)-m 1k (x,y) c1(x,y)=h 1k (x,y) Among them, h 1(k-1) (x, y) represents the k-1th intermediate variable value at (x, y) obtained when calculating the first component of IMF, m 1k (x,y) represents the kth average envelope value at (x,y) obtained when calculating the first component of IMF. The residual r1(x, y) of the eroded image is obtained by the following formula, which represents the residual value obtained by subtracting the value at the first component (x, y) of the IMF from the value at the eroded image (x, y): r1(x,y)=f(x,y)-c1(x,y) Use r1(x,y) as the eroded image to remove noise.

6. A method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method according to claim 5, characterized in that: The first component of the IMF is Gaussian filtered, and the pixels in the neighborhood are weighted averaged by the Gaussian function to smooth the eroded image.

7. A method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method according to claim 6, characterized in that: The degree of corrosion is evaluated by the grayscale contrast between the corroded area and the uncorroded area.

8. A method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method according to claim 7, characterized in that: Grayscale contrast uses the following formula: Among them, L M and L N They represent the average grayscale values ​​of the corroded area and the non-corroded area in the corrosion image, L represents the average grayscale value of the corrosion image, M represents the corroded area, N represents the non-corroded area, and δ M and δ N represents the standard deviation of the average gray value of the corroded area and the non-corroded area, respectively, l(b) is the gray value of the b-th pixel, C represents the gray contrast value, and g is the number of pixels in the corresponding corroded area or non-corroded area.

9. A method for characterizing and evaluating marine corrosion under paint based on pulsed eddy current thermal imaging and BEMD noise reduction method according to claim 8, characterized in that: The degree of corrosion was assessed by differential peak shape deviation.

Images