A method and system for identifying aging of a sealing element of a ring main unit

Abstract

The application relates to the technical field of visual detection, in particular to a ring main unit sealing piece aging identification method and system, which comprises the following steps: constructing a gradient deviation degree in a local window of any pixel point in a surface image; calculating a skewness feature value difference of the pixel point in a preset direction, combining the gradient deviation degree to calculate a crack saliency; constructing a local penalty factor negatively correlated with the crack saliency; replacing a global penalty factor in a low-rank sparse decomposition algorithm with the local penalty factor, using the low-rank sparse decomposition algorithm to decompose the surface image to obtain a sparse matrix, and judging an aging state of the sealing piece according to a proportion of a number of non-zero elements in the sparse matrix. The technical scheme can improve the aging identification efficiency of the low-rank sparse decomposition algorithm on the sealing piece.

Description

Technical Field

[0001] This application relates to the field of visual inspection technology. More specifically, this application relates to a method and system for identifying aging of ring main unit seals. Background Technology

[0002] Ring main units (RMNs) are critical nodes in power distribution networks, and the sulfur hexafluoride (SF6) gas filling them plays a vital role in insulation and arc extinguishing. The RMN seals are the core barrier maintaining the airtightness of the gas chamber, their fundamental function being to prevent the internal insulating gas from leaking to the outside, while simultaneously blocking moisture and dust from entering the cabinet. However, these seals are mostly made of polymer materials such as EPDM rubber. During long-term operation, under the combined effects of high and low temperature cycles, strong electric field ozone corrosion, and mechanical stress, the seal surface inevitably undergoes aging phenomena such as hardening, cracking, powdering, and permanent deformation. Once the seals fail due to aging, it will directly lead to a decrease in gas chamber pressure and a sharp drop in insulation capacity, easily inducing internal flashover or even short-circuit explosion accidents. Therefore, timely and accurate identification of the aging status of RMN seals is of great significance for preventing major power grid outages.

[0003] In existing technologies, low-rank sparse decomposition algorithms, due to their solid mathematical foundation and strong interpretability, can be used for aging identification of sealing components. Normal surface textures of sealing components exhibit low rank, while local aging defects are characterized by sparsity, allowing aging features to be separated through matrix decomposition. However, ring main unit sealing components typically exhibit black or dark gray light-absorbing properties and are often irregular curved surfaces. Early aging features manifest as extremely fine microcracks, requiring the acquisition of high-resolution images. However, when directly using traditional low-rank sparse decomposition algorithms to process high-resolution image matrices, solving the kernel norm minimization problem requires performing singular value decomposition of the high-dimensional matrix in each iteration. This causes the computational complexity of the algorithm to increase cubically with the image size, resulting in extremely low computational efficiency and excessively long computation time. This cannot meet the engineering requirements of power inspection robots or online monitoring systems for real-time and rapid diagnosis of sealing component conditions. Summary of the Invention

[0004] This application provides a method and system for identifying the aging of seals in ring main units, aiming to solve the problem of low efficiency of low-rank sparse decomposition algorithms when identifying the aging of seals.

[0005] In a first aspect, this application provides a method for identifying the aging of ring main unit seals. The method includes: constructing a local window for any pixel in a surface image to calculate the gradient deviation of each pixel; calculating the mean of the gradient deviation of the pixel in a preset direction as the skewness feature value of the preset direction; calculating the gradient deviation difference of the pixel based on the difference of the skewness feature values ​​in each preset direction; and using the product of the gradient deviation difference and the gradient deviation as the crack saliency of the pixel; constructing a local penalty factor for the pixel, wherein the local penalty factor is negatively correlated with the crack saliency and negatively correlated with the mean of the crack saliency of each pixel in the surface image; replacing the global penalty factor in the low-rank sparse decomposition algorithm with the local penalty factor, so that different pixels in the surface image correspond to different local penalty factors; using the low-rank sparse decomposition algorithm to decompose the surface image to obtain a sparse matrix; calculating the proportion of non-zero elements in the sparse matrix; and comparing the proportion with a first threshold and a second threshold to determine the aging state of the seal.

[0006] By constructing gradient deviation, the gradient changes when aging features appear are accurately quantified; by analyzing the linear extension and directional features of aging cracks in space, crack saliency is constructed, significantly improving the accuracy of aging feature extraction; a local penalty factor based on crack saliency replaces the global penalty factor in the low-rank sparse decomposition algorithm, improving the low-rank sparse decomposition algorithm. This allows the improved algorithm to intelligently adjust the iteration convergence speed according to the aging complexity of the sealing surface, accelerating convergence in smooth background regions to save computing power, and performing fine calculations in complex aging regions to preserve details, thus achieving efficient and accurate diagnosis of the aging state of the sealing components.

[0007] Furthermore, the method for calculating the gradient deviation includes: for any pixel within a local window, calculating the mean of the gradient magnitudes of all pixels, and taking the mean of the absolute values ​​of the differences between the gradient magnitudes of all pixels and the mean as the gradient deviation of the pixel.

[0008] By constructing a gradient deviation, the gradient change when aging characteristics of the seal appear can be accurately quantified while suppressing random noise from the sensor.

[0009] Furthermore, the size of the local window is either 7×7 or 9×9.

[0010] Furthermore, the preset direction includes: , , and direction.

[0011] Furthermore, the step of calculating the gradient deviation difference of the pixel based on the difference in skewness feature values ​​in each preset direction includes: obtaining the maximum value, minimum value, and mean value of all preset direction skewness feature values, taking the difference between the maximum value and the minimum value as the first difference value, and taking the ratio of the first difference value to the mean value as the gradient deviation difference of the pixel.

[0012] By constructing gradient deviation differences, the directional characteristics of cracks appearing during the aging of seals can be reflected, enabling a more accurate assessment of the aging characteristics of seals.

[0013] Furthermore, the neighboring pixels of the pixel in the preset direction include: for the pixel, the pixels within a preset distance in the preset direction are the neighboring pixels.

[0014] Furthermore, determining the aging state of the seal includes: calculating the ratio of the number of non-zero elements in the sparse matrix to the total number of pixels in the surface image, as the aging ratio; if the aging ratio is greater than a first threshold, the ring main unit seal is determined to be in a severely aged state; if the aging ratio is greater than a second threshold but less than or equal to the first threshold, the ring main unit seal is determined to be in a slightly aged state; if the aging ratio is less than or equal to the second threshold, the ring main unit seal is in a normal state.

[0015] By comparing the aging ratio with the first and second thresholds, the aging state of the seal can be accurately classified.

[0016] Furthermore, the method for calculating the local penalty factor includes: taking the square root of the total number of pixels in the surface image as the numerator; calculating the mean of the crack saliency of all pixels in the surface image, taking the sum of the mean and the crack saliency of the pixels as the denominator; and taking the ratio of the numerator to the denominator as the local penalty factor.

[0017] By constructing a local penalty factor instead of the globally uniform penalty factor in the low-rank sparse decomposition algorithm, a larger local penalty factor can be used in smooth regions during each iteration of the algorithm to accelerate the convergence of the low-rank matrix, while a smaller local penalty factor can be used in aging and complex regions to finely separate sparse aging textures. This avoids high-intensity computation of the globally uniform factor during the iteration process and improves the execution efficiency of the algorithm.

[0018] Furthermore, the recognition method also includes: preprocessing the surface image; the preprocessing includes enhancement using an adaptive histogram equalization algorithm and denoising using median filtering.

[0019] In a second aspect, this application also provides a ring main unit sealing component aging identification system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a ring main unit sealing component aging identification method according to the first aspect of this application is implemented.

[0020] This application has the following technical effects:

[0021] The gradient deviation constructed using a logarithmic nonlinear weighting strategy accurately quantifies gradient changes when aging features appear while suppressing random noise from the sensor. Crack saliency is constructed by analyzing the linear extension and directional characteristics of aging cracks in space, thus directionally enhancing the spatially connected real crack signals, significantly suppressing isotropic isolated noise, and significantly improving the accuracy of aging feature extraction. A local penalty factor based on crack saliency replaces the globally uniform penalty factor in the low-rank sparse decomposition algorithm, improving the algorithm. This allows the improved algorithm to intelligently adjust the iteration convergence speed according to the aging complexity of the sealing surface, accelerating convergence in smooth background regions to save computational power, and performing fine calculations in complex aging regions to preserve details, achieving efficient and accurate diagnosis of the aging state of the sealing component. Attached Figure Description

[0022] Figure 1 This is a flowchart of a method for identifying the aging of a ring main unit seal according to an embodiment of this application.

[0023] Figure 2 This is a diagram showing the aging status identification of the ring main unit seal according to an embodiment of this application.

[0024] Figure 3 This is a structural block diagram of a ring main unit sealing component aging identification system according to an embodiment of this application. Detailed Implementation

[0025] Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] The first aspect of this application provides a method for identifying the aging of ring main unit seals. Figure 1 This is a flowchart of a method for identifying the aging of a ring main unit seal according to an embodiment of this application. The specific implementation process of this method will be described in detail below.

[0027] S101: Obtain a surface image of the ring main unit's sealing components.

[0028] In this embodiment, images are captured using an image acquisition device mounted on a track-mounted inspection robot or a portable handheld terminal. The acquisition device includes a high-resolution industrial area array camera, a macro lens, and a ring-shaped shadowless LED light source. During acquisition, an infrared rangefinder sensor controls the distance between the lens and the sealing surface to maintain it within the optimal depth of field, such as between 15cm and 20cm. The ring light source provides supplementary illumination to the sealing surface from multiple angles to eliminate shadows caused by the curved surface structure and highlight surface texture details. By adjusting the camera exposure parameters, a high-resolution original image covering the entire area of ​​the sealing surface is obtained.

[0029] It should be noted that, in order to improve image quality, the obtained original image needs to be preprocessed, including: converting the color image to grayscale to reduce the computational dimensionality and obtain a grayscale image; using an adaptive histogram equalization algorithm to enhance the local contrast of the grayscale image to improve the problem of insufficient contrast of the black rubber material surface details due to light absorption characteristics; and performing median filtering on the grayscale image to remove the salt-and-pepper noise generated by the sensor while preserving the edge features of the sealing surface, providing a high-quality data foundation for subsequent feature index calculations.

[0030] At this point, a surface image of the pre-processed ring main unit seal is obtained.

[0031] S102: Construct a local window for any pixel in the surface image to calculate the gradient deviation of each pixel.

[0032] In this embodiment, the ring main unit seals are mostly made of black rubber. In their unaged state, their surface is relatively smooth and dense, reflecting light uniformly, and the pixel grayscale values ​​in local areas change gradually. When the seals age, their surface exhibits powdering and micropores, leading to a sharp increase in surface roughness. This change in physical surface is manifested in the image as: the pixel grayscale values ​​in the aged area exhibit high-frequency oscillations within a very small neighborhood, meaning the gradient direction and amplitude between adjacent pixels are chaotic, losing the consistency of a smooth surface.

[0033] Therefore, this step constructs the gradient deviation of any pixel in the surface image to reflect the early pulverization and microcrack trends of the seal surface.

[0034] Specifically, for any pixel in the surface image, a local window of size 7×7 is taken with the pixel as the center, and the Sobel operator is used to obtain the gradient magnitude of each pixel in the local window.

[0035] The gradient deviation of the pixel is constructed based on the above features, and the calculation method is as follows:

[0036]

[0037] In the formula This indicates the gradient deviation of the pixel. This represents the number of pixels within a local window; in this embodiment, the value is 49. Indicates the first [number] within a local window Gradient magnitude of each pixel This represents the average gradient magnitude of all pixels within a local window. This represents a logarithmic function with the natural constant as its base. This represents a preset hyperparameter, used to avoid the denominator being zero, which would prevent calculation. In this embodiment, the value is 1.

[0038] The deeper the aging of the seal surface, the coarser the micro-texture, and the more dramatic and irregular the gradient differences between pixels, leading to... The item becomes larger, and at the same time through The method weights pixels with high gradient amplitudes, removes the influence of high gradient differences and low gradient amplitudes of noise points, effectively identifies gradient amplitude changes caused by aging of the seal, and the greater the corresponding calculated gradient deviation.

[0039] This embodiment also provides a method for calculating gradient deviation, including: for any pixel point within a local window, calculating the mean value of the gradient magnitude of all pixels, and taking the mean value of the absolute values ​​of the differences between the gradient magnitudes of all pixels and the mean value as the gradient deviation of the pixel point.

[0040] S103: Calculate the mean of the gradient deviation of the neighboring pixels of the pixel in the preset direction as the skewness feature value of the preset direction, calculate the gradient deviation difference of the pixel based on the difference of the skewness feature values ​​in each preset direction, and use the product of the gradient deviation difference and the gradient deviation as the crack saliency of the pixel.

[0041] In this embodiment, the aging of ring main unit seals is often accompanied by cracking. These cracks are physically continuous and directional, rather than isolated points. That is, if a region has a high gradient deviation, and the texture dispersion of its surrounding neighborhood is also generally high, and these high gradient deviation regions exhibit a linear or mesh-like connected structure in space, then the probability of severe aging cracks in that region will greatly increase. That is, the gradient deviation remains at a high level along the crack extension direction, while it rapidly decreases in the direction perpendicular to the crack. Conversely, sensor noise or dust often manifests as isolated high gradient deviations.

[0042] Therefore, this step constructs the crack saliency of the pixel points to reflect the connectivity of the local aging characteristics of the ring main unit seal.

[0043] Specifically, for each pixel, within a local window, in a preset direction, the set of all pixels on the line connecting the pixel is denoted as the direction point set. The mean of the gradient deviation of all pixels in the direction point set is calculated as the skewness feature value of the preset direction; the skewness feature values ​​are calculated for different preset directions. The preset direction includes: , , and direction.

[0044] The crack saliency of the pixel is constructed based on the above features, and the calculation method is as follows:

[0045]

[0046] In the formula This indicates the salience of the crack at the pixel. This indicates the gradient deviation of the pixel. , and These represent the maximum, minimum, and mean values ​​of all preset directional skewness characteristic values, respectively. This represents a preset hyperparameter, used to avoid the denominator being zero, which would prevent calculation. In this embodiment, the value is 1.

[0047] When the pixel meets the aging characteristics of the seal, i.e., the gradient deviation remains at a high level along the crack extension direction, but rapidly decreases in the direction perpendicular to the crack, then there is a high gradient deviation in the crack extension direction, resulting in a large deviation characteristic value in that direction and a small deviation characteristic value in the direction perpendicular to the crack. The term is significantly increased, and at the same time, the gradient deviation can ensure that the pixel itself has a strong gradient difference relative to other pixels in the local window, so that when the pixel meets the aging characteristics of the seal, the calculated crack significance is large.

[0048] When the pixel is isolated random noise, the gradient deviation of the pixel within the local window does not have directional characteristics, making... The term is close to 0, which corresponds to a relatively small crack significance in the calculation.

[0049] S104: Construct a local penalty factor for the pixel, wherein the local penalty factor is negatively correlated with the crack significance and negatively correlated with the mean crack significance of each pixel in the surface image.

[0050] In this embodiment, the crack saliency of any pixel in the surface image is obtained through the above steps, reflecting the degree of conformity between the pixel and the aging characteristics of the seal. That is, when the surface image is decomposed using the low-rank sparse decomposition algorithm, the crack saliency of pixels in unaged or slightly aged regions is small, and these regions mainly exhibit low-rank characteristics with fewer sparse components; while the crack saliency of pixels in severely aged regions is large, and these regions contain complex sparse textures. To adjust the penalty intensity of the low-rank sparse decomposition algorithm when decomposing the surface image, the constraint balance between sparsity and low-rank is dynamically adjusted during the iteration process, thereby accelerating convergence and improving the execution efficiency of the algorithm.

[0051] Therefore, for any pixel in the surface image, this step constructs a local penalty factor for the pixel to achieve adaptive adjustment of the penalty constraint strength of each pixel in the low-rank sparse decomposition algorithm, thereby accelerating the low-rank convergence of the background region while ensuring the fine sparse reconstruction of complex aging textures.

[0052] Specifically, the crack significance of all pixels in the surface image is normalized using the minimax normalization method to avoid extreme values ​​and improve the stability of the calculated local penalty factor.

[0053] Based on the above features, the local penalty factor for the pixel is constructed, and the calculation method is as follows:

[0054]

[0055] In the formula This represents the local penalty factor for the pixel. This represents a preset constant, which in this embodiment is taken as the square root of the total number of pixels in the surface image. This indicates the salience of the crack at the pixel. The mean value representing the crack significance of all pixels in the surface image is used as a global adjustment factor. This represents a preset hyperparameter, used to avoid the denominator being zero, which would prevent calculation. In this embodiment, the value is 1.

[0056] If the aging characteristics of the pixel are obvious, that is A larger value corresponds to a smaller calculated local penalty factor, allowing for the preservation of more detailed information when processing the complex texture structure of aging regions during the iterative process of the low-rank sparse decomposition algorithm; if the aging features of the pixel are weak, i.e. The smaller value corresponds to a smaller calculated local penalty factor, which enables the algorithm to quickly converge to the low-rank solution during the iteration process of the low-rank sparse decomposition algorithm. This achieves adaptive adjustment of the aging features of different pixels, significantly reducing unnecessary iteration oscillations while ensuring accuracy.

[0057] Simultaneously through global adjustment factors The crack saliency is adjusted to improve the robustness of the local penalty factor.

[0058] S105: Replace the global penalty factor in the low-rank sparse decomposition algorithm with a local penalty factor so that different pixels in the surface image correspond to different local penalty factors. Use the low-rank sparse decomposition algorithm to decompose the surface image to obtain a sparse matrix. Calculate the proportion of non-zero elements in the sparse matrix. Compare the proportion with the first threshold and the second threshold respectively to determine the aging state of the seal.

[0059] In this embodiment, the local penalty factor of each pixel in the surface image is replaced with the global penalty factor in the low-rank sparse decomposition algorithm. During the iteration process of the algorithm, when updating the low-rank matrix and sparse matrix in each iteration, a larger local penalty factor is used in the smooth region to accelerate the convergence of the low-rank matrix, and a smaller local penalty factor is used in the aging complex region to finely separate the sparse aging texture. This avoids high-intensity calculations on the global uniformity during the iteration process and improves the execution efficiency of the algorithm.

[0060] Following the steps described above, the surface image of the preprocessed ring main unit seal is input into the improved low-rank sparse decomposition algorithm. After iterative convergence, a sparse matrix is ​​output. The sparse matrix retains only the aging texture and cracks on the surface of the seal. The Otsu threshold segmentation method is used to binarize the sparse matrix to obtain the number of non-zero elements in the sparse matrix. The ratio of the number of non-zero elements to the total number of pixels in the surface image is calculated as the aging ratio.

[0061] If the aging ratio is greater than the first threshold, the ring main unit seal is determined to be in a severely aged state; if the aging ratio is greater than the second threshold but less than or equal to the first threshold, the ring main unit seal is determined to be in a slightly aged state; if the aging ratio is less than or equal to the second threshold, the ring main unit seal is in a normal state. In this embodiment, the first and second thresholds are set to 3% and 1%, respectively, and can be selected according to actual conditions.

[0062] Figure 2 This is a result diagram of the aging state identification of the ring main unit seal according to the embodiments of this application. The result of aging identification of the ring main unit seal according to the above steps is as follows. Figure 2 As shown.

[0063] According to a second aspect of this application, this application also provides a ring main unit seal aging identification system. Figure 3 This is a structural block diagram of a ring main unit sealing component aging identification system according to an embodiment of this application. Figure 3As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the aging identification method for ring main unit seals according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface; their configurations and functions are known in the art and will not be described further here.

Claims

1. A method for identifying aging of ring main unit seals, characterized in that, Identification methods include: Construct a local window for any pixel in the surface image to calculate the gradient deviation of each pixel; The mean of the gradient deviation of a pixel in a preset direction is calculated as the skewness feature value of the preset direction. The gradient deviation difference of the pixel is calculated based on the difference of the skewness feature values ​​in each preset direction. The product of the gradient deviation difference and the gradient deviation is used as the crack significance of the pixel. A local penalty factor for each pixel is constructed. The local penalty factor is negatively correlated with the crack significance and negatively correlated with the mean crack significance of each pixel in the surface image. The local penalty factor is replaced by the global penalty factor in the low-rank sparse decomposition algorithm, so that different pixels in the surface image correspond to different local penalty factors. The surface image is decomposed using the low-rank sparse decomposition algorithm to obtain a sparse matrix. The proportion of the number of non-zero elements in the sparse matrix is ​​calculated, and the proportion is compared with the first threshold and the second threshold respectively to determine the aging state of the seal. The gradient deviation is calculated as follows: , This represents the gradient deviation of a pixel. This indicates the number of pixels within a local window. Indicates the first [number]th [unit] within a local window Gradient magnitude of each pixel This represents the average gradient magnitude of all pixels within a local window. This represents a logarithmic function with the natural constant as its base. Indicates the preset hyperparameters; The determination of the aging state of the seals includes: calculating the ratio of the number of non-zero elements in the sparse matrix to the total number of pixels in the surface image, as the aging ratio; if the aging ratio is greater than a first threshold, the ring main unit seal is determined to be in a severely aged state; if the aging ratio is greater than a second threshold but less than or equal to the first threshold, the ring main unit seal is determined to be in a slightly aged state; if the aging ratio is less than or equal to the second threshold, the ring main unit seal is in a normal state; the first threshold and the second threshold are set to 3% and 1%, respectively.

2. The method for identifying aging of ring main unit seals according to claim 1, characterized in that, The size of the local window is either 7×7 or 9×9.

3. The method for identifying aging of ring main unit seals according to claim 1, characterized in that, Preset directions include: , , and direction.

4. The method for identifying aging of ring main unit seals according to claim 1, characterized in that, The gradient deviation difference of a pixel is calculated based on the difference in skewness feature values ​​in each preset direction. This includes: obtaining the maximum, minimum, and average values ​​of all preset skewness feature values; taking the difference between the maximum and minimum values ​​as the first difference; and taking the ratio of the first difference to the average value as the gradient deviation difference of the pixel.

5. The method for identifying aging of ring main unit seals according to claim 1, characterized in that, The neighboring pixels of a pixel in a preset direction include: for a pixel, pixels within a preset distance in the preset direction are considered neighboring pixels.

6. The method for identifying aging of ring main unit seals according to claim 1, characterized in that, The calculation method for the local penalty factor includes: taking the square root of the total number of pixels in the surface image as the numerator; calculating the mean of the crack significance of all pixels in the surface image, and taking the sum of the mean and the crack significance of the pixels as the denominator; and taking the ratio of the numerator to the denominator as the local penalty factor.

7. The method for identifying aging of ring main unit seals according to claim 1, characterized in that, The recognition method also includes: preprocessing the surface image; the preprocessing includes enhancement using an adaptive histogram equalization algorithm and denoising using median filtering.

8. A ring main unit sealing component aging identification system, characterized in that, include: The system includes a processor, a memory, and a communication interface. The memory stores a computer program, which, when executed by the processor, implements a method for identifying aging of ring main unit seals as described in any one of claims 1 to 7.

Images