Array eddy current C-scan image enhancement method based on image pyramid reconstruction

Abstract

The invention discloses an array eddy current C-scan image enhancement method based on image pyramid reconstruction, and the method carries out the combined bilateral filtering of an array eddy current C-scan original image and a C-scan image after Gaussian filtering, and can remove most noisy points while keeping the detail information of small defects. According to the method, an image pyramid algorithm is adopted to carry out Laplacian pyramid decomposition, detail images under different scales are obtained, detail information of the images is further extracted, and noise is effectively removed. According to the method, the detail images under different scales are subjected to linear superposition reconstruction according to the proportion of L2: L1: L0 being 1: 0.2: 0.1, and detail information under multiple scales can be combined, so that the defect detection rate is improved to the maximum extent, the signal-to-noise ratio is further improved, and the array eddy current C-scan image with the high signal-to-noise ratio is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of eddy current detection, and in particular to an array eddy current C-scan image enhancement method based on image pyramid reconstruction. Background Art

[0002] Array eddy current testing is a non-destructive testing method for surface and near-surface defects. It inherits the advantages of eddy current testing such as fast, convenient, and no need for coupling agents. At the same time, through the application of eddy current coil arrays, it can achieve large-area synchronous scanning and real-time imaging of defects. It is widely used in aerospace, special equipment, petrochemical and other testing fields.

[0003] However, directly performing amplitude imaging on the array eddy current signal will be affected by the roughness of the surface of the workpiece to be inspected, the uneven conductivity of the material itself, and the system noise, resulting in the defect signal being submerged, causing a large number of noise points in the original C-scan image, unclear defect edges, and low signal-to-noise ratio. At present, commonly used image enhancement algorithms include histogram equalization and median filtering, but these conventional filtering methods are not effective in processing and optimizing array eddy current C-scan images with multiple noise points and details. This is manifested in that while filtering out noise points, small defects and defect detail information are also filtered out, which easily leads to missed defects.

[0004] Therefore, an array eddy current C-scan image enhancement method based on image pyramid reconstruction is needed to solve the above problems. Summary of the invention

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] An array eddy current C-scan image enhancement method based on image pyramid reconstruction comprises the following steps:

[0007] Step 1: Perform an array eddy current C scan on a workpiece with surface defects through an array eddy current testing experiment, obtain the C scan raw data of the defects on the surface of the tested workpiece, and perform amplitude imaging on the C scan raw data;

[0008] Step 2: Gaussian filtering is performed on the C-scan raw data to be processed, and the Gaussian filtered data is used as a guide map to perform joint bilateral filtering on the C-scan raw data to obtain an image after joint bilateral filtering;

[0009] Step 3: Use the image pyramid algorithm to perform Laplacian pyramid decomposition on the data after joint bilateral filtering, perform Gaussian filtering and downsampling on the image after joint bilateral filtering three times to obtain a Gaussian pyramid, then perform Gaussian filtering and upsampling on the Gaussian pyramid, subtract the upsampling result from each layer of the Gaussian pyramid to obtain a Laplacian pyramid, that is, obtain detailed images of different scales;

[0010] Step 4: up-sampling the detail images at different scales so that the size of the detail images at different scales is consistent with the size of the original C-scan image;

[0011] Step 5: Linearly stack and reconstruct the detail images at different scales, and perform RGB amplitude mapping on the grayscale image, so as to finally obtain a high signal-to-noise ratio array eddy current C-scan image that retains the image details.

[0012] Preferably, the array eddy current detection frequency in step 1 is 200 kHz, the detected workpiece material is 20# steel, and the amplitude imaging is mapped into an RGB image through pseudo-color mapping.

[0013] Preferably, the Gaussian kernel of the Gaussian filter in step 2 is 5×5, the filter diameter is 5, the standard deviation of the color space is 10, and the standard deviation of the coordinate space is 5.

[0014] Preferably, in step 2, the calculation formula for joint bilateral filtering is as follows:

[0015]

[0016] Where I represents the input image, p and q represent the spatial positions of pixels, and I q represents the grayscale of the corresponding position of the guide map, J represents the output, f and g represent the spatial weight distribution function and the pixel local range weight distribution function respectively, both of which are Gaussian functions. The distribution represents the grayscale value of the corresponding coordinate of the guidance graph.

[0017] Preferably, the algorithm function formula of the Gaussian filter in step 2 is:

[0018]

[0019] Among them, σ is the variance of the gray value of each pixel.

[0020] Preferably, the function expression of the Laplace pyramid obtained in step 3 is:

[0021] Li=G i -UP(DOWN(G i ));

[0022] Among them, Li represents the i-th layer in the Laplace pyramid, G i Represents the i-th layer in the Gaussian pyramid, UP represents upsampling, and DOWN represents downsampling + Gaussian filtering.

[0023] Preferably, in step 4, L1 of the obtained detail images of different scales is upsampled by 4 times, and L2 is upsampled by 8 times, so that the image size is consistent with the original C-scan image size.

[0024] Preferably, in step 5, the detail images at different scales are linearly superimposed and reconstructed at a ratio of 1:0.2:0.1 with respect to the edge clarity. The detail images at different scales are linearly superimposed and reconstructed at this ratio, which can not only retain the multi-scale detail information, but also maximize the image clarity, improve the signal-to-noise ratio, and improve the defect detection rate.

[0025] Preferably, the defects in step 1 are surface opening defects, including vertical linear defects, oblique linear defects, and circular defects.

[0026] Due to the adoption of the above technical solution, the present invention has the following technical advances compared with the prior art:

[0027] 1. This method performs joint bilateral filtering on the original array eddy current C-scan image and the C-scan image after Gaussian filtering, which can remove most of the noise points while retaining the detail information of small defects.

[0028] 2. This method uses the image pyramid algorithm to perform Laplacian pyramid decomposition to obtain detailed images at different scales, further extract detailed information of the image, and effectively remove noise.

[0029] 3. This method can combine the detail information at multiple scales to maximize the defect detection rate, further improve the signal-to-noise ratio, and obtain an array eddy current C-scan image with a high signal-to-noise ratio by linearly superimposing and reconstructing the detail images at different scales in the ratio of L2:L1:L0=1:0.2:0.1. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A flow chart of an array eddy current C-scan image enhancement method based on image pyramid reconstruction provided by the present invention;

[0031] Figure 2 is the Laplacian image pyramid structure;

[0032] Figure 3 Schematic diagram of the joint bilateral filtering structure;

[0033] Figure 4 This is a schematic diagram of the image pyramid algorithm;

[0034] Figure 5 It is the original image of array eddy current C scan;

[0035] Figure 6 It is the image after the array eddy current C-scan image is combined with bilateral filtering;

[0036] Figure 7 The array eddy current C scan after the final image enhancement; DETAILED DESCRIPTION

[0037] The following embodiments may help those skilled in the art to more fully understand the present invention, but shall not limit the present invention in any way.

[0038] Example

[0039] An array eddy current C-scan image enhancement method based on image pyramid reconstruction comprises the following steps:

[0040] Step 1: Perform an array eddy current C scan on a workpiece with surface defects through an array eddy current testing experiment, obtain the C scan raw data of the defects on the surface of the tested workpiece, and perform amplitude imaging on the C scan raw data;

[0041] Step 2: Gaussian filtering is performed on the C-scan raw data to be processed, and the Gaussian filtered data is used as a guide map to perform joint bilateral filtering on the C-scan raw data to obtain an image after joint bilateral filtering;

[0042] Step 3: Use the image pyramid algorithm to perform Laplacian pyramid decomposition on the data after joint bilateral filtering. Perform Gaussian filtering and downsampling on the image after joint bilateral filtering three times to obtain a Gaussian pyramid. Then, perform Gaussian filtering and upsampling on the Gaussian pyramid. Subtract the upsampling result from each layer of the Gaussian pyramid to obtain a Laplacian pyramid, that is, obtain detailed images of different scales. The calculation process is as follows: Figure 4 As shown;

[0043] Step 4: up-sampling the detail images at different scales so that the size of the detail images at different scales is consistent with the size of the original C-scan image;

[0044] Among them, after upsampling, the overall size of the image becomes 4 times of the original size; after downsampling, the overall size becomes one quarter of the original size, but image details will be lost after downsampling and upsampling, so the subtraction after sampling is the lost detail information of the image.

[0045] Figure 2 The grayscale image in is the detail image at different scales, and the RGB image is the image after joint bilateral filtering.

[0046] Step 5: Linearly stack and reconstruct the detail images at different scales, and perform RGB amplitude mapping on the grayscale image, and finally obtain a high signal-to-noise ratio array eddy current C-scan image that retains image details, such as Figure 7 shown.

[0047] In one embodiment of the present invention, the frequency of the array eddy current detection in step 1 is 200kHz, the detected workpiece material is 20# steel, and the amplitude imaging is mapped into an RGB image through pseudo color, such as Figure 5 shown.

[0048] In one embodiment of the present invention, the Gaussian kernel of the Gaussian filter in step 2 is 5×5, the filter diameter is 5, the standard deviation of the color space is 10, and the standard deviation of the coordinate space is 5. The algorithm model of step 2 is as follows Figure 3 As shown in the figure, the image obtained after joint bilateral filtering is as follows Figure 6 shown.

[0049] In one embodiment of the present invention, in step 2, the calculation formula of the joint bilateral filtering is as follows:

[0050]

[0051] Where I represents the input image, p and q represent the spatial positions of pixels, and I q represents the grayscale of the corresponding position of the guide map, J represents the output, f and g represent the spatial weight distribution function and the pixel local range weight distribution function respectively, both of which are Gaussian functions. The distribution represents the grayscale value of the corresponding coordinate of the guidance graph.

[0052] In one embodiment of the present invention, the algorithm function formula of the Gaussian filter in step 2 is:

[0053]

[0054] Among them, σ is the variance of the gray value of each pixel.

[0055] In one embodiment of the present invention, the function expression of the Laplacian pyramid obtained in step 3 is:

[0056] Li=G i -UP(DOWN(G i ));

[0057] Among them, Li represents the i-th layer in the Laplace pyramid, G i Represents the i-th layer in the Gaussian pyramid, UP represents upsampling, and DOWN represents downsampling + Gaussian filtering.

[0058] In one embodiment of the present invention, in step 4, L1 of the obtained detail images of different scales is upsampled by 4 times, and L2 is upsampled by 8 times, so that the image size is consistent with the original C-scan image size.

[0059] In one embodiment of the present invention, in step 5, the detail images at different scales are linearly superimposed and reconstructed at a ratio of 1:0.2:0.1 with respect to the edge clarity. The detail images at different scales are linearly superimposed and reconstructed at this ratio, which can not only ensure the retention of multi-scale detail information, but also maximize the image clarity, improve the signal-to-noise ratio, and improve the defect detection rate.

[0060] In one embodiment of the present invention, the defect in step 1 is a surface opening defect, including a vertical linear defect, an oblique linear defect, and a circular defect.

[0061] The above generally describes the present invention in detail, but it is obvious to a person skilled in the art that some modifications or improvements can be made to the present invention. Therefore, modifications or improvements that do not depart from the spirit of the present invention are within the scope of protection of the present invention.

Claims

1. An array eddy current C-scan image enhancement method based on image pyramid reconstruction, characterized in that: The following steps are involved: Step 1: Perform an array eddy current C scan on a workpiece with surface defects through an array eddy current testing experiment, obtain the C scan raw data of the defects on the surface of the tested workpiece, and perform amplitude imaging on the C scan raw data; Step 2: Gaussian filtering is performed on the C-scan raw data to be processed, and the Gaussian filtered data is used as a guide map to perform joint bilateral filtering on the C-scan raw data to obtain an image after joint bilateral filtering; Step 3: Use the image pyramid algorithm to perform Laplacian pyramid decomposition on the data after joint bilateral filtering, perform Gaussian filtering and downsampling on the image after joint bilateral filtering three times to obtain a Gaussian pyramid, then perform Gaussian filtering and upsampling on the Gaussian pyramid, subtract the upsampling result from each layer of the Gaussian pyramid to obtain a Laplacian pyramid, that is, obtain detailed images of different scales; Step 4: up-sampling the detail images at different scales so that the size of the detail images at different scales is consistent with the size of the original C-scan image; Step 5: Linearly stack and reconstruct the detail images at different scales, and perform RGB amplitude mapping on the grayscale image, so as to finally obtain a high signal-to-noise ratio array eddy current C-scan image that retains the image details.

2. The array eddy current C-scan image enhancement method based on image pyramid reconstruction according to claim 1, characterized in that: In step 2, the calculation formula for joint bilateral filtering is as follows: Where I represents the input image, p and q represent the spatial positions of pixels, and I q represents the grayscale of the corresponding position of the guide map, J represents the output, f and g represent the spatial weight distribution function and the pixel local range weight distribution function respectively, both of which are Gaussian functions. The distribution represents the grayscale value of the corresponding coordinate of the guidance graph.

3. The array eddy current C-scan image enhancement method based on image pyramid reconstruction according to claim 1, characterized in that: The algorithm function formula of the Gaussian filter in step 2 is: Among them, σ is the variance of the gray value of each pixel.

4. The array eddy current C-scan image enhancement method based on image pyramid reconstruction according to claim 1, characterized in that: The function expression of the Laplace pyramid obtained in step 3 is: Li=G i -UP(DOWN(G i )); Among them, Li represents the i-th layer in the Laplace pyramid, G i Represents the i-th layer in the Gaussian pyramid, UP represents upsampling, and DOWN represents downsampling + Gaussian filtering.

5. The array eddy current C-scan image enhancement method based on image pyramid reconstruction according to claim 1, characterized in that: In step 5, the detail images at different scales are linearly superimposed and reconstructed in a ratio of 1:0.2:0.1 with respect to the edge clarity.

6. The array eddy current C-scan image enhancement method based on image pyramid reconstruction according to claim 1, characterized in that: The defects in step 1 are surface opening defects, including vertical linear defects, oblique linear defects, and circular defects.

7. The array eddy current C-scan image enhancement method based on image pyramid reconstruction according to claim 4, characterized in that: In step 4, L1 of the obtained detail images of different scales is upsampled by 4 times, and L2 is upsampled by 8 times, so that the image size is consistent with the original C-scan image size.

8. The array eddy current C-scan image enhancement method based on image pyramid reconstruction according to claim 1, characterized in that: The frequency of the array eddy current detection in step 1 is 200kHz, the material of the detected workpiece is 20# steel, and the amplitude imaging is mapped into an RGB image through pseudo-color mapping.

9. The array eddy current C-scan image enhancement method based on image pyramid reconstruction according to claim 1, characterized in that: The Gaussian kernel of the Gaussian filter in step 2 is 5×5, the filter diameter is 5, the standard deviation of the color space is 10, and the standard deviation of the coordinate space is 5.

Images