Computer vision-based transformer core detection method

Abstract

The present application relates to the technical field of industrial component detection, in particular to a transformer core detection method based on computer vision, which obtains core images, extracts key pixel features, screens directional scattering abnormal areas to generate interference atlas, extracts consistent gradient areas to correct images to generate reconstructed images, integrates gray scale differences to locate symmetric disturbances to generate structure atlas, superimposes structure atlas to analyze overlapping relationship to generate abnormal component atlas, matches reference indexes to evaluate risk levels to generate early warning layers. Through linkage analysis of pixel direction vector and brightness change frequency, the present application can distinguish interference reflection and structure features, correct distortion areas in images based on gray scale statistical stable values, construct symmetric point position atlas and analyze gradient difference value sequence change trend, and combine component mapping contour atlas to quantitatively judge structure overlapping relationship, so that the correlation between abnormal areas and key components can be clearly expressed, and the grading evaluation ability for multiple fault risks in the core can be improved.

Description

Technical Field

[0001] This invention relates to the field of industrial component inspection technology, and in particular to a transformer core inspection method based on computer vision. Background Technology

[0002] Industrial component testing technology encompasses non-destructive testing methods based on principles such as electromagnetics, ultrasound, eddy current, infrared, optics, and lasers. It also includes technical systems that utilize sensor arrays, signal acquisition and processing systems, and intelligent algorithm models to continuously monitor and analyze the state parameters of components during operation. Component testing is widely used in high-voltage equipment such as transformers, circuit breakers, cables, instrument transformers, and insulators to improve the safety, maintainability, and accuracy of lifespan prediction, thereby preventing sudden failures and unplanned downtime.

[0003] As a critical component of power equipment, the transformer core can be effectively inspected using industrial component testing technology to avoid structural and technical defects and ensure its safe operation. Specifically, transformer core inspection refers to the technical methods for identifying and assessing the structural condition, electromagnetic performance, and defects of the transformer's internal core components. Its main uses include detecting problems such as loosening, misalignment, localized overheating, abnormal clamping, and abnormal electromagnetic vibration within the core, preventing equipment risks such as increased eddy current losses, increased noise, and insulation breakdown caused by core failures. This inspection method enables accurate diagnosis of the core's operating status, assisting maintenance personnel in developing maintenance plans and improving the stability and safety of transformer operation.

[0004] Traditional detection methods rely solely on electromagnetic performance or structural state signals for core evaluation. However, they cannot reliably identify abnormal areas in transformer cores in environments with optical interference, structural obstruction, or unstable vibration frequencies. Especially when clamping devices are loose or there is slight displacement of core laminations, the lack of image analysis and symmetry modeling makes it difficult to accurately capture local disturbance signals, leading to an increased rate of false positives. For example, under conditions of light source interference, reflective areas may be mistakenly identified as defective locations, or the influence range of key components may be missed in areas of structural overlap, resulting in distorted risk assessment results. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a transformer core detection method based on computer vision.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is as follows:

[0007] A computer vision-based method for transformer core inspection includes the following steps:

[0008] S1: Obtain the image of the transformer core, call the brightness gradient direction vector of each pixel and the brightness gradient change amplitude in the surrounding neighborhood, combine the gray-scale fluctuation frequency in the time series, filter the pixel areas with directional scattering anomalies, and generate the core reflection interference area map.

[0009] S2: Based on the iron core reflection interference region map, extract the pixel grayscale set of regions with consistent gradients in three adjacent directions, replace the original abnormal region pixel grayscale value with the stable grayscale value, correct the image distortion caused by reflection or interference, and generate the iron core edge structure reconstruction image.

[0010] S3: Call the edge pixel information of the longitudinal or transverse extension area of ​​the iron core in the image of the reconstructed image of the iron core edge structure, extract the corresponding continuous change point sequence, and filter it in combination with the gradient mean square error threshold to locate the boundary of the abnormal region that cannot converge to the symmetric trend, and generate the iron core symmetry perturbation structure map.

[0011] S4: Based on the core symmetry disturbance structure diagram, calculate the percentage of overlap area between the abnormal area and the key component and the intersection length of the structural boundary, identify the abnormal area covering the key part, and generate a distribution diagram of the core abnormality affecting the component.

[0012] In a further technical solution of the present invention, the iron core reflection interference region map includes the boundary of the local brightness change region, the high-frequency pixel distribution with consistent direction, the brightness fluctuation sequence between image frames, and the suspicious reflection area identification mask; the iron core edge structure reconstruction image includes the gray-level correction block of the interference region, the gradient direction matching region, the stable gray-level region of the non-interference frame, and the image complete edge reconstruction region; the iron core symmetry disturbance structure map includes the symmetry mapping path annotation layer, the gray-level difference map of symmetry points, the structural offset response mask, and the continuous jump contour of the abnormal region; the iron core abnormality affected component distribution map includes the boundary projection map of key components, the overlapping label of abnormal components, the spatial index of the influence range, and the overlap rate distribution layer.

[0013] In a further technical solution of the present invention, the step of obtaining the spectrum of the iron core reflection interference region specifically includes:

[0014] S111: Acquire images of the transformer core, extract pixel sets of the contact boundary of the clamping component, the lamination area of ​​the core, and the edge of the magnetic flux channel, monitor image frame data of pixels in the target area, calculate the inter-frame difference of the brightness value sequence of each pixel, and perform frequency statistics on the number of grayscale fluctuations in the time series to obtain the pixel brightness time change feature information.

[0015] S112: Based on the pixel brightness time change feature information, call the brightness gradient direction vector value of the pixel in the current frame, and construct the set of angle vectors between the gradient direction of the pixel and the surrounding neighboring pixels. Compare whether the angle difference in the angle set is within the brightness direction concentration determination threshold, filter out pixels that meet the brightness gradient direction consistency condition, and obtain pixel direction consistency area information.

[0016] S113: Call the frequency values ​​in the pixel orientation consistency region information and pixel brightness time change feature information, and jointly judge the time fluctuation frequency and orientation concentration of the pixels in each region. When the fluctuation frequency is higher than the high-frequency reflection recognition threshold and the orientation consistency meets the condition, the region is marked as a reflection abnormal region, and a core reflection interference region map is generated.

[0017] In a further technical solution of the present invention, the process of determining whether the angle difference in the set of compared angles is within the threshold for determining the concentration of brightness direction specifically involves:

[0018] Call the brightness gradient direction vector value and the brightness gradient direction vector value corresponding to the eight neighboring pixels, calculate the direction angle value between each group of center pixels and neighboring pixels, and generate an angle difference set including eight angle differences.

[0019] Calculate the maximum difference value and standard deviation of the angle difference set according to the degree of dispersion of the included angle values ​​in each direction in the angle difference set;

[0020] When either the maximum difference value or the standard deviation value of the angle difference set is lower than a preset brightness direction consistency judgment threshold, the center pixel is determined to meet the brightness gradient direction consistency condition; the set value of the brightness direction consistency judgment threshold is less than 15 degrees. The specific steps for obtaining the reconstructed image of the iron core edge structure are as follows:

[0021] S211: Based on the abnormal pixel regions marked in the iron core reflection interference region map, extract the pixels in the adjacent regions outside the boundary, call the gradient direction vector value of the pixel in the current frame image, and calculate the angle difference between the gradient direction and the center pixel. Filter out three directional regions whose angle difference does not exceed the brightness direction consistency judgment threshold, extract the pixel gray values ​​respectively, and generate a direction consistency gray set.

[0022] S212: Based on the pixel gray value sequence extracted from the directional consistent gray set, retrieve the gray value evolution trajectory of the same spatial location in consecutive image frames, remove frames with instantaneous brightness surges, filter pixel gray value sequences within the stable frame range, perform an arithmetic average on the sequences, obtain the central stable gray value, calculate the difference between the maximum and minimum values ​​of the sequence, construct a gray value stability fluctuation factor, and generate a stable gray value replacement set;

[0023] S213: Call the center gray value in the stable gray value replacement value set to replace the gray value information of each abnormal region in the iron core reflection interference region map, optimize the continuity of the region boundary and the consistency of the texture, stitch and integrate the replacement region on the basis of the original image, output the reconstructed image frame sequence, and generate the reconstructed image of the iron core edge structure.

[0024] In a further technical solution of the present invention, the step of obtaining the core symmetry disturbance structure diagram specifically includes:

[0025] S311: Call the edge pixels of the region where the iron core extends longitudinally and laterally in the image of the reconstructed image of the iron core edge structure, collect the two-dimensional coordinate index and corresponding gray-level gradient value of the edge pixel points, and establish equidistant point index pairs based on the central axis of the image to construct a symmetrical structure point mapping map and obtain the symmetrical structure index distribution information.

[0026] S312: Based on the equidistant point pairs in the symmetrical structure index distribution information, extract the gray gradient value sequence, perform difference operation and integral normalization according to the index order, calculate and obtain the path symmetry offset of each symmetrical scanning path, filter the path symmetry offset greater than the symmetry offset threshold, mark it as the structural imbalance area, and extract the jump points on the path to obtain the symmetry gradient offset point set.

[0027] S313: Call the coordinate index of the marked points in the symmetric gradient offset point set, and construct the jump point cluster block based on the horizontal and vertical adjacency features. Filter out segments with a span smaller than the region continuity judgment threshold, extract the continuously distributed edge point sequence that is stable in the offset direction with the central axis, and establish the core symmetry perturbation structure diagram.

[0028] In a further technical solution of the present invention, the step of obtaining the distribution diagram of the components affected by the core anomaly specifically includes:

[0029] S411: Based on the marked abnormal region position coordinates in the core symmetry disturbance structure diagram, extract the boundary index range of each abnormal region in the two-dimensional coordinate system, and retrieve the structural contour boundary map of the clamping unit arrangement area, magnetic channel, and insulation overlap boundary of the corresponding image frame in the core structure database. Overlay the boundary map and the abnormal region boundary map to generate an abnormal component overlap index set.

[0030] S412: Call the pixel coordinates recorded in the abnormal component overlap index set, calculate the boundary intersection length between the abnormal region and the component region respectively, and count the proportion of the pixel set corresponding to the intersection region in the component region. Based on the cumulative area ratio, calculate the percentage of pixel area covered by each type of component and generate abnormal component coverage index information.

[0031] S413: Based on the coverage ratio and boundary intersection length of the components in the abnormal component coverage index information, filter the component areas with a coverage ratio greater than the influence judgment threshold, record the associated component names and location coordinate indexes of the corresponding abnormal areas, identify them as key parts with correlation influence areas, and generate a distribution map of components affected by iron core anomalies.

[0032] In a further embodiment of the present invention, the method further includes the following steps:

[0033] S5: Call the anomaly type and the type of covered component in the distribution map of the components affected by the core anomaly, and combine the corresponding brightness reflectance ratio, symmetry offset amplitude and key structure overlap ratio. The brightness reflectance ratio is the ratio of the average brightness value of the abnormal area to the average stable gray value of the surrounding normal area. The interval is matched with the risk assessment reference indicators commonly used in transformer core detection applications. The label is assigned according to the risk level interval of the combined parameters to obtain the multi-type anomaly warning level layer of the core.

[0034] The aforementioned multi-type anomaly early warning level layer for iron cores specifically refers to the risk level color partition map, the anomaly type level label group, the component corresponding risk label set, and the early warning level interval distribution map.

[0035] In a further technical solution of the present invention, the specific steps for obtaining the multi-type anomaly warning level layer of the iron core are as follows:

[0036] S511: Call the identified abnormal areas in the distribution map of the components affected by the iron core abnormality, extract the abnormal type and the type of the covered components in each area, and associate the corresponding feature indicators such as brightness reflectance ratio, symmetry offset amplitude, and structural overlap ratio of the area. The brightness reflectance ratio is the ratio of the average brightness value of the abnormal area to the average stable gray value of the surrounding normal area. Construct a multi-dimensional attribute vector set and generate a feature parameter set of abnormal components.

[0037] S512: Based on the set of characteristic parameters of the abnormal components, calculate and obtain the risk measurement value of each abnormal region based on brightness, structure and region proportion, perform interval matching with the risk level interval boundary value set, mark the corresponding risk label type, and generate an abnormal component risk level label set.

[0038] S513: Call the risk level identifier and corresponding abnormal area coordinate information of each risk level label in the abnormal component risk level label set, bind the risk level label to the abnormal area and draw the risk layer, and overlay it on the iron core structure reference map to establish a multi-type abnormal early warning level layer for the iron core.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] In this invention, by analyzing the linkage between pixel direction vectors and brightness change frequency, interference reflections and structural features can be distinguished, improving the ability to identify local disturbance areas. Based on the grayscale statistical stability value, the distortion areas in the image are corrected, effectively restoring the actual boundary contour of the iron core. By constructing a symmetrical point map and analyzing the gradient difference sequence change trend, the symmetrical disturbance boundary is accurately captured. Then, combined with the component mapping contour map, the structural overlap relationship is quantitatively judged, so that the correlation between abnormal areas and key components can be clearly expressed. By matching multi-dimensional indicators such as brightness reflection, symmetry offset and structural overlap with risk intervals, the abnormal warning level is accurately divided, improving the ability to classify and assess the risk of multiple types of faults inside the iron core. Attached Figure Description

[0041] Figure 1 This is a flowchart of the method of the present invention;

[0042] Figure 2 This is a flowchart illustrating the process of obtaining the spectrum of iron core reflection interference regions according to the present invention;

[0043] Figure 3 This is a flowchart illustrating the process of obtaining a reconstructed image of the iron core edge structure according to the present invention;

[0044] Figure 4 This is a flowchart illustrating the process of obtaining the core symmetry disturbance structure diagram according to the present invention;

[0045] Figure 5 This is a flowchart illustrating the distribution of components affected by core anomalies according to the present invention.

[0046] Figure 6 This is a flowchart illustrating the process of obtaining multiple anomaly warning level layers for iron cores according to the present invention. Detailed Implementation

[0047] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] like Figure 1 As shown, the computer vision-based transformer core detection method of the present invention includes the following steps:

[0049] S1: Obtain the image of the transformer core, extract the pixel set of the contact boundary of the clamping component, the core lamination area, and the edge of the magnetic flux channel, call the brightness gradient direction vector of each pixel and the brightness gradient change amplitude in the surrounding neighborhood, combine the gray-scale fluctuation frequency in the time series, determine whether the direction concentration and high-frequency brightness fluctuation characteristics are satisfied at the same time, filter the pixel area with directional scattering anomaly, and generate the core reflection interference area map.

[0050] S2: Based on the pixel regions marked in the iron core reflection interference region map, extract the pixel grayscale set of regions with consistent gradients in three adjacent directions, average the grayscale intensity values ​​of the regions in the non-interference frame, calculate the upper and lower limits of grayscale intensity fluctuation, replace the original abnormal region pixel grayscale values ​​with stable grayscale values, correct the image distortion caused by reflection or interference, and generate the iron core edge structure reconstruction image.

[0051] S3: Call the edge pixel information of the longitudinal or transverse extension area of ​​the iron core in the image to reconstruct the iron core edge structure, construct a symmetrical point map, integrate and accumulate the gray-level gradient difference between equidistant point pairs, extract the corresponding continuous change point sequence, and filter it in combination with the gradient mean square error threshold to locate the boundary of abnormal regions that cannot converge to the symmetrical trend, and generate the iron core symmetry perturbation structure map.

[0052] S4: Based on the coordinates of the abnormal area in the core symmetry disturbance structure diagram, superimpose the structural mapping contour diagram of the matching core internal clamping unit arrangement area, magnetic channel, and insulation overlap boundary, calculate the percentage of overlap area between the abnormal area and the key component and the intersection length of the structural boundary, identify the abnormal area covering the key part, and generate a core abnormality affected component distribution diagram.

[0053] S5: Call the anomaly type and the type of the covered component in the distribution map of the components affected by the core anomaly. Combine the corresponding brightness reflectance ratio, symmetry offset amplitude and key structure overlap ratio. The brightness reflectance ratio is the ratio of the average brightness value of the abnormal area to the average stable gray value of the surrounding normal area. Match the interval with the risk assessment reference indicators commonly used in transformer core detection applications. Assign labels according to the risk level interval of the combined parameters to obtain the multi-type anomaly warning level layer of the core.

[0054] The iron core reflection interference area map includes the boundary of the local brightness change region, the high-frequency pixel distribution with directional consistency, the brightness fluctuation sequence between image frames, and the mask for identifying suspicious reflection areas. The iron core edge structure reconstruction image includes the gray-level correction block of the interference area, the gradient direction matching area, the stable gray-level area of ​​the non-interference frame, and the image complete edge reconstruction area. The iron core symmetry disturbance structure map includes the symmetry mapping path annotation layer, the gray-level difference map of symmetry points, the structural offset response mask, and the continuous jump contour of the abnormal area. The iron core anomaly affected component distribution map includes the boundary projection map of key components, the overlapping labels of abnormal components, the spatial index of the impact range, and the overlap rate distribution layer. The iron core multi-type anomaly warning level layer specifically refers to the risk level color partition map, the anomaly type level annotation group, the risk label set corresponding to the component, and the warning level interval distribution map.

[0055] like Figure 2 As shown, the specific steps for obtaining the spectrum of the iron core reflection interference region are as follows:

[0056] S111: Acquire images of the transformer core, extract pixel sets of the contact boundary of the clamping component, the lamination area of ​​the core, and the edge of the magnetic flux channel, monitor image frame data of pixels in the target area, calculate the inter-frame difference of the brightness value sequence of each pixel, and perform frequency statistics on the number of grayscale fluctuations in the time series to obtain the pixel brightness time change feature information.

[0057] The obtained transformer core image is a grayscale image with a resolution of 1920×1080 pixels. First, based on the preset structural mask information, the pixel set of the contact boundary of the clamping component is extracted. This set consists of pixels located in the image coordinate ranges (210, 300) to (210, 780) and (1710, 300) to (1710, 780). At the same time, the core lamination area is extracted, with pixel coordinates ranging from (215, 50) to (1705, 1030). Finally, the core lamination area is extracted... The magnetic flux channel edge is located within a rectangular region with pixel coordinates between (880, 50) and (1040, 1030). Within these three target regions, for each pixel, its luminance value data is continuously monitored across a 100-frame image sequence. Taking the pixel at coordinates (450, 600) as an example, its luminance value sequence over 10 consecutive frames is recorded as {128, 127, 129, 185, 188, 132, 130, 129, 192, 135}. This luminance value sequence is then processed... Frame-by-frame interpolation is performed, resulting in a sequence of nine interpolation values: {-1, 2, 56, 3, -56, -2, -1, 63, -57}. Next, the absolute value of each value in this interpolation sequence is statistically analyzed, and a grayscale fluctuation baseline is established. This baseline is based on the brightness value sequence of 50 normal, non-reflective pixels within 1000 frames under stable lighting conditions. The standard deviation of all sequences is calculated, and three times the average value is taken as the baseline. In this example, the brightness standard deviation of 50 pixels... The average standard deviation is 1.5, so the fluctuation baseline is set to 4.5. The absolute values ​​of the aforementioned difference sequence {1, 2, 56, 3, 56, 2, 1, 63, 57} are compared with 4.5, and the number of times the value exceeds the baseline is counted. In this example, there are 4 values ​​(56, 56, 63, 57) greater than 4.5. Therefore, the grayscale fluctuation frequency of this pixel is recorded as 4. This calculation is repeated for all pixels in the target area to finally obtain the brightness time change feature information of all monitored pixels.

[0058] S112: Based on the pixel brightness time change feature information, call the brightness gradient direction vector value of the pixel in the current frame, and construct the set of angle vectors between the gradient direction of the pixel and the surrounding neighboring pixels. Compare whether the angle difference in the angle set is within the brightness direction concentration determination threshold, filter out the pixels that meet the brightness gradient direction consistency condition, and obtain the pixel direction consistency area information.

[0059] The process of comparing whether the angle differences in the set of included angles are within the threshold for determining the concentration of light direction is as follows:

[0060] Call the brightness gradient direction vector value and the brightness gradient direction vector value corresponding to the eight neighboring pixels, calculate the direction angle value between each group of center pixels and neighboring pixels, and generate an angle difference set including eight angle differences.

[0061] Calculate the maximum difference value and standard deviation of the angle difference set based on the degree of dispersion of the included angle values ​​in each direction in the angle difference set;

[0062] When either the maximum difference value or the standard deviation value of the angle difference set is lower than the preset brightness direction consistency judgment threshold, the center pixel is determined to meet the brightness gradient direction consistency condition; the set value of the brightness direction consistency judgment threshold is less than 15 degrees.

[0063] Based on the acquired pixel brightness temporal variation characteristics, pixels with fluctuation frequencies greater than 0 are selected as the objects to be analyzed. Any qualified center pixel, such as the pixel with coordinates (450, 600), is used. The brightness gradient direction vector value in the current image frame is calculated. The Sobel gradients Gx and Gy of this pixel in the x and y directions are calculated, where Gx is 48 and Gy is -64. The gradient direction angle is calculated as atan2(-64, 48), approximately -53.1 degrees. Then, a set of angle vectors between the gradient directions of this center pixel and its eight neighboring pixels is constructed. Specifically, the angle vectors between the gradient directions of these eight neighboring pixels (coordinates...) are calculated. The brightness gradient direction vector values ​​are (449, 599) to (451, 601), excluding the center point. For example, the Gx value of its right neighboring pixel (451, 600) is 52, the Gy value is -60, and its gradient direction angle is approximately -49.1 degrees. Therefore, the angle between this neighboring pixel and the center pixel is |-53.1-(-49.1)|, which is 4.0 degrees. This process is repeated for all eight neighboring pixels to generate an angle difference set containing eight angle difference values, for example, {5.2, 4.5, 4.0, 6.1, 5.5, 4.8, 6.5, 5.8}, in degrees. Then, based on the discreteness of each direction angle value in this angle difference set... The maximum difference and standard deviation of the set are calculated. The maximum difference is the maximum value minus the minimum value in the set, i.e., 6.5 - 4.0 = 2.5 degrees. The standard deviation is obtained by the standard deviation calculation formula and is approximately 0.85 degrees. The calculated maximum difference of 2.5 degrees and standard deviation of 0.85 degrees are then compared with a preset brightness direction consistency judgment threshold, which is set to 12 degrees. The setting process is as follows: 1000 image samples of the iron core lamination area of ​​a transformer under normal operating conditions without reflective interference are collected. The same angle difference calculation is performed on the pixels and their neighborhoods in each sample, forming a set containing a large number of maximum differences and standard deviations. The database of differences was used to plot their cumulative distribution function curves. The value corresponding to the 90% cumulative probability was selected as the threshold to ensure that the small directional changes introduced by the normal vibration of most devices or sensor noise could be filtered out. Experimental data showed that the maximum difference value of 90% of the samples was less than 13.5 degrees and the standard deviation was less than 12.8 degrees. To improve the strictness of the screening, the smaller integer value of 12 degrees was taken as the final threshold. The calculated maximum difference value of 2.5 degrees and the standard deviation of 0.85 degrees were both less than 12 degrees. Therefore, it was determined that the central pixel (450, 600) met the brightness gradient direction consistency condition. After traversing all high-frequency fluctuating pixels, the pixel direction consistency region information was obtained.

[0064] S113: Call the frequency values ​​in the pixel orientation consistency region information and the pixel brightness time change feature information, and jointly judge the time fluctuation frequency and orientation concentration of the pixels in each region. When the fluctuation frequency is higher than the high-frequency reflection recognition threshold and the orientation consistency meets the condition, the region is marked as an abnormal reflection region, and a core reflection interference region map is generated.

[0065] The algorithm retrieves the frequency values ​​from the obtained pixel orientation consistency region information and the brightness time variation feature information of each pixel. For each pixel that passes the orientation consistency judgment, such as pixel (450, 600), its fluctuation frequency value is 4. This value is compared with a high-frequency reflection recognition threshold. This threshold is set by analyzing 200 image samples containing obvious reflection interference and 200 normal samples without reflection interference. Frequency statistics are performed on the pixels in these two types of samples. It is found that the pixel fluctuation frequency values ​​of normal samples are distributed in the range of 0 to 2, while the pixel fluctuation frequency values ​​of reflective samples are... The fluctuation frequency values ​​are concentrated in the range of 3 to 15. In order to effectively distinguish between these two types of cases, the identification threshold is set to 3. That is, when the fluctuation frequency is higher than 3, it is identified as high-frequency fluctuation. In this example, the fluctuation frequency of pixel (450, 600) is 4, which is higher than the high-frequency reflection identification threshold of 3. Moreover, this pixel has been determined to meet the directional consistency condition. Therefore, this pixel is marked as a reflection anomaly. All pixels that meet this joint judgment condition are aggregated to form several regions. These regions are marked as reflection anomaly regions. Finally, all the marked regions jointly generate the iron core reflection interference region map.

[0066] like Figure 3 As shown, the specific steps for obtaining the reconstructed image of the iron core edge structure are as follows:

[0067] S211: Based on the abnormal pixel regions marked in the iron core reflection interference region map, extract the pixels in the adjacent regions outside the boundary, call the gradient direction vector value of the pixel in the current frame image, and calculate the angle difference between the gradient direction and the center pixel. Filter the three directional regions whose angle difference does not exceed the brightness direction consistency judgment threshold, extract the pixel gray value respectively, and generate a direction consistency gray set.

[0068] Based on an anomalous pixel region marked in the iron core reflection interference area map, the region ranges from image coordinates (450, 600) to (460, 610). Pixels within the neighboring region outside its boundary are extracted. Specifically, for each pixel on the boundary of the anomalous region, a range extending outwards from its normal direction by 5 pixels is selected as the neighboring region. For example, for pixel (455, 600) in the anomalous region, its neighboring pixels are (455, 595) to (455, 599). The gradient direction vector values ​​of these neighboring pixels in the current frame image are retrieved, and the distance between each neighboring pixel and the center pixel (455, 600) is calculated. 00) The angle difference between gradient directions, where the gradient direction of the center pixel is -52.5 degrees and the gradient direction of its neighboring pixel (455, 598) is -54.2 degrees, with an angle difference of 1.7 degrees. This angle difference is compared with the brightness direction consistency judgment threshold, and neighboring pixels with an angle difference of no more than 12 degrees are selected. Based on the orientation of these pixels, the three directional regions with the most pixels are selected. For example, after screening, it is found that the top, upper left, and left directions have the most qualified neighboring pixels. Then, the gray values ​​of all qualified pixels in these three directional regions are extracted respectively. Finally, these gray values ​​are integrated to generate a direction-consistent gray set.

[0069] S212: Based on the pixel gray value sequence extracted from the directional consistent gray set, retrieve the gray value evolution trajectory of the same spatial location in consecutive image frames, remove frames with instantaneous brightness surges, filter pixel gray value sequences within the stable frame range, perform arithmetic mean on the sequences, obtain the central stable gray value, calculate the difference between the maximum and minimum values ​​of the sequence, construct a gray value stability fluctuation factor, and generate a stable gray value replacement set;

[0070] Based on a directionally consistent grayscale set containing the grayscale values ​​of 75 pixels from three directional regions, the grayscale evolution trajectory of each of these 75 pixels in 20 consecutive frames is retrieved. Taking a pixel located at (455, 598) in the set as an example, its grayscale value sequence for 20 consecutive frames is {135, 136, 134, 135, 210, 137, 136, ..., 138}. To remove frames with sudden increases in brightness, the inter-frame difference of this sequence is calculated, resulting in the difference sequence {1, -2, 1, 75, -73, ...}. The standard deviation of the entire sequence is also calculated, which is 16.5. The sudden increase criterion is set as the absolute value of the inter-frame difference exceeding three times the standard deviation, i.e., 3 * 16.5 = 49.5. The value in the sequence is 75. The 5th and 6th frames, corresponding to the two differences of -73, are identified as having instantaneous brightness spikes and are therefore removed. Within the remaining 18 stable frames, the grayscale sequence of the pixel is selected, for example, {135, 136, 134, 135, 137, 136, ..., 138}. The arithmetic mean of this stable sequence is calculated, and the center stable grayscale value is 136.2, which is rounded to 136. At the same time, the difference between the maximum value 138 and the minimum value 134 in the stable sequence is calculated, and the difference is 4. This difference is the grayscale stability fluctuation factor. The above process is repeated for each pixel in the direction-consistent grayscale set, and finally a stable grayscale replacement value set containing 75 center stable grayscale values ​​and 75 fluctuation factors is generated.

[0071] S213: Call the center gray value in the stable gray value replacement set to replace the gray value information of each abnormal region in the iron core reflection interference region map, optimize the continuity of the region boundary and the consistency of the texture, stitch and integrate the replacement region on the basis of the original image, output the reconstructed image frame sequence, and generate the reconstructed image of the iron core edge structure.

[0072] The system calls upon the center stable grayscale value from the stable grayscale replacement value set. For example, for the 75 center stable grayscale values ​​calculated with pixel (455, 600) as the center, these 75 values ​​are weighted and averaged. The weights are allocated based on the calculated grayscale stability fluctuation factor. The smaller the fluctuation factor, the more stable the historical grayscale value of that pixel, and the higher its weight. The weight calculation formula is W_i=(1 / F_i) / sum(1 / F_j), where F_i is the fluctuation factor of the i-th pixel. If the final weighted average grayscale value is set to 137, then this value is used to replace the iron core reflection interference. The grayscale information of the abnormal pixel (455, 600) in the region map is used to replace each pixel in the abnormal region with the weighted average grayscale value of its surrounding stable pixels. After replacing the grayscale information of all pixels in the abnormal region, the boundary of the replacement region is subjected to 8-neighbor mean filtering to smooth the grayscale transition of the boundary pixels and optimize the continuity of the region boundary and texture consistency. Finally, all regions that have undergone grayscale replacement and boundary optimization are stitched together and integrated back into the original image frame to output a reconstructed image frame. This operation is performed frame by frame in the entire image sequence to finally generate the reconstructed image sequence of the iron core edge structure.

[0073] like Figure 4 As shown, the specific steps for obtaining the core symmetry disturbance structure diagram are as follows:

[0074] S311: Call the iron core edge structure to reconstruct the edge pixels of the region extending longitudinally and laterally in the image, collect the two-dimensional coordinate index and corresponding gray-level gradient value of the edge pixel points, and establish equidistant point index pairs based on the image central axis to construct a symmetrical structure point mapping map and obtain the symmetrical structure index distribution information.

[0075] The image is reconstructed using the iron core edge structure. Edge pixels extending longitudinally and laterally from this image are extracted using the Canny edge detection operator. A Gaussian filter kernel size of 5x5, a low threshold of 50, and a high threshold of 150 are set to obtain accurate single-pixel width edges. The two-dimensional coordinate indices and corresponding grayscale gradient values ​​of all edge pixels are collected. The grayscale gradient values ​​are calculated using the Sobel operator in the x and y directions, and their modulus is taken. Subsequently, equidistant point index pairs are established based on the geometric center of the image (for a 1920x1080 image, the central axis is x=960). Taking the 400th row of the image (y=400) as an example, if an edge point coordinate is detected on the left edge of this row... If the coordinates are (220, 400), then the coordinates of its symmetrical point on the central axis x=960 should be (960+(960-220), 400), i.e. (1700, 400). The program will search for whether there is an edge point within a specified tolerance range (e.g., ±3 pixels) around (1700, 400). If it exists, for example, an edge point is detected at (1702, 400), then the two pixels (220, 400) and (1702, 400) will be established as a pair of equidistant point index pairs. This operation is performed on all horizontal and vertical edges to construct a complete symmetrical structure point mapping map. This map records the coordinate information of all successfully matched symmetrical point pairs, and finally obtains the symmetrical structure index distribution information.

[0076] S312: Based on the equidistant point pairs in the symmetric structure index distribution information, extract the gray-level gradient value sequence, perform difference calculations and integral normalization according to the index order, and adopt an improved formula:

[0077] ;

[0078] The path symmetry offset of each symmetric scan path is obtained through calculation. Paths with path symmetry offset greater than the symmetry offset threshold are filtered out and marked as structural imbalance areas. Jump points on the path are extracted to obtain the set of symmetric gradient offset points.

[0079] in, Indicates the first The first path The grayscale gradient values ​​of each pixel at the symmetrical points on the left are obtained by calculating the Sobel gradient values ​​of the equidistant symmetrical points on the left side of the image frame. Indicates the first The first path The grayscale gradient value of each pixel at the symmetrical point on the right. Indicates the first The number of pixel pairs whose absolute value of the gray-level gradient difference in a path exceeds the threshold for determining a jump is defined as follows: the threshold is 1.5 times the average pixel gradient within the local window. This represents the number of pixel pairs on each symmetrical path. This represents the path symmetry offset. The symmetry offset threshold is set by extracting the corresponding value from multiple normal region sample paths. The value is used to form a sample distribution curve, and the 95% position of its cumulative distribution function is taken as the threshold to exclude normal fluctuations and screen abnormal structural offset paths.

[0080] Based on the acquired symmetrical structure index distribution information, a horizontal symmetrical scanning path is selected, such as all matched equidistant point pairs on the y=400 row in the image. The gray-level gradient value sequences of these point pairs are extracted to form the left gradient sequence. and right-side gradient sequence ,in Represents the path index. The index of the pixel pair on this path is set to the total number of pixels on this path. For pixels, in index order Perform difference operations and integral normalization from 1 to 1480 using an improved formula:

[0081] ;

[0082] Perform the calculation. The parameters in the formula. Indicates the first The first path The grayscale gradient value of a pixel at a point symmetrical to the left is obtained by calculating the Sobel gradient value of the equidistant symmetrical point on the left side of the image frame. Similarly, This represents the grayscale gradient value of the corresponding point on the right. It is the first The number of pixel pairs whose absolute value of grayscale gradient difference in a path exceeds the threshold for determining a transition is defined as 1.5 times the average gradient of all pixels within a 5x5 local window on the path. The total number of pixel pairs on each symmetrical path. The final calculated value is the path symmetry offset. The calculation logic of this formula lies in the numerator. By accumulating the absolute differences in gradient values ​​between symmetric points, the overall asymmetry of the entire path is quantified. (The denominator is...) As a normalization term, the square root of the sum of squares of all gradient values ​​is used to characterize the overall gradient energy along the path, thereby eliminating the influence of differences in edge sharpness in different regions on the result. The final multiplication term... It is a penalty term applied when there are "jump points" on the path where the gradient changes drastically. As the logarithmic term increases, it will non-linearly amplify the final offset. In a specific example, for the path , The setting is obtained through calculation. ,at the same time , ,for The calculation shows that the average gradient within the 5x5 window on this path is 25, therefore the jump determination criterion is... All paths that satisfy The number of pixel pairs, assuming a total of 10 pairs are found, then Substitute these values ​​into the formula:

[0083] ;

[0084] Calculate the path symmetry offset Then, it is compared with a symmetry offset threshold, which is set by randomly selecting 5000 symmetrical scan paths from 100 core image samples confirmed to be free of structural defects, and calculating their respective... Values, forming a sample distribution curve, and these Sort the values ​​in ascending order and take the first value after sorting. Location The value was used as a threshold, and the experimental statistics showed that this value was 2.5. In this example, This value is greater than the threshold of 2.5, therefore the path... The area was marked as structurally unbalanced, and the coordinates of 10 abrupt transition points (i.e., pixel pairs with an absolute gradient difference exceeding 37.5) calculated along this path were extracted to obtain a set of symmetric gradient offset points. This result indicates that path 400 exhibits significant structural asymmetry. The advantage of the formula lies in the fact that, by introducing a gradient energy term for normalization, the symmetry assessment is unaffected by the strength of the edges themselves, while also incorporating a logarithmic penalty term. The design makes the measurement results more sensitive to local, drastic structural changes, and can effectively identify symmetrical damage caused by local deformation or foreign object obstruction.

[0085] S313: Call the coordinate index of the marked points in the symmetric gradient offset point set, and construct the jump point cluster block based on the horizontal and vertical adjacency features. Filter the segments with a span smaller than the region continuity judgment threshold, extract the continuously distributed edge point sequence that is stable with the offset direction of the central axis, and establish the core symmetry perturbation structure diagram.

[0086] The system retrieves a set of symmetric gradient offset points, containing the coordinate indices of all jump points marked on structural imbalance paths, such as the 10 jump points on path 400, and jump points on other imbalance paths. Based on the lateral and vertical adjacency features of these marked points, a jump point cluster is constructed. Specifically, each jump point is traversed, and its 8-neighborhood is checked for other jump points. If any exist, these two points are grouped into one category. Neighborhood checks are continuously performed on newly added points until no new points are added to a category, thus forming a cluster. During this process, segments with a span smaller than the region continuity threshold (set to 5 pixels) need to be filtered out. The determination is based on the minimum process size of the iron core lamination. Any isolated disturbance smaller than this size is not considered a structural defect. For example, a cluster consisting of 4 adjacent jump points is filtered out because its size is less than 5 pixels. The remaining clusters are all composed of 5 or more consecutive jump points. Subsequently, the edge point sequences that are continuously distributed and stable in the offset direction from the central axis are extracted from these clusters. The offset direction stability judgment means that within a cluster, more than 90% of the points have the same offset direction (i.e., the difference between the left gradient and the right gradient) (e.g., the left gradient is greater than the right gradient). Finally, these edge point sequences that meet the conditions are integrated to establish the iron core symmetry disturbance structure diagram.

[0087] like Figure 5 As shown, the specific steps for obtaining the component distribution diagram affected by core anomalies are as follows:

[0088] S411: Based on the marked abnormal region location coordinates in the core symmetry disturbance structure diagram, extract the boundary index range of each abnormal region in the two-dimensional coordinate system, and retrieve the structural contour boundary map of the clamping unit arrangement area, magnetic channel, and insulation overlap boundary of the corresponding image frame in the core structure database. Overlay the boundary map and the abnormal region boundary map to generate an abnormal component overlap index set.

[0089] Based on the coordinates of the marked abnormal regions in the core symmetry disturbance structure diagram, one abnormal region is extracted. Its boundary index range in the two-dimensional coordinate system is from 218 to 245 in the x-direction and from 398 to 422 in the y-direction. Then, the structural outline boundary diagrams of the clamping unit arrangement area, magnetic channel, and insulation overlap boundary corresponding to the currently processed image frame are retrieved from the core structure database. This database pre-stores the precise location information of each component marked in the transformer design drawings. For example, the database displays the outline boundary of a clamping unit. The boundary in the image ranges from 200 to 250 in the x-direction and from 390 to 440 in the y-direction. The boundary map of the abnormal region (x: 218-245, y: 398-422) is superimposed with the boundary map of the clamping unit (x: 200-250, y: 390-440) to find the pixel region that overlaps with the other region in the coordinate system. The range of this overlapping region is from 218 to 245 in the x-direction and from 398 to 422 in the y-direction. All pixel coordinates of this overlapping region are recorded to generate an abnormal component overlap index set.

[0090] S412: Call the pixel coordinates recorded in the abnormal component overlap index set, calculate the boundary intersection length between the abnormal region and the component region respectively, and count the proportion of the pixel set corresponding to the intersection region in the component region. Based on the cumulative area ratio, calculate the percentage of pixel area covered by each type of component and generate abnormal component coverage index information.

[0091] The system calls the abnormal component overlap index set, which records the pixel coordinates of the overlap between the abnormal region and the clamping unit region. It calculates the intersection length between these two regions by counting the number of consecutive pixels on the boundary of the overlapping region (e.g., an intersection length of 108 pixels). It also counts the proportion of the pixel set corresponding to the intersection region (x: 218-245, y: 398-422) within the component region (x: 200-250, y: 390-440). First, it calculates the pixel area of ​​the intersection region, which is (245-218+...). 1)*(422-398+1)=28*25=700 pixels. Then calculate the total pixel area of ​​the clamping unit component, which is (250-200+1)*(440-390+1)=51*51=2601 pixels. Based on the cumulative area ratio, calculate the percentage of pixel area of ​​the clamping unit covered by the abnormal area, which is (700 / 2601)*100%≈26.9%. Record the two indicators of intersection length 108 pixels and coverage 26.9% to generate the abnormal component coverage index information of the abnormal area.

[0092] S413: Based on the coverage ratio and boundary intersection length of components in the abnormal component coverage index information, filter component areas with a coverage ratio greater than the impact judgment threshold, record the associated component names and location coordinate indexes of the corresponding abnormal areas, identify them as key parts with correlation, and generate a distribution map of components affected by iron core anomalies.

[0093] Based on the abnormal component coverage index information, which recorded the coverage ratio of the abnormal area to the clamping unit as 26.9% and the boundary intersection length as 108 pixels, component areas with a coverage ratio greater than the impact judgment threshold were filtered out. This impact judgment threshold was set according to the importance of different components. For critical load-bearing and fixing components such as clamping units, the threshold was set to 15%. This setting refers to the structural integrity requirements in the relevant equipment maintenance manual, that is, any unknown structural coverage exceeding 15% must be subject to risk assessment. Since the calculated coverage ratio of 26.9% is greater than the set impact judgment threshold of 15%, the clamping unit area was filtered out. The associated component name "top left corner clamping unit" and the location coordinate index of its associated abnormal area (x: 218-245, y: 398-422) were recorded, and this area was marked as a critical part of the related influence area. Finally, all the marked areas were integrated to generate a distribution map of the components affected by the iron core anomaly.

[0094] like Figure 6 As shown, the specific steps for obtaining the multi-type anomaly warning level layer of the iron core are as follows:

[0095] S511: Call the identified abnormal areas in the distribution map of components affected by iron core anomalies, extract the anomaly type and the covered component type in each area, and associate the corresponding feature indicators such as brightness reflectance ratio, symmetry offset amplitude, and structural overlap ratio of the area. The brightness reflectance ratio is the ratio of the average brightness value of the abnormal area to the average stable gray value of the surrounding normal area. Construct a multi-dimensional attribute vector set and generate a feature parameter set of abnormal components.

[0096] The system retrieves the identified abnormal areas from the core anomaly impact distribution map, such as the "upper left corner clamping unit" impact area. It extracts the anomaly type from these areas, classifying it as "symmetrical structural imbalance caused by surface reflection," and identifies the covered component type as "clamping unit." It then associates these with corresponding characteristic indicators, including: brightness-reflectance ratio (calculated by comparing the average brightness of the abnormal area (e.g., 195) with the average stable grayscale value of the surrounding normal area (e.g., 135), with a ratio of 195 / 135 ≈ 1.44); and symmetry offset amplitude (calculated path symmetry offset). The maximum value is 5.362; the structural overlap ratio is the calculated coverage of 26.9%. The heterogeneous index data are integrated to construct a multi-dimensional attribute vector in the form of {abnormal region ID: 001, abnormal type: symmetry imbalance, component type: clamping unit, brightness reflectance: 1.44, symmetry offset: 5.362, overlap ratio: 0.269}. This operation is performed on all identified abnormal regions to finally generate a set of abnormal component feature parameters.

[0097] S512: Based on the set of characteristic parameters of abnormal components, and taking into account brightness, structure, and area proportion, the following formula is used:

[0098] ;

[0099] The risk metric value of each abnormal region is obtained through calculation, and it is matched with the risk level interval boundary value set to mark the corresponding risk label type and generate an abnormal component risk level label set.

[0100] in, Indicates the first Risk measurement value for each abnormal region The normalized value representing the brightness-reflectance ratio is obtained by linearly normalizing the ratio of the gray level at the center of the abnormal region to the average gray level of the corresponding background region. This represents the normalized value of the symmetry offset magnitude, obtained by using the path structure offset from the preceding steps. After normalization to the maximum value, This represents the normalized value indicating the overlap ratio of critical structures. It is obtained as the ratio of the area of ​​overlap between the abnormal region and the critical component structure to the total area of ​​the component. (Parameter) , These are adjustment factors for brightness and structural disturbance, used to adjust their contribution to the overall risk. The risk level interval boundary value set is obtained by extracting risk metrics from a large number of classified iron core anomaly samples. Establish its cumulative probability distribution curve, and set risk interval boundary points based on engineering experience, such as low risk, medium risk, and high risk, to form a segmented level standard value set;

[0101] Based on the set of characteristic parameters of abnormal components, for the abnormal region with ID 001, the formula is used based on its brightness, structure, and region proportion:

[0102] ;

[0103] Perform risk metric calculation. (The formula contains...) Indicates the first Risk measurement value for each abnormal region This is the normalized value of the luminance-reflectance ratio. It is obtained by linearly normalizing the luminance-reflectance ratio of 1.44. The normalization interval is set based on statistics of luminance ratios in historical data, typically ranging from 1.0 (no reflection) to 3.0 (strong specular reflection). This interval is then linearly mapped to [0, 1]. , It is a normalized value of the symmetry offset magnitude, which is obtained by converting the path structure offset. Perform maximum value normalization, setting the maximum value calculated across all paths in the current entire image. If the value is 8.0, then , This is the normalized value of the critical structure overlap ratio, i.e., coverage 0.269, parameter and These are adjustment factors for brightness and structural disturbance, respectively, used to adjust their contribution to the overall risk. Based on expert experience and historical failure data analysis, the hazards of structural deformation are generally higher than those of surface reflection; therefore, they are set as follows: , This makes the structural disturbance have a larger weight. The calculation logic of this formula is that the core term... It is a weighted sum of brightness risk and structural risk, representing the severity of the anomaly itself. Squaring it aims to amplify the risk metric for high-severity events, making it more distinguishable from low-severity events. The denominator... The risk value is then adjusted based on the overlap ratio between the abnormal area and the critical component. The larger the value, the larger the denominator, and the higher the final risk value. The risk will decrease accordingly. This design reflects a risk adjustment mechanism, that is, the risk of an anomaly with a wide impact but low severity may be lower than that of an anomaly with a small impact but extremely high severity.

[0104] Substitute the above parameter values ​​into the formula to calculate:

[0105] ;

[0106] Calculate the risk metric After setting the value to 0.254, it is matched against the risk level interval boundary value set. This boundary value set is obtained by extracting the risk metric values ​​from 500 iron core anomaly samples in the database that have been manually marked by experts as "low risk," "medium risk," and "high risk." A cumulative probability distribution curve was plotted. Based on engineering experience, the 30% and 80% quantiles were selected as the boundary points of the risk interval. The final determined boundary value set was: low risk ( ), medium risk High risk ), due to the calculated The value is between 0.2 and 0.6, therefore the abnormal area is labeled as "medium risk," ultimately generating a risk level label set for abnormal components. This result indicates that the abnormal area with ID 001 has a medium level of risk.

[0107] Table 1 presents a summary of the sample data used to define risk level ranges:

[0108] Table 1 Summary of Sample Data for Risk Level Range Setting

[0109]

[0110] As shown in Table 1, statistical analysis of a large number of classified samples provides data support for the classification of risk levels, ensuring the objectivity and rationality of the boundary value setting.

[0111] S513: Call the risk level label of each risk level and the corresponding abnormal area coordinate information in the abnormal component risk level label set, bind the risk level label to the abnormal area and draw the risk layer, and overlay it on the iron core structure reference map to establish a multi-type abnormal early warning level layer for the iron core.

[0112] The system calls upon the risk level label set for abnormal components, which contains the risk level identifier (e.g., "medium risk") and corresponding coordinate information (e.g., x: 218-245, y: 398-422) for each abnormal area. The risk level labels are then linked to the abnormal areas. A semi-transparent yellow layer is overlaid on a rectangular area between coordinates (218, 398) and (245, 422) to represent "medium risk" (where green represents low risk and red represents high risk). This process generates a risk layer. Finally, all similarly generated risk layers are layered and overlaid on a pre-prepared core structure baseline image, which is a clear, interference-free standard image of the core. Ultimately, a multi-type abnormality early warning level layer for the core is established and output.

[0113] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any simple modifications and substitutions made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A transformer core detection method based on computer vision, characterized in that, Includes the following steps: S1: Obtain the image of the transformer core, call the brightness gradient direction vector of each pixel and the brightness gradient change amplitude in the surrounding neighborhood, combine the gray-scale fluctuation frequency in the time series, filter the pixel areas with directional scattering anomalies, and generate the core reflection interference area map. S2: Based on the iron core reflection interference region map, extract the pixel grayscale set of the three adjacent regions with consistent gradients to obtain the central stable grayscale value. Use the central stable grayscale value to replace the original abnormal region pixel grayscale value to correct the image distortion caused by reflection or interference and generate the iron core edge structure reconstruction image. S3: Call the edge pixel information of the longitudinal and transverse extension areas of the iron core in the image of the reconstructed image of the iron core edge structure, extract the corresponding continuous change point sequence, and filter it in combination with the gradient mean square error threshold to locate the boundary of the abnormal region that cannot converge to the symmetric trend, and generate the iron core symmetry perturbation structure map. S4: Based on the core symmetry disturbance structure diagram, calculate the percentage of overlap area between the abnormal area and the key component and the intersection length of the structural boundary, identify the abnormal area covering the key part, and generate a distribution diagram of the core abnormality affecting the component.

2. The transformer core detection method based on computer vision according to claim 1, characterized in that, The iron core reflection interference region map includes the boundary of the local brightness change region, the high-frequency pixel distribution with consistent direction, the brightness fluctuation sequence between image frames, and the mask for identifying suspicious reflection areas. The iron core edge structure reconstruction image includes the gray-level correction block of the interference region, the gradient direction matching region, the stable gray-level region of the non-interference frame, and the image complete edge reconstruction region. The iron core symmetry disturbance structure map includes the symmetry mapping path annotation layer, the gray-level difference map of symmetry points, the structural offset response mask, and the continuous jump contour of the abnormal region. The iron core abnormality affected component distribution map includes the boundary projection map of key components, the overlapping label of abnormal components, the spatial index of the influence range, and the overlap rate distribution layer.

3. The transformer core detection method based on computer vision according to claim 2, characterized in that, The specific steps for obtaining the core reflection interference region map are as follows: S111: Acquire images of the transformer core, extract pixel sets of the contact boundary of the clamping component, the lamination area of ​​the core, and the edge of the magnetic flux channel, monitor image frame data of pixels in the target area, calculate the inter-frame difference of the brightness value sequence of each pixel, and perform frequency statistics on the number of grayscale fluctuations in the time series to obtain the pixel brightness time change feature information. S112: Based on the pixel brightness time change feature information, call the brightness gradient direction vector value of the pixel in the current frame, and construct the set of angle vectors between the gradient direction of the pixel and the surrounding neighboring pixels. Compare whether the angle difference of the set of angle vectors is within the brightness direction concentration determination threshold, filter out pixels with angle difference less than the brightness direction concentration determination threshold, and obtain pixel direction consistency area information. S113: Call the frequency values ​​in the pixel orientation consistency region information and pixel brightness time change feature information, and jointly judge the time fluctuation frequency and orientation concentration of the pixels in each region. When the fluctuation frequency is higher than the high-frequency reflection recognition threshold and the orientation consistency meets the condition, the region is marked as a reflection abnormal region, and a core reflection interference region map is generated.

4. The transformer core detection method based on computer vision according to claim 3, characterized in that, The process of comparing whether the angle difference in the set of angle vectors is within the threshold for determining the concentration of brightness direction is as follows: Call the brightness gradient direction vector value and the brightness gradient direction vector value corresponding to the eight neighboring pixels, calculate the direction angle value between each group of center pixels and neighboring pixels, and generate an angle difference set including eight angle differences. Calculate the maximum difference value and standard deviation of the angle difference set according to the degree of dispersion of the included angle values ​​in each direction in the angle difference set; When either the maximum difference value or the standard deviation value of the angle difference set is lower than the preset brightness direction concentration determination threshold, the center pixel is determined to meet the brightness gradient direction consistency condition; the set value of the brightness direction concentration determination threshold is less than 15 degrees.

5. The transformer core detection method based on computer vision according to claim 4, characterized in that, The specific steps for obtaining the reconstructed image of the iron core edge structure are as follows: S211: Based on the abnormal pixel regions marked in the iron core reflection interference region map, extract the pixels in the adjacent regions outside the boundary, call the gradient direction vector value of the pixel in the current frame image, and calculate the angle difference between the gradient direction and the center pixel. Filter the three directional regions whose angle difference does not exceed the brightness direction concentration determination threshold, extract the pixel gray values ​​respectively, and generate a directional consistent gray set. S212: Based on the pixel gray value sequence extracted from the directional consistent gray set, retrieve the gray value evolution trajectory of the same spatial location in consecutive image frames, remove frames with instantaneous brightness surges, filter pixel gray value sequences within the stable frame range, perform an arithmetic average on the sequences, obtain the central stable gray value, calculate the difference between the maximum and minimum values ​​of the sequence, construct a gray value stability fluctuation factor, and generate a stable gray value replacement set; S213: Call the central stable gray value in the stable gray value replacement value set to replace the gray value information of each abnormal region in the iron core reflection interference region map, optimize the continuity of the region boundary and the consistency of the texture, stitch and integrate the replacement region on the basis of the original image, output the reconstructed image frame sequence, and generate the reconstructed image of the iron core edge structure.

6. The transformer core detection method based on computer vision according to claim 5, characterized in that, The specific steps for obtaining the core symmetry disturbance structure diagram are as follows: S311: Call the edge pixels of the region where the iron core extends longitudinally and laterally in the image of the reconstructed image of the iron core edge structure, collect the two-dimensional coordinate index and corresponding gray-level gradient value of the edge pixel points, and establish equidistant point index pairs based on the central axis of the image to construct a symmetrical structure point mapping map and obtain the symmetrical structure index distribution information. S312: Based on the equidistant point pairs in the symmetrical structure index distribution information, extract the gray gradient value sequence, perform difference operation and integral normalization according to the index order, calculate and obtain the path symmetry offset of each symmetrical scanning path, filter the path symmetry offset greater than the symmetry offset threshold, mark it as the structural imbalance area, and extract the jump points on the path to obtain the symmetry gradient offset point set. S313: Call the coordinate index of the marked points in the symmetric gradient offset point set, and construct the jump point cluster block based on the horizontal and vertical adjacency features. Filter out segments with a span smaller than the region continuity judgment threshold, extract the continuously distributed edge point sequence that is stable in the offset direction with the central axis, and establish the core symmetry perturbation structure diagram.

7. The transformer core detection method based on computer vision according to claim 6, characterized in that, The specific steps for obtaining the component distribution diagram affected by the core anomaly are as follows: S411: Based on the marked abnormal region position coordinates in the core symmetry disturbance structure diagram, extract the boundary index range of each abnormal region in the two-dimensional coordinate system, and retrieve the structural contour boundary map of the clamping unit arrangement area, magnetic channel, and insulation overlap boundary of the corresponding image frame in the core structure database. Overlay the boundary map and the abnormal region boundary map to generate an abnormal component overlap index set. S412: Call the pixel coordinates recorded in the abnormal component overlap index set, calculate the boundary intersection length between the abnormal region and the component region respectively, and count the proportion of the pixel set corresponding to the intersection region in the component region. Based on the cumulative area ratio, calculate the percentage of pixel area covered by each type of component and generate abnormal component coverage index information. S413: Based on the coverage ratio and boundary intersection length of the components in the abnormal component coverage index information, filter the component areas with a coverage ratio greater than the influence judgment threshold, record the associated component names and location coordinate indexes of the corresponding abnormal areas, identify them as key parts with correlation influence areas, and generate a distribution map of components affected by iron core anomalies.

8. The transformer core detection method based on computer vision according to claim 7, characterized in that, The method further includes the following steps: S5: Call the anomaly type and the type of covered component in the distribution map of the components affected by the core anomaly, and combine the corresponding brightness reflectance ratio, symmetry offset amplitude and key structure overlap ratio. The brightness reflectance ratio is the ratio of the average brightness value of the abnormal area to the average stable gray value of the surrounding normal area. The interval is matched with the risk assessment reference indicators commonly used in transformer core detection applications. The label is assigned according to the risk level interval of the combined parameters to obtain the multi-type anomaly warning level layer of the core. The aforementioned multi-type anomaly early warning level layer for iron cores specifically refers to the risk level color partition map, the anomaly type level label group, the component corresponding risk label set, and the early warning level interval distribution map.

9. The transformer core detection method based on computer vision according to claim 8, characterized in that, The specific steps for obtaining the multi-type anomaly warning level layer of the iron core are as follows: S511: Call the identified abnormal areas in the distribution map of the components affected by the iron core abnormality, extract the abnormal type and the type of the covered components in each area, and associate the brightness reflectance ratio, symmetry offset amplitude and structural overlap ratio characteristic index of the corresponding area respectively. The brightness reflectance ratio is the ratio of the average brightness value of the abnormal area to the average stable gray value of the surrounding normal area. Construct a multi-dimensional attribute vector set and generate a set of abnormal component feature parameters. S512: Based on the set of characteristic parameters of the abnormal components, and based on brightness, structure and area proportion, calculate and obtain the risk measurement value of each abnormal area, perform interval matching with the risk level interval boundary value set, mark the corresponding risk label type, and generate a set of risk level labels for abnormal components. S513: Call the coordinate information of each risk level label and the corresponding abnormal area in the abnormal component risk level label set, bind the risk level label and the abnormal area to the location and draw the risk layer, and overlay it on the iron core structure reference map to establish a multi-type abnormal early warning level layer for the iron core.

Images