Porcelain insulator microdefect intelligent diagnosis system and method based on industrial CT detection

Abstract

The invention belongs to the technical field of insulator detection, and relates to a porcelain insulator microdefect intelligent diagnosis system and method based on industrial CT detection. The method comprises the following steps: acquiring an industrial CT detection picture of a porcelain insulator, and extracting a cement layer and a porcelain piece layer in a cross section and a longitudinal section; and identifying a defect area contained in the industrial CT detection picture through a yov11 instance segmentation algorithm, and classifying the defects into a cement layer pore defect and a porcelain layer crack defect in combination with the cement layer and the porcelain layer. According to the method, the problem of low efficiency of manual identification of the porcelain insulator industrial CT detection photo defect is solved, full-automatic defect identification is realized, and the workload of manual defect rechecking is greatly reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of non-destructive testing of insulators, and relates to an intelligent diagnosis system and method for micro-defects of porcelain insulators based on industrial CT detection. Background Art

[0002] Currently, the diagnosis of micro-defects of porcelain insulators based on industrial CT detection mainly relies on technicians to manually identify the detection results of industrial CT, which has a large workload and low efficiency. Currently, the indicators reflecting the micro-defects of porcelain insulators are mainly porcelain cracks and cement layer pores. When cracks appear in the porcelain, it indicates that there are major defect hidden dangers inside the insulator.

[0003] When using the manual identification method to carry out defect detection, since the number of industrial CT detection photos formed after industrial CT detection of a single sample is extremely large, the workload of manually identifying defects in each photo is huge and the efficiency is low. Summary of the Invention

[0004] The present invention provides an intelligent diagnosis system and method for micro-defects of porcelain insulators based on industrial CT detection, which realizes the fully automated detection of cracks and cement layer pores existing in the industrial CT detection photos of porcelain insulators.

[0005] The present invention is realized through the following technical solutions. An intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection includes: Step 1: Collect industrial CT detection photos of porcelain insulators, and extract the cement layer and porcelain layer in the cross-section and longitudinal section; Step 2: Identify the defect areas contained in the industrial CT detection photos through the yolov11 instance segmentation algorithm, and combine with the cement layer and porcelain layer to classify the defects into cement layer pore defects and porcelain layer crack defects; The process of extracting the cement layer and porcelain layer in the cross-section and longitudinal section includes the following sub-steps: Step S1: Extract the parameter information in each industrial CT detection photo, and the parameter information includes scene depth and pixel size information; Step S2: Extract the size features of the maximum longitudinal and cross-sections; Step S2-1: Find two maximum longitudinal section photos and one maximum cross-section photo from the industrial CT detection photos; Step S2-2: Extract the non-steel cap area, porcelain layer area, and steel foot area; Step S2-3: Extract the size features of the maximum cross-section photo and the maximum longitudinal section; Step S2-3-1: Calculate the distance between the steel foot and the steel cap of the maximum cross-section photo on the X-axis and Y-axis; Step S2-3-2: Calculate the distance between the widest position of the steel foot in the maximum longitudinal section and the top of the photo, as well as the horizontal distances between the widest position and the left and right sides of the steel cap; Step S3: Extract the cement layer and porcelain part layer of any longitudinal and transverse sections; Step S3-1: Calculate the positions of the horizontal projection lines of any transverse section on two maximum longitudinal sections; Step S3-2: Calculate the porcelain part layer of the transverse section; Step S3-3: Calculate the cement layer of the transverse section; Step S3-4: Calculate the porcelain part layer of the longitudinal section; Step S3-5: Calculate the cement layer of the longitudinal section.

[0006] Further preferably, the industrial CT detection photos include two groups of longitudinal section photos and one group of transverse section photos. The two groups of longitudinal section photos are respectively called the longitudinal section group 1 photos and the longitudinal section group 2 photos. The transverse section photo with the largest steel foot area in the same group of photos is called the maximum transverse section; the longitudinal section photo with the largest steel foot area in the same group of photos is called the maximum longitudinal section.

[0007] Further preferably, use the calibrated data set to train the yolov11 instance segmentation algorithm to obtain the non-steel cap area recognition model, the porcelain part layer area recognition model, and the steel foot area recognition model respectively; by calling the non-steel cap area recognition model, the porcelain part layer area recognition model, and the steel foot area recognition model, perform instance segmentation on the industrial CT detection photos, and sequentially extract the non-steel cap area, the porcelain part layer area, and the steel foot area.

[0008] Further preferably, the calculation of the porcelain part layer of the transverse section includes: Step S3-2-1: Calculate the distance between the porcelain part layer and the steel foot at the positions of the horizontal projection lines of any transverse section on two maximum longitudinal sections; Step S3-2-2: Calculate the distances between the porcelain part layer in any transverse section and the steel foot on the X-axis and Y-axis; Step S3-2-3: Draw the porcelain part layer area of any transverse section.

[0009] Further preferably, the calculation of the porcelain part layer of the longitudinal section includes: Step S3-4-1: Calculate the distances between the inner and outer boundaries of the porcelain part layer at any height of any longitudinal section and the central axis of the photo; Step S3-4-2: Calculate the porcelain part layer area of any longitudinal section.

[0010] The present invention solves the problem of low efficiency of manual identification of defects in industrial CT detection photos of porcelain insulators, realizes fully automated defect identification, and greatly reduces the workload of manual defect review. Description of the Drawings

[0011] Figure 1 This is the architecture diagram of the present invention.

[0012] Figure 2 This is the flowchart for extracting the cement layer and porcelain layer in the cross-section and longitudinal section.

[0013] Figure 3 This is the schematic diagram for extracting the steel feet using the yolov11 instance segmentation algorithm.

[0014] Figure 4 This is the mask diagram output by the yolov11 instance segmentation algorithm.

[0015] Figure 5 This is the schematic diagram of the distances between the steel feet and steel caps of the maximum cross-section in the X-axis and Y-axis directions.

[0016] Figure 6 This is the schematic diagram of the distances between the widest position of the steel feet and the top of the photo and the distance between the widest position and the steel cap in the maximum longitudinal section.

[0017] Figure 7 This is the schematic diagram of the intersection points of the horizontal line at any Y-axis value with the mask of the porcelain layer area and the steel foot area.

[0018] Figure 8 This is the schematic diagram of the distances between the porcelain parts and the steel feet in the X-axis and Y-axis directions on the cross-section.

[0019] Figure 9 This is the schematic diagram of 8 points on the boundary of the porcelain layer.

[0020] Figure 10 This is the mask diagram of the porcelain layer area.

[0021] Figure 11 For any longitudinal section Z i and the maximum longitudinal section Z max projected onto M j The projection relationship.

[0022] Figure 12 For connecting P 0 and the intersection points of the projection line of any longitudinal section with the inner and outer boundaries of the porcelain layer.

[0023] Figure 13 For the cross-section M j projected onto the maximum longitudinal section Z max The position of the projection line and the intersection points with the boundary of the porcelain layer.

[0024] Figure 14 This is the schematic diagram of the positional relationship of the coordinate points on the boundary. Detailed implementation method

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of this application.

[0026] Embodiment 1 As Figure 1 shown, an intelligent diagnosis system for micro-defects of porcelain insulators based on industrial CT detection includes: Porcelain insulator structure analysis module: Collect industrial CT detection photos of porcelain insulators, and extract the cement layer and porcelain part layer in the cross-section and longitudinal section; Porcelain insulator defect detection module: Identify the defect areas contained in the industrial CT detection photos through the yolov11 instance segmentation algorithm, and combine the cement layer and the porcelain part layer to classify the defects into cement layer pore defects and porcelain part layer crack defects.

[0027] Embodiment 2 This embodiment provides an intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection, including: Step 1: Collect industrial CT detection photos of porcelain insulators, and extract the cement layer and porcelain part layer in the cross-section and longitudinal section; Step 2: Identify the defect areas contained in the industrial CT detection photos through the yolov11 instance segmentation algorithm, and combine the cement layer and the porcelain part layer to classify the defects into cement layer pore defects and porcelain part layer crack defects.

[0028] Before detecting defects, it is necessary to know which areas in the photo are the porcelain part layer and which areas are the cement layer. This facilitates subsequent judgment of the defect type after detecting the defects in the industrial CT detection photos. Since the boundary between the cement layer and the porcelain part layer in the industrial CT detection photo is not very clear, it is very difficult to distinguish the cement layer and the porcelain part layer in the industrial CT detection photo simply by image recognition. To solve this problem, taking advantage of the feature that each layer of the porcelain insulator is distributed around the central axis of the porcelain insulator, the cement layer and the porcelain part layer in the industrial CT detection photo are calculated by calculating the size parameters of the porcelain part layer and the cement layer.

[0029] The process of extracting the cement layer and the porcelain part layer in the cross-section and longitudinal section includes the following sub-steps: Step S1: Extract the parameter information in each industrial CT detection photo, and the parameter information includes the scene depth and pixel size information; The scene depth refers to the distance between the cross section and the detector of the industrial CT when shooting. The scene depth is generally in the upper left corner of the industrial CT detection photo. Since the coordinate information area accounts for a small proportion of the photo, it cannot be directly extracted through OCR recognition. Therefore, it is necessary to combine the coordinate arrangement characteristics of the industrial CT detection photo to achieve accurate extraction. The 100x100 pixel area in the upper left corner of the industrial CT detection photo is cut out as a new photo and recognized by OCR, the text is extracted, the extracted text is traversed, and the numerical part of the text is found. The numerical part is the scene depth of the industrial CT detection photo. Since the pixel size of the industrial CT detection photo is represented by the scale bar located below the industrial CT detection photo, and since the area is relatively small, the proportional size cannot be effectively extracted directly through the AI ​​detection algorithm. Therefore, the bottom area of ​​the industrial CT detection photo is first cut out as a new photo, and then recognized by OCR.

[0030] Step S2: extracting the size features of the maximum longitudinal and transverse sections; Step S3: extracting cement layers and porcelain layers in any longitudinal and cross sections.

[0031] Specifically, the step S2 of extracting the size features of the maximum longitudinal and transverse sections includes the following sub-steps: Step S2-1: Find two maximum longitudinal sections and one maximum cross-section from the industrial CT detection photos; When doing industrial CT inspection, in order to ensure that all defects can be scanned, three mutually perpendicular directions will be selected for scanning, and two groups of longitudinal section photos and one group of cross-section photos will be obtained. The two longitudinal section photos are called longitudinal section group 1 photos and longitudinal section group 2 photos respectively. The cross-section photo with the largest area of ​​the steel foot area in the same group of photos is called the largest cross-section. The longitudinal section photo with the largest area of ​​the steel foot area in the same group of photos is called the largest longitudinal section.

[0032] In order to extract the maximum cross-section photo and the maximum longitudinal section, the yolov11 instance segmentation algorithm in this embodiment trains the corresponding recognition model to identify the steel feet in the industrial CT detection photos; instance segmentation is a complex task in the field of computer vision, which requires the model to be able to identify objects of different categories in the image and classify each individual object at the pixel level. Figure 3 As shown, the steel foot part in the industrial CT inspection photo is extracted using the yolov11 instance segmentation algorithm.

[0033] The construction of instance segmentation detection algorithm is similar to that of target detection, which mainly consists of three steps: sample annotation, model training and model detection. The differences between each step and target detection are as follows: 1. The sample annotation stage annotates polygonal areas instead of rectangular areas.

[0034] 2. The segmentation model rather than the object detection model is used in the model training stage. Its principle is mainly to obtain the bounding box of each object through the detection branch first. Then, the segmentation branch is used to extract which pixels in the photo belong to the instance pixels. Finally, the pixels within the region are extracted into a mask image through the bounding box.

[0035] 3. What the model detection stage outputs is a mask image (binary image, 1 represents the object, 0 represents not), as Figure 4 shown.

[0036] Due to the maximum cross-sectional photo and the maximum longitudinal section, the main feature is that the area of the steel foot region in this photo is the largest among the photos in the same group. Therefore, the area of the steel foot region in each industrial CT detection photo is calculated based on the steel foot region mask image (the sum of the number of pixels with a brightness value of 1 in the steel foot region mask image); finally, the photo with the largest area occupied by the steel foot in each group of photos is found.

[0037] The extraction process of the maximum cross-sectional photo is as follows: Traverse each set of cross-sectional photos (where i is the serial number of the cross-sectional photo in the set, is the i-th cross-sectional photo in the set, and n is the number of cross-sectional photos in the set). Use the yolov11 instance segmentation algorithm to extract the steel foot region mask image ; Calculate the total brightness of each steel foot region mask image , where is a function for summing the brightness of the picture, and what it returns is the sum of the brightness of the input picture; Traverse each , find the serial number i of the cross-sectional photo in the set when the value is the largest, and mark the corresponding cross-sectional photo as the maximum cross-sectional photo.

[0038] The extraction process of the maximum longitudinal section in the first group of longitudinal sections is as follows: Traverse the set of photos in the first group of longitudinal sections , where is the q-th longitudinal section photo in the first group of longitudinal sections, and k is the number of photos in the first group of longitudinal sections. Use the yolov11 instance segmentation algorithm to extract the steel foot region mask image ; Calculate the sum of its brightness for each steel foot region mask image ; Traverse each , find the serial number q of the photo in the first group of longitudinal sections when the value is the largest, and mark the corresponding photo in the first group of longitudinal sections as the maximum longitudinal section.

[0039] Similarly, the extraction process of the maximum longitudinal section in the second group of longitudinal sections is as follows: Traverse the set of photos in the second group of longitudinal sections , where is the e-th longitudinal section photo in the second group of longitudinal sections, r is the number of photos in the second group of longitudinal sections, and the yolov11 instance segmentation algorithm is used to extract the mask map of the steel foot area ; Calculate the sum of brightness for each mask map of the steel foot area ; Traverse each to find the serial number e of the photo in the second group of longitudinal sections when the sum of brightness reaches the maximum value, and the corresponding photo in the second group of longitudinal sections is the maximum longitudinal section of the second group of longitudinal sections.

[0040] Step S2-2: Extraction of non-steel cap area, porcelain layer area, and steel foot area; By sorting out and calibrating the non-steel cap area, porcelain layer area, and steel foot area in various industrial CT detection photos of porcelain insulators, a data set is formed. Then, the calibrated data set is used to train the yolov11 instance segmentation algorithm to obtain a non-steel cap area recognition model, a porcelain layer area recognition model, and a steel foot area recognition model respectively. By calling the non-steel cap area recognition model, the porcelain layer area recognition model, and the steel foot area recognition model to perform instance segmentation on the industrial CT detection photos, the non-steel cap area, the porcelain layer area, and the steel foot area are extracted in sequence; the corresponding maximum cross-section mask map sets are obtained. The maximum longitudinal section mask map set of the first group of longitudinal sections The maximum longitudinal section mask map set of the second group of longitudinal sections , where are the mask maps of the non-steel cap area of the maximum cross-section, the maximum longitudinal section of the first group of longitudinal sections, and the maximum longitudinal section of the second group of longitudinal sections respectively. are the mask maps of the porcelain layer area of the maximum cross-section, the maximum longitudinal section of the first group of longitudinal sections, and the maximum longitudinal section of the second group of longitudinal sections respectively. are the mask maps of the steel foot area of the maximum cross-section, the maximum longitudinal section of the first group of longitudinal sections, and the maximum longitudinal section of the second group of longitudinal sections respectively.

[0041] Step S2-3: Extraction of size features of the maximum cross-section photo and the maximum longitudinal section; Step S2-3-1: Calculation of the distances between the steel foot and the steel cap of the maximum cross-section photo on the X-axis and Y-axis: Let the origin O of the plane coordinate system A in the maximum cross-section photo be the upper left corner of the photo, the positive direction of the X-axis be the horizontal direction from left to right, the positive direction of the Y-axis be the vertical direction from top to bottom, and the length unit on the X-axis direction and the Y-axis direction be mm. The distances between the steel foot and the steel cap of the maximum cross-section on the X-axis and Y-axis include 4 distances (as Figure 5 shown), which are the distance between the steel foot and the steel cap in the negative X-axis direction , the distance between the steel foot and the steel cap in the positive X-axis direction , the distance between the steel foot and the steel cap in the negative Y-axis direction , the distance between the steel foot and the steel cap in the positive Y-axis direction . To calculate these 4 distances, first calculate the bounding box of the steel foot area in the plane coordinate system A of the maximum cross-sectional photo , and the bounding box of the non-steel cap area in the plane coordinate system A of the maximum cross-sectional view . Among them is the leftmost X coordinate of the steel foot bounding box, is the rightmost X coordinate of the steel foot bounding box, is the uppermost Y coordinate of the steel foot bounding box, is the lowermost Y coordinate of the steel foot bounding box. is the leftmost X coordinate of the non-steel cap area bounding box, is the rightmost X coordinate of the non-steel cap area bounding box, is the uppermost Y coordinate of the non-steel cap area bounding box, is the lowermost Y coordinate of the non-steel cap area bounding box.

[0042] Then, through the formulas , , , calculate the 4 distances between the steel foot and the steel cap on the X-axis and Y-axis of the maximum cross-section.

[0043] Calculate the bounding box of the steel foot area through the steel foot area mask map of the maximum cross-section : Traverse each pixel of the steel foot area mask map of the maximum cross-section , record the pixel coordinates of each pixel with a non-zero brightness value, and form a set of pixel coordinates: . Among them is the pixel coordinate of the steel foot area mask map, is the pixel abscissa of the steel foot area mask map, is the pixel ordinate of the steel foot area mask map, is the number of pixels in the horizontal direction, is the number of pixels in the vertical direction; traverse set, and respectively find the maximum value and the minimum value of the pixel abscissa of the steel foot area mask map, the maximum value and the minimum value of the pixel ordinate of the steel foot area mask map; then , , , , where F is the pixel size of the image.

[0044] Mask diagram of the non-steel cap area through the maximum cross-section Perform the calculation of the bounding box for the non-steel cap area: Traverse the mask diagram of the non-steel cap area of the maximum cross-section For each pixel, record the pixel coordinates of each pixel with a non-zero luminance value to form a set of pixel coordinates: , where is the pixel coordinate of the non-steel cap area mask diagram, is the abscissa of the pixel of the non-steel cap area mask diagram, is the ordinate of the pixel of the non-steel cap area mask diagram, is the number of pixels in the horizontal direction, is the number of pixels in the vertical direction; traverse the set, and respectively find the maximum value and the minimum value of the abscissa of the pixel of the non-steel cap area mask diagram, and the minimum value of the ordinate of the pixel of the non-steel cap area mask diagram, then , , , , where F is the pixel size of the image.

[0045] Step S2-3-2: Calculate the distance between the widest position of the steel leg and the top of the photo and the horizontal distances between the widest position and the left and right sides of the steel cap in the maximum longitudinal section: Let there be a two-dimensional plane coordinate system in the maximum longitudinal section, with the origin O of the coordinate system at the upper left corner of the photo, the positive direction of the X-axis being the horizontal direction from left to right, and the positive direction of the Y-axis being the vertical direction from top to bottom; the distance between the widest position of the steel leg and the top of the photo in the maximum longitudinal section , the distance between the widest position of the steel leg and the left side of the steel cap and the distance between the widest position of the steel leg and the right side of the steel cap , as shown in Figure 6 . Since there are photos in 2 longitudinal section groups, therefore , , also have 2 sets of corresponding data. Let the distance between the widest position of the steel leg and the top of the photo in the maximum longitudinal section of longitudinal section group 1 be , and the horizontal distances between the widest position of the steel leg and the left and right sides of the steel cap in the maximum longitudinal section of longitudinal section group 1 be , .

[0046] Similarly, let the distance between the widest position of the steel leg and the top of the photo in the maximum longitudinal section of longitudinal section group 2 be , the horizontal distances between the widest position of the steel foot and the left and right sides of the steel cap in the largest longitudinal section of the two groups of longitudinal sections are , .

[0047] The distance between the widest position of the steel foot in the largest longitudinal section of the first group of longitudinal sections and the top of the photo The calculation process is as follows: Statistically calculate the sum of the brightness of the steel foot area mask map of the largest longitudinal section of the first group of longitudinal sections row by row to obtain the set of brightness values of each row of the steel foot area mask map , where is the row number of any row of the steel foot area mask map , is the sum of the brightness of all pixels in the th row of the steel foot area mask map, is the pixel height of the steel foot area mask map ; find the row number with the highest brightness value in the set , and let this row number be , .

[0048] Similarly, the distance between the widest position of the steel foot in the largest longitudinal section of the second group of longitudinal sections and the top of the photo The calculation process is as follows: Statistically calculate the sum of the brightness of the steel foot area mask map of the largest longitudinal section of the second group of longitudinal sections row by row to obtain the set of brightness values of each row of the steel foot area mask map . Among them is the row number of any row of the steel foot area mask map , is the sum of the brightness of all pixels in the th row of the steel foot area mask map, is the pixel height of the steel foot area mask map ; find the row number with the highest brightness value in the set , and let this row number be , then .

[0049] The horizontal distances between the steel foot and the left and right sides of the steel cap at the widest position of the steel foot in the largest longitudinal section of the first group of longitudinal sections are , The calculation process is as follows: Read the steel foot area mask map of the largest longitudinal section of the first group of longitudinal sections , and find the Y-axis coordinate value of the widest position of the steel foot in the pixel coordinate system ; traverse all pixels in the th row to find the pixel coordinates of the first pixel with a non-zero brightness value; Find the pixel coordinates of the last pixel with a non-zero brightness value ; Traverse the non-steel cap area mask map of the maximum longitudinal section in Group 1 of the longitudinal section The coordinates of the first pixel with a non-zero brightness value among all pixels in the row are found ; Then ; .

[0050] Similarly, the horizontal distances between the widest positions of the steel feet and the left and right sides of the steel cap in the maximum longitudinal section of Group 2 of the longitudinal section 、 are calculated as follows: Read the steel foot area mask map of the maximum longitudinal section in Group 2 of the longitudinal section , and find the Y-axis coordinate value of the widest position of its steel foot in the pixel coordinate system ; Traverse the coordinates of the first pixel with a non-zero brightness value and the pixel coordinates of the last pixel with a non-zero brightness value among all pixels in the row are found ; Traverse the non-steel cap area mask map of the maximum longitudinal section in Group 2 of the longitudinal section the coordinates of the first pixel with a non-zero brightness value are found ; The pixel coordinates of the last pixel with a non-zero brightness value are found , then: ; .

[0051] Specifically, extracting the cement layer and porcelain part layer of any longitudinal and transverse sections includes the following sub-steps: Step S3-1: Calculation of the horizontal projection line position of any transverse section on two maximum longitudinal sections: Step S3-2: Calculation of the porcelain part layer of the transverse section; Step S3-3: Calculation of the cement layer of the transverse section; Step S3-4: Calculation of the porcelain part layer of the longitudinal section; Step S3-5: Calculation of the cement layer of the longitudinal section.

[0052] Among them, the calculation of the horizontal projection line position of any transverse section on two maximum longitudinal sections: Given the scene depth value of any transverse section , find the Y-axis coordinate value of the horizontal projection line of this transverse section in the plane coordinate system A on two maximum longitudinal sections. The calculation process is as follows: For any transverse section , the calculation formula for the Y coordinate value of the horizontal projection line in the 1st group of maximum longitudinal sections in the longitudinal section (Y-axis coordinate in the plane coordinate system A) is: , where is the y-axis coordinate of the horizontal projection line of the maximum cross-section in the 1st group of maximum longitudinal sections in the longitudinal section, and its value is numerically equal to , g is the scene depth distance between any cross-section and the maximum cross-section .

[0053] For any cross-section , the calculation formula for the Y coordinate value of the horizontal projection line in the 2nd group of maximum longitudinal sections in the longitudinal section (Y-axis coordinate in the plane coordinate system A) is: , where is the y-axis coordinate of the horizontal projection line of the maximum cross-section in the 2nd group of maximum longitudinal sections in the longitudinal section, and its value is numerically equal to .

[0054] Among them, the cross-section porcelain part layer calculation includes: Step S3-2-1: Calculation of the distance between the porcelain part layer and the steel foot at the position of the horizontal projection line of any cross-section on two maximum longitudinal sections: It is known that the Y coordinate values of the horizontal projection lines of the cross-section on two maximum longitudinal sections are , , and find the distances between the inner and outer boundaries of the porcelain part layer and the steel foot at the position of the horizontal projection line.

[0055] In the maximum longitudinal section, after superimposing the porcelain part layer area mask map and the steel foot area mask map, for any given Y-axis coordinate value , making a horizontal projection line, 4 intersection points of the horizontal projection line of with the porcelain part layer mask boundary and 2 intersection points of the steel foot mask boundary can be obtained. As shown in Figure 7 , the intersection points from left to right are the intersection point of the horizontal projection line of with the left outer boundary of the porcelain part layer mask, the intersection point of the horizontal projection line of with the left inner boundary of the porcelain part layer mask, the intersection point of the horizontal projection line of with the left boundary of the steel foot mask boundary, the intersection point of the horizontal projection line of with the right boundary of the steel foot mask boundary, the intersection point of the horizontal projection line of with the right inner boundary of the porcelain part layer mask, and the intersection point of the horizontal projection line of with the right outer boundary of the porcelain part layer mask.

[0056] Since the industrial CT detection image contains two sets of longitudinal sections corresponding to two maximum longitudinal sections, there will also be two sets of the above six intersection points. For the photo of the maximum longitudinal section of one set of longitudinal sections, any given Y-axis coordinate value , let the intersection points be the intersection points of the horizontal projection lines of in the maximum longitudinal section of one set of longitudinal sections and the left outer boundary of the porcelain part layer mask , the intersection points of the horizontal projection lines of in the maximum longitudinal section of one set of longitudinal sections and the left inner boundary of the porcelain part layer mask , the intersection points of the horizontal projection lines of in the maximum longitudinal section of one set of longitudinal sections and the left boundary of the steel foot mask boundary , the intersection points of the horizontal projection lines of in the maximum longitudinal section of one set of longitudinal sections and the right boundary of the steel foot mask boundary , the intersection points of the horizontal projection lines of in the maximum longitudinal section of one set of longitudinal sections and the right inner boundary of the porcelain part layer mask , the intersection points of the horizontal projection lines of in the maximum longitudinal section of one set of longitudinal sections and the right outer boundary of the porcelain part layer mask ; correspondingly, in the maximum longitudinal section of one set of longitudinal sections, the distance from the outside of the porcelain part layer to the left side of the steel foot is , the distance from the inside of the porcelain part layer to the left side of the steel foot is , the distance from the outside of the porcelain part layer to the right side of the steel foot is , the distance from the inside of the porcelain part layer to the right side of the steel foot is , where Dist is the distance calculation formula , is the X-axis coordinate of point , is the Y-axis coordinate of point , is the X-axis coordinate of point , is the Y-axis coordinate of point , where, .

[0057] For the photo of the maximum longitudinal section in the second set of longitudinal sections, any given Y-axis coordinate value , let the intersection points be the intersection points of the horizontal projection lines of in the maximum longitudinal section of the second set of longitudinal sections and the left outer boundary of the porcelain part layer mask , the intersection points of the horizontal projection lines of in the maximum longitudinal section of the second set of longitudinal sections and the left inner boundary of the porcelain part layer mask , the intersection points of the horizontal projection lines of in the maximum longitudinal section of the second set of longitudinal sections and the left boundary of the steel foot mask boundary , in the maximum longitudinal section of the second set of longitudinal sections The intersection point of the horizontal projection line and the right boundary of the steel foot mask Among the two maximum longitudinal sections of the longitudinal section The intersection point of the horizontal projection line and the right inner boundary of the porcelain part layer mask Among the two maximum longitudinal sections of the longitudinal section The intersection point of the horizontal projection line and the right outer boundary of the porcelain part layer mask ; Correspondingly, in the maximum longitudinal section of the two groups of longitudinal sections, the distance from the outside of the porcelain part layer to the left side of the steel foot is , the distance from the inside of the porcelain part layer to the left side of the steel foot is , the distance from the outside of the porcelain part layer to the right side of the steel foot is , the distance from the inside of the porcelain part layer to the right side of the steel foot is .

[0058] Step S3-2-2: Calculation of the distances between the porcelain part layer and the steel foot on the X-axis and Y-axis in any cross-section: It is known that the distances between the porcelain part layer and the steel foot at the positions of the horizontal projection lines of any cross-section on the two maximum longitudinal sections are respectively: , . It is necessary to calculate the distances between the inner and outer boundaries of the porcelain part layer of this cross-section and the steel foot on the X-axis and Y-axis as shown in Figure 8 . There are a total of 8 distance parameters, which are in turn: the distance from the outside of the porcelain part layer to the left side of the steel foot in the X-axis direction ; the distance from the inside of the porcelain part layer to the left side of the steel foot in the X-axis direction ; the distance from the outside of the porcelain part layer to the right side of the steel foot in the X-axis direction ; the distance from the inside of the porcelain part layer to the right side of the steel foot in the X-axis direction ; the distance from the outside of the porcelain part layer to the upper side of the steel foot in the Y-axis direction ; the distance from the inside of the porcelain part layer to the upper side of the steel foot in the Y-axis direction ; the distance from the outside of the porcelain part layer to the lower side of the steel foot in the Y-axis direction ; the distance from the inside of the porcelain part layer to the lower side of the steel foot in the Y-axis direction .

[0059] Since the widest position of the steel foot in the maximum longitudinal section is the horizontal projection position of the maximum cross-section, therefore , , , and , , , are equal in pairs. Therefore, by matching , , , and , , , The value, and then the horizontal distances between the widest positions of the steel feet and the left and right sides of the steel cap in the two largest longitudinal sections can be found. Are they the projections of the steel feet and the steel cap in the X-axis direction or the Y-axis direction of the largest cross-section?

[0060] In addition, since the projection direction of any cross-section on the largest longitudinal section is the same as the projection direction of the largest cross-section on the largest longitudinal section, the projection direction of any cross-section on the largest longitudinal section can be obtained based on the projection direction of the projection of the largest cross-section on the largest longitudinal section.

[0061] After obtaining the projection directions of any cross-section on the two largest longitudinal sections, combined with the horizontal projection positions of this cross-section on the 2 largest longitudinal sections , and the porcelain layer area mask diagrams and steel foot area mask diagrams of the 2 largest longitudinal sections, the distances between the porcelain layer and the steel feet at the horizontal projection positions , can be obtained. , , , , , , , .

[0062] Finally, by combining the projection directions of the first group of the largest longitudinal sections and the second group of the largest longitudinal sections on any cross-section, the , , , , , , , and , , , , , , , relationship can be found.

[0063] The specific matching rules are as follows: When , When; for the maximum steel foot position of the largest longitudinal section in the first group of longitudinal sections, it is the projection in the X-axis direction of the cross-section; for the maximum steel foot position of the largest longitudinal section in the second group of longitudinal sections, it is the projection in the Y-axis direction of the cross-section. And the X-axis direction of the largest longitudinal section in the first group of longitudinal sections is the same as the X-axis direction of this cross-section, and the X-axis of the largest longitudinal section in the second group of longitudinal sections is the same as the Y-axis direction of this cross-section. Then the distance parameter values between the porcelain part and the steel foot on the X-axis and Y-axis in this cross-section are: 、 、 、 、 、 、 、 。

[0064] When , ; for the maximum steel foot position of the largest longitudinal section in the first group of longitudinal sections, it is the projection in the X-axis direction of the cross-section. For the maximum steel foot position of the largest longitudinal section in the second group of longitudinal sections, it is the projection in the Y-axis direction of the cross-section. And the X-axis direction of the largest longitudinal section in the first group of longitudinal sections is opposite to the X-axis direction of this cross-section, and the X-axis of the largest longitudinal section in the second group of longitudinal sections is the same as the Y-axis direction of this cross-section. Then the distance parameter values between the porcelain part and the steel foot on the X-axis and Y-axis in this cross-section are: 、 、 、 、 、 、 、 。

[0065] When , ; for the maximum steel foot position of the largest longitudinal section in the first group of longitudinal sections, it is the projection in the X-axis direction of the cross-section. For the maximum steel foot position of the largest longitudinal section in the second group of longitudinal sections, it is the projection in the Y-axis direction of the cross-section. And the X-axis direction of the largest longitudinal section in the first group of longitudinal sections is the same as the X-axis direction of this cross-section, and the X-axis of the largest longitudinal section in the second group of longitudinal sections is opposite to the Y-axis direction of this cross-section. Then the distance parameter values between the porcelain part and the steel foot on the X-axis and Y-axis in this cross-section are: 、 、 、 、 、 、 、 。

[0066] When , When; for the first set of longitudinal sections, the position of the maximum steel feet of the maximum longitudinal section is the projection in the X-axis direction of the cross-section. For the second set of longitudinal sections, the position of the maximum steel feet of the maximum longitudinal section is the projection in the Y-axis direction of the cross-section. And the X-axis direction of the maximum longitudinal section of the first set of longitudinal sections is opposite to the X-axis direction of this cross-section, and the X-axis of the maximum longitudinal section of the second set of longitudinal sections is opposite to the Y-axis direction of this cross-section. Then the distance parameter values between the porcelain part and the steel feet on the X-axis and Y-axis in this cross-section are: 、 、 、 、 、 、 、 。

[0067] When , ; for the first set of longitudinal sections, the position of the maximum steel feet of the maximum longitudinal section is the projection in the Y-axis direction of the cross-section. For the second set of longitudinal sections, the position of the maximum steel feet of the maximum longitudinal section is the projection in the X-axis direction of the cross-section. And the X-axis direction of the maximum longitudinal section of the first set of longitudinal sections is the same as the Y-axis direction of this cross-section, and the X-axis of the maximum longitudinal section of the second set of longitudinal sections is the same as the X-axis direction of this cross-section. Then the distance parameter values between the porcelain part and the steel feet on the X-axis and Y-axis in this cross-section are: 、 、 、 、 、 、 、 。

[0068] When , ; for the first set of longitudinal sections, the position of the maximum steel feet of the maximum longitudinal section is the projection in the Y-axis direction of the cross-section. For the second set of longitudinal sections, the position of the maximum steel feet of the maximum longitudinal section is the projection in the X-axis direction of the cross-section. And the X-axis direction of the maximum longitudinal section of the first set of longitudinal sections is opposite to the Y-axis direction of this cross-section, and the X-axis of the maximum longitudinal section of the second set of longitudinal sections is the same as the X-axis direction of this cross-section. Then the distance parameter values between the porcelain part and the steel feet on the X-axis and Y-axis in this cross-section are: 、 、 、 、 、 、 、 。

[0069] When , When; for the first set of longitudinal sections, the position of the maximum steel leg of the maximum longitudinal section is the projection in the Y-axis direction of the cross-section. For the second set of longitudinal sections, the position of the maximum steel leg of the maximum longitudinal section is the projection in the X-axis direction of the cross-section. And the X-axis direction of the maximum longitudinal section of the first set of longitudinal sections is the same as the Y-axis direction of this cross-section, while the X-axis of the maximum longitudinal section of the second set of longitudinal sections is opposite to the X-axis direction of this cross-section. Then the distance parameter values between the porcelain part and the steel legs on the X-axis and Y-axis in this cross-section are: 、 、 、 、 、 、 、 。

[0070] When , ; for the first set of longitudinal sections, the position of the maximum steel leg of the maximum longitudinal section is the projection in the Y-axis direction of the cross-section. For the second set of longitudinal sections, the position of the maximum steel leg of the maximum longitudinal section is the projection in the X-axis direction of the cross-section. And the X-axis direction of the maximum longitudinal section of the first set of longitudinal sections is opposite to the Y-axis direction of this cross-section, while the X-axis of the maximum longitudinal section of the second set of longitudinal sections is opposite to the X-axis direction of this cross-section. Then the distance parameter values between the porcelain part and the steel legs on the X-axis and Y-axis in this cross-section are: 、 、 、 、 、 、 、 。

[0071] Step S3-2-3: Drawing the porcelain part layer area in any cross-section: When calculating the distances between the porcelain part layer in any cross-section and the steel legs in the X-axis and Y-axis directions 、 、 、 、 、 、 、 are obtained, 8 points on the inner and outer boundaries of the porcelain part layer can be determined on the cross-section. Among them, the 4 upper, lower, left, and right points on the outer boundary are respectively , and the 4 upper, lower, left, and right points on the inner boundary are respectively: , as shown in Figure 9 . Create a mask image with the same size as this cross-section, and the default brightness value is all 0. On this mask image, using the coordinate values of the 4 points on the outer boundary: , and the 4 points on the inner boundary: , draw the inner and outer rings of the porcelain part respectively. And fill the brightness between the inner and outer rings with 1. Finally, a mask image of the porcelain part layer area matching this cross-section will be obtained, as shown inFigure 10 as shown

[0072] Among them, the calculation process of the cross-sectional cement layer is as follows: After obtaining the mask image of the porcelain part layer area of the cross-section , by calling the segmentation algorithm, the mask image of the non-steel cap area of the cross-section is extracted ; Subtract the brightness of the corresponding coordinate positions of the two mask images of the porcelain part layer area and the non-steel cap area, and the obtained result is the mask image of the cement layer of the cross-section .

[0073] Among them, the calculation process of the longitudinal-section porcelain part layer is as follows: Step S3-4-1: Calculate the distances between the inner and outer boundaries of the porcelain part layer at any height of any longitudinal section and the central axis of the photo

[0074] Taking a group of photos of the longitudinal section as an example, taking any cross-section M j as the projection plane, project any longitudinal section Z i and the maximum longitudinal section Z max onto M j ; Then the projection relationship is as Figure 11 shown, and there is the following relationship: 1. The two projection lines are parallel, and the distance between the two projection lines is the difference in the scene depth values between any longitudinal section and the maximum longitudinal section

[0075] 2. When taking any longitudinal section, the shooting range of its detector is the same. Therefore, the lengths of the projection lines of Z i and Z max on M j are the same

[0076] 3. Based on the above two points, it can be inferred that the connection line between the center position P i of the projection line of any longitudinal section Z j on M i and the center position P j of the projection line of the maximum longitudinal section on M 0 is perpendicular to the X-axis of this cross-section M j , that is, the line segment P i P 0 is perpendicular to the X-axis

[0077] As Figure 12 shown, based on Figure 11 , draw the area of the porcelain part layer of the cross-section, and connect P 0 and the intersection points of the projection line of any longitudinal section Z i with the inner and outer boundaries of the porcelain part layer Figure 12 The meanings of each intersection point in a are: P jThe inner side of the porcelain part layer and the longitudinal section Z i The left intersection point of the projection line, P b Is the cross-section M j The outer side of the porcelain part layer and the longitudinal section Z i The right intersection point of the projection line, P c Is the cross-section M j The inner side of the porcelain part layer and the longitudinal section Z max The left intersection point of the projection line, P d Is the cross-section M j The outer side of the porcelain part layer and the longitudinal section Z max The right intersection point of the projection line.

[0078] Based on Figure 12 , there are the following conclusions: 1. The length of the line segment Satisfies , where Is the scene depth value of the largest longitudinal section, Is the scene depth value of any longitudinal section.

[0079] 2. The length of the line segment Satisfies the formula , where Is the length of the line segment , Is the length of the line segment .

[0080] 3. Since the porcelain part layer is circular, therefore , so .

[0081] 4. Similarly, the length of the line segment Satisfies the formula , where Is the length of the line segment , Is the length of the line segment .

[0082] 5. Similarly, since the porcelain part layer is circular, therefore , therefore .

[0083] 6. In the above process, The value of is equal to the distance from the center point P j Of the cross-section M on the projection line position of the largest longitudinal section Z max To the right outer boundary point P of the porcelain part layer cj , fj The value of is equal to the distance from the center point P of the projection line of the cross-section M j On the largest longitudinal section Z max ​At the position of the upper projection line, the center point P of the projection line cj to the left inner boundary point P of the porcelain part layer bj The distance, as Figure 13 shown

[0084] Among them, the distance from the projection line to the top of the photo , is the y-axis coordinate of the horizontal projection line of the largest cross-section in the first group of the largest longitudinal sections of the longitudinal section, and its numerical value is the same as equal

[0085] Therefore Figure 13 in , .

[0086] Also, since the height of the photo is the same in the same group of longitudinal sections. Therefore, for the cross-section M j at the projection position of any longitudinal section Zi, the distance R j between the right outer boundary of the porcelain part layer of the longitudinal section Zi and the center point of the projection position of the cross-section M mj is numerically equal to , and the distance E mj between the left inner boundary of its porcelain part layer and the center point of this horizontal position is numerically equal to .

[0087] In the same way, the distance between the left outer boundary of the porcelain part layer of the longitudinal section Zi and the center point of this horizontal position at the horizontal position of the distance from the top of the photo can be calculated. Suppose it is B mj , and the distance between the left inner boundary of its porcelain part layer and the center point of this horizontal position, suppose it is Q mj .

[0088] Through the above process, the distances of the inner and outer boundaries of the porcelain part layer of any longitudinal section Z i from the center point of the photo at any height can be calculated, and finally a set i of the distances of the inner and outer boundaries of the porcelain part layer of the photo Z from the center point of the photo at any height is formed, where is the distance between the left outer boundary of its porcelain part layer and the center point of the horizontal position at the height , is the distance between the left inner boundary of its porcelain part layer and the center point of the horizontal position at the height , is the distance between the right inner boundary of its porcelain part layer and the center point of the horizontal position at the height , is the distance between the right outer boundary of its porcelain part layer and the center point of the horizontal position at the height The distance between the right outer boundary of the porcelain part layer at this height and the center point of the horizontal position at this height.

[0089] Step S3-4-2: Calculation of the porcelain part layer area of any longitudinal section: Step S3-4-2-1: Create a mask image i with the same size as the longitudinal section Z , and its brightness is defaulted to 0; Step S3-4-2-2: Assume that in the plane coordinate system A, the width of the longitudinal section is , and the height is , and the pixel size is ; Step S3-4-2-3: Create an empty set to record the point coordinates on the boundary of the porcelain part layer; Step S3-4-2-4: Based on the plane coordinate system A, define a variable to record the y-axis coordinate value, with an initial value of 0; Step S3-4-2-5: Calculate the set of the distances from the inner and outer boundaries of the porcelain part layer at the height of to the central axis of the photo; ; Step S3-4-2-6: When is not empty, then put the coordinate points , , , into the set , and the positional relationship of the four coordinate points on the boundary is as shown in Figure 14 ; Step S3-4-2-7: Increase the value of by the size of one pixel , and then repeat steps S3-3-2-5 to S3-4-2-6 until ends; Step S3-4-2-8: Draw a closed polygon on the mask image with the point coordinates in the set , fill the brightness of the polygon area with 1, and finally this mask image is the mask image of the porcelain part layer area of the longitudinal section Z i .

[0090] The calculation of the cement layer of the longitudinal section referred to in Step S3-5 is: Draw the mask image of the cement layer area of all longitudinal section pictures. For each longitudinal section picture, subtract the brightness of the corresponding pixel of the steel foot area mask image and the porcelain part layer area mask image from the brightness of each pixel of the non-steel cap area mask image, then the mask image of the cement layer area of the longitudinal section picture is obtained.

[0091] In this embodiment, the specific process of step 2 is as follows: Step A1: For any industrial CT inspection photo, use the yolov11 instance segmentation algorithm for defect detection. Let each defect mask map in the industrial CT inspection photo be , and the mask map of the porcelain part layer area of the industrial CT inspection photo be ; Step A2: Traverse , find the pixels with non-zero brightness values among them, and record the pixel coordinates in a set , is the abscissa of the i-th pixel of the s-th defect mask map, is the abscissa of the i-th pixel of the s-th defect mask map; Step A3: Traverse each pixel coordinate in the set , for each pixel coordinate, read the brightness value of the pixel at the same position in its porcelain part layer mask map; if the brightness value of the pixel at the corresponding position in the porcelain part layer mask map is not 0, then the defect is a crack in the porcelain part layer, otherwise it is a pore in the cement layer.

[0092] What is disclosed above are only some preferred embodiments of the present invention. Of course, the scope of the rights of the present invention cannot be limited thereby. Those of ordinary skill in the art can understand the entire or partial processes of implementing the above embodiments, and the equivalent changes made according to the claims of the present invention still fall within the scope covered by the invention.

Claims

1. An intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection, characterized in that: include: Step 1: Collect industrial CT test photos of porcelain insulators and extract cement layers and porcelain layers in cross sections and longitudinal sections; Step 2: Use the yolov11 instance segmentation algorithm to identify the defect areas contained in the industrial CT inspection photos, and combine the cement layer and the porcelain layer to classify the defects into pore defects in the cement layer and crack defects in the porcelain layer; The process of extracting the cement layer and the porcelain layer in the cross section and the longitudinal section comprises the following sub-steps: Step S1: extracting parameter information from each industrial CT detection photo, the parameter information including scene depth and pixel size information; Step S2: extracting the size features of the maximum longitudinal and transverse sections; Step S2-1: Find two maximum longitudinal sections and one maximum cross-section from the industrial CT detection photos; Step S2-2: extracting the non-steel cap area, the porcelain layer area, and the steel foot area; Step S2-3: Extracting the size features of the largest cross-section photo and the largest longitudinal section; Step S2-3-1: Calculate the distance between the steel foot and the steel cap on the X-axis and Y-axis of the maximum cross-section photo; Step S2-3-2: Calculate the distance between the widest position of the steel foot and the top of the photo in the maximum longitudinal section, and the horizontal distance between the widest position and the left and right sides of the steel cap; Step S3: extracting cement layers and porcelain layers in any longitudinal and cross sections; Step S3-1: Calculate the position of the horizontal projection line of any cross section on the two largest longitudinal sections; Step S3-2: Calculation of cross-section porcelain layers; Step S3-3: Calculation of cross-section cement layer; Step S3-4: longitudinal section porcelain layer calculation; Step S3-5: Calculation of the longitudinal section cement layer.

2. The intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection according to claim 1 is characterized in that: Industrial CT detection photos include 2 groups of longitudinal section photos and one group of cross-section photos. The 2 groups of longitudinal section photos are respectively called longitudinal section group 1 photos and longitudinal section group 2 photos. The cross-section photo with the largest steel foot area in the same group of photos is called the largest cross-section; the longitudinal section photo with the largest steel foot area in the same group of photos is called the largest longitudinal section.

3. The intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection according to claim 2 is characterized in that: The calibrated data set is used to train the yolov11 instance segmentation algorithm, and the non-steel cap area recognition model, the porcelain layer area recognition model, and the steel foot area recognition model are obtained respectively; the industrial CT detection photos are segmented by instance by calling the non-steel cap area recognition model, the porcelain layer area recognition model, and the steel foot area recognition model, and the non-steel cap area, the porcelain layer area, and the steel foot area are extracted in turn; Get the corresponding maximum cross-section mask map set respectively ; Longitudinal section 1 set of maximum longitudinal section mask map collection ; 2 sets of maximum longitudinal section mask maps for longitudinal sections ,in They are the non-steel cap area mask images of the maximum cross section, the maximum longitudinal section of the longitudinal section 1 group, and the maximum longitudinal section of the longitudinal section 2 group; They are the mask images of the porcelain layer area of ​​the largest cross section, the largest longitudinal section of the longitudinal section 1 group, and the largest longitudinal section of the longitudinal section 2 group; They are the steel foot area mask images of the maximum cross section, the maximum longitudinal section of the longitudinal section 1 group, and the maximum longitudinal section of the longitudinal section 2 group.

4. The intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection according to claim 3 is characterized in that: The distance calculation process between the steel foot and the steel cap on the X-axis and Y-axis of the maximum cross-section photo is as follows: Let the origin O of the plane coordinate system A in the maximum cross-section photo be the upper left corner of the photo, the positive direction of the X-axis be the horizontal direction from left to right, and the positive direction of the Y-axis be the vertical direction from top to bottom. The distance between the steel foot and the steel cap on the X-axis and Y-axis of the maximum cross-section includes 4 distances, which are the distance between the steel foot and the steel cap in the negative direction of the X-axis. , the distance between the steel foot and the steel cap in the positive direction of the X axis , the distance between the steel foot and the steel cap in the negative direction of the Y axis , the distance between the steel foot and the steel cap in the positive direction of the Y axis ; The steel foot area is bounded by the plane coordinate system A in the maximum cross-section photo , and the bounding box of the non-steel cap area in the plane coordinate system A in the maximum cross-sectional view ;in is the leftmost X coordinate of the steel foot bounding box, is the rightmost X coordinate of the steel foot bounding box, is the top Y coordinate of the steel foot bounding box, The bottom Y coordinate of the steel foot bounding box; is the leftmost X coordinate of the bounding box of the non-steel cap area, is the rightmost X coordinate of the bounding box of the non-steel cap area, is the top Y coordinate of the bounding box of the non-steel cap area, is the bottom Y coordinate of the bounding box of the non-steel cap area; , , , Calculate the four distances between the steel foot and the steel cap on the X-axis and Y-axis of the maximum cross-section.

5. The intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection according to claim 4 is characterized in that: Steel foot area mask through the largest cross section Calculate the bounding box of the steel foot area: Mask map of the steel foot area traversing the maximum cross section For each pixel, record the pixel coordinates of each pixel whose brightness value is not 0 to form a pixel coordinate set: ;in is the pixel coordinate of the steel foot area mask map, is the pixel horizontal coordinate of the steel foot area mask map, is the pixel ordinate of the steel foot area mask image, is the number of pixels in the horizontal direction, is the number of pixels in the vertical direction; traverse Set, find the maximum value of the pixel horizontal coordinate of the steel foot area mask map and minimum value , the maximum value of the pixel vertical coordinate of the steel foot area mask map and minimum value ;but , , , , where F is the pixel size of the image; Mask image of non-steel cap area through the maximum cross section Calculate the bounding box of the non-steel cap area: Traverse the non-steel cap area mask map of the maximum cross section For each pixel, record the pixel coordinates of each pixel whose brightness value is not 0 to form a pixel coordinate set: ,in is the pixel coordinate of the mask image of the non-steel cap area, is the pixel horizontal coordinate of the mask image of the non-steel cap area, is the pixel ordinate of the mask image of the non-steel cap area, is the number of pixels in the horizontal direction, is the number of pixels in the vertical direction; traverse Set, find the maximum value of the pixel horizontal coordinate of the mask map of the non-steel cap area and minimum value , the maximum value of the vertical coordinate of the pixel in the non-steel cap area mask and minimum value ,but , , , , where F is the pixel size of the image.

6. The intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection according to claim 5 is characterized in that: Step S2-3-2 specifically includes: Longitudinal section 1: The distance between the widest position of the steel foot and the top of the photo in the largest longitudinal section The calculation process is as follows: Count the steel foot area mask map of the largest longitudinal section of a group of longitudinal sections by row The sum of the brightness of the steel foot area mask image is obtained by ,in Mask map of the steel foot area The line number of any line of Mask map of the steel foot area No. The sum of the brightness of all pixels in the row, Mask map of the steel foot area Pixel height; find the set The row number with the highest brightness value in is , ; The distance between the widest position of the steel foot and the top of the photo in the largest longitudinal section of the 2 groups of longitudinal sections The calculation process is as follows: Count the steel foot area mask map of the two groups of maximum longitudinal sections by row The sum of the brightness of the steel foot area mask image is obtained by ;in Mask map of the steel foot area The line number of any line of Mask map of the steel foot area No. The sum of the brightness of all pixels in the row, Mask map of the steel foot area Pixel height; find the set The row number with the highest brightness value in is ,but ; The horizontal distances between the steel foot and the left and right sides of the steel cap at the widest position of the steel foot in the largest longitudinal section of the longitudinal section 1 are , The calculation process is as follows: Read the steel foot area mask map of the largest longitudinal section of the longitudinal section 1 group , find the Y-axis coordinate value of the widest position of the steel foot in the pixel coordinate system ; Traverse No. For all pixels in the row, find the pixel coordinates of the first pixel whose brightness value is not 0 ; Find the pixel coordinates of the last pixel whose brightness value is not 0 ; Traverse the non-steel cap area mask map of the largest longitudinal section of the longitudinal section 1 group No. For all pixels in the row, find the pixel coordinates of the first pixel whose brightness value is not 0 ; Find the pixel coordinates of the last pixel whose brightness value is not 0 ;but ; ; Horizontal distance between the widest position of the steel foot and the left and right sides of the steel cap in the largest longitudinal section of the 2nd group of longitudinal sections , The calculation process is as follows: Read the steel foot area mask map of the two largest longitudinal sections of the longitudinal section , find the Y-axis coordinate value of the widest position of the steel foot in the pixel coordinate system ; Traverse No. For all pixels in the row, find the pixel coordinates of the first pixel whose brightness value is not 0 and the pixel coordinates of the last pixel whose brightness value is not 0 ; Traverse the non-steel cap area mask map of the two largest longitudinal sections of the longitudinal section No. Line, find the pixel coordinates of the first pixel whose brightness value is not 0 ; Find the pixel coordinates of the last pixel whose brightness value is not 0 ,but: ; .

7. The intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection according to claim 6 is characterized in that: The calculation of the cross-section porcelain layer includes: Step S3-2-1: Calculation of the distance between the porcelain layer and the steel foot at the position of the horizontal projection line of any cross section on the two largest longitudinal sections; Step S3-2-2: Calculate the distance between the ceramic layer and the steel foot on the X-axis and Y-axis in any cross section; Step S3-2-3: Draw the ceramic layer area in any cross section.

8. The intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection according to claim 7 is characterized in that: The longitudinal section porcelain layer calculation includes: Step S3-4-1: Calculate the distance between the inner and outer boundaries of the porcelain layer at any longitudinal section at any height and the central axis of the photograph; Step S3-4-2: Calculation of the ceramic layer area of ​​any longitudinal section.

9. The intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection according to claim 1 is characterized in that: The specific process of step 2 is as follows: Step A1: For any industrial CT inspection photo, use the yolov11 instance segmentation algorithm to perform defect detection; suppose each defect mask in the industrial CT inspection photo is The mask image of the ceramic layer area of ​​the industrial CT inspection photo is as follows: ; Step A2: Traverse , find the pixels whose brightness value is not 0, and set the pixel coordinates Record in a collection , is the horizontal coordinate of the i-th pixel of the s-th defect mask image, is the horizontal coordinate of the i-th pixel of the s-th defect mask image; Step A3: Traverse the collection For each pixel coordinate in , for each pixel coordinate, read the brightness value of the pixel at the same position in the porcelain layer area mask map; if the pixel brightness value at the corresponding position in the porcelain layer area mask map is not 0, the defect is a porcelain layer crack, otherwise it is a cement layer pore.

10. A system for implementing the intelligent diagnosis method for micro-defects of porcelain insulators based on industrial CT detection according to any one of claims 1 to 9, characterized in that: include: Porcelain insulator structure analysis module: collect industrial CT inspection photos of porcelain insulators, and extract cement layers and porcelain layers in cross sections and longitudinal sections; Porcelain insulator defect detection module: The defect area contained in the industrial CT detection photo is identified through the yolov11 instance segmentation algorithm, and the defects are classified into cement layer pore defects and porcelain layer crack defects based on the cement layer and porcelain layer.

Images