Grading ring deformation identification method in electric power intelligent inspection

Abstract

The invention relates to a grading ring deformation identification method in electric power intelligent inspection. The method comprises the following steps: detecting and segmenting an original image acquired by inspection equipment; smoothing the inner and outer contours of each grading ring mask, and removing noise points; carrying out skeleton extraction on the mask subjected to noise removal; performing ellipse fitting on the grading ring framework; calculating the shortest distance from each pixel on the grading ring skeleton to the ellipse, and normalizing the distance; and counting the number N of shortest distances greater than a distance threshold T1 in the shortest distance set of the grading ring skeleton and the ellipse, and if N exceeds a number threshold T2, considering that the grading ring is deformed. According to the method, whether the grading ring deformation exists in the intelligent inspection image of the transformer substation is automatically analyzed, so that the efficiency and accuracy of inspection work are greatly improved, the deformation position of the grading ring is automatically identified, an engineer can be helped to quickly position a problem, and the troubleshooting time and cost are reduced; the inspection personnel can visually and clearly see the deformation part and degree in a thermodynamic diagram mode.

Description

Technical Field

[0001] The invention relates to the technical field of electric power intelligent inspection, and in particular to a method for identifying deformation of a grading ring in electric power intelligent inspection. Background Art

[0002] The grading ring is a key component in high-voltage electrical equipment. The accuracy of its shape and position is directly related to the electric field distribution and voltage balance of the equipment. If the grading ring is deformed, it will lead to uneven electric field distribution, increase local voltage stress, and may cause insulation breakdown or other failures of the equipment. Timely identification of grading ring deformation can prevent equipment failure and ensure the safe operation of the power system.

[0003] At present, the traditional methods of identifying the deformation of the grading ring include manual inspection and intelligent inspection. Among them, the manual inspection method is not only time-consuming and labor-intensive, but also relies on the experience of the inspectors, and is prone to missed detection or misjudgment. The intelligent inspection method uses digital image processing, machine learning and other technologies to automatically, quickly and accurately identify the deformation of the grading ring, greatly improving the inspection efficiency and accuracy, and reducing the workload of the inspectors. At present, the intelligent inspection of power transmission and transformation mainly adopts the deep learning target detection method. However, the use of convolutional neural networks to detect the rectangular frame of the defective grading ring cannot provide the judgment basis and key intermediate information. There is also a method of using ellipse fitting to calculate the intersection of the ellipse and the grading ring to determine whether the grading ring is missing. However, considering the different thicknesses of the grading rings, the foreground of the grading ring will still intersect with the fitted ellipse, resulting in only judging the missing but not the deformation, and the degree of deformation cannot be quantified. Summary of the invention

[0004] In order to solve the problems of missed detection or misjudgment of deformed equalizing rings, and inability to visualize the position and degree of deformation of the equalizing rings, the purpose of the present invention is to provide a method for identifying the deformation of equalizing rings in intelligent power inspections, which can improve the efficiency and accuracy of inspections, reduce the work pressure of inspectors, and automatically identify the position of the deformed equalizing rings and visualize them.

[0005] To achieve the above object, the present invention adopts the following technical solution: a method for identifying deformation of a grading ring in electric power intelligent inspection, comprising the following steps in order: (1) Use the instance segmentation network to detect and segment the original image collected by the inspection equipment to obtain the mask of each balancing ring; (2) Smoothing the inner and outer contours of each balancing ring mask, removing noise points, and obtaining a denoised mask; (3) Use the refinement algorithm to extract the skeleton of the de-dried mask and obtain the pressure-equalizing ring skeleton; (4) Perform ellipse fitting on the skeleton of the pressure equalizing ring to obtain the major axis, minor axis, center point coordinates and rotation angle of the ellipse in the image coordinate system; (5) Calculate the shortest distance from each pixel on the pressure equalizing ring skeleton to the ellipse and normalize the distance to obtain the set of shortest distances between the pressure equalizing ring skeleton and the ellipse; (6) Count the number N of the shortest distances between the pressure equalizing ring skeleton and the ellipse that are greater than the distance threshold T1. If N exceeds the number threshold T2, it is considered that the pressure equalizing ring is deformed.

[0006] In step (1), the instance segmentation network adopts any one of the Mask-RCNN network, the HTC network, and the YOLOv8-seg network.

[0007] The step (2) specifically refers to: using a Gaussian filter to smooth the contour curves of the inner and outer contours of each balancing ring mask to remove noise and burrs, and then binarizing the blurred result through a binarization algorithm; the window size of the Gaussian filter is set to 3, σ is set to 2; the threshold of the binarization algorithm is set to 1, the foreground is set to 255, and the background is set to 0.

[0008] The step (3) specifically refers to: using a thinning algorithm to extract the skeleton of the denoised mask to obtain a pressure equalizing ring skeleton with a width of 1 pixel.

[0009] The step (4) specifically refers to: using an ellipse fitting algorithm based on the least squares method, fitting the coordinates of the ellipse center, the lengths of the major axis and the minor axis, and the rotation angle according to the skeleton of the pressure equalizing ring, and drawing an ellipse with a foreground of 255 on an image with a background of 0.

[0010] The step (5) specifically refers to: traversing each pixel on the equalizing ring skeleton, using Euclidean distance to calculate the shortest distance from the pixel to the elliptical foreground, and normalizing the distance value, the numerator of the distance value is the Euclidean distance, the denominator of the distance value is the size of the original image, and the normalized maximum and minimum interval is 0 to 255.

[0011] The step (6) specifically refers to: counting the number N of the shortest distances between the equalizing ring skeleton and the ellipse that is greater than the distance threshold T1, and judging whether the equalizing ring is deformed based on whether the number N is greater than T2, the distance threshold T1 is selected as 20, and the number threshold T2 is selected as 10% of the sampling points of the equalizing ring skeleton.

[0012] It can be seen from the above technical scheme that the beneficial effects of the present invention are: First, the present invention greatly improves the efficiency and accuracy of the inspection work by automatically analyzing whether there is deformation of the equalizing ring in the intelligent inspection image of the substation, reduces the work pressure of the inspectors, and effectively reduces the omissions caused by human factors; Second, the present invention can automatically identify the position where the equalizing ring is deformed, help engineering personnel to quickly locate the problem, and reduce the time and cost of troubleshooting; Third, the present invention can quantify the degree of deformation and identify subtle deformations, so as to timely discover potential safety hazards of the equipment, avoid major losses caused by equipment failures, and provide a scientific basis for equipment maintenance; Fourth, the present invention provides a visualization method for the deformation of the equalizing ring, which allows inspectors to intuitively and clearly see the location and degree of deformation through a thermal map, helping decision makers to make faster and more accurate maintenance decisions, and effectively ensure the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a normal pressure equalizing ring image collected by the patrol system; Figure 2 is the deformed pressure equalizing ring image collected by the patrol system; Figure 3 This is a visualization of the normal distance of the equalizing ring; Figure 4 is the distance visualization diagram of the deformation of the equalizing ring; Figure 5 It is the normal Euclidean distance line graph of the equalizing ring; Figure 6 It is the Euclidean distance line graph of the deformation of the equalizing ring. DETAILED DESCRIPTION

[0014] A method for identifying deformation of a grading ring in electric power intelligent inspection comprises the following steps in order: (1) Use the instance segmentation network to detect and segment the original image collected by the inspection equipment to obtain the mask of each balancing ring; Figure 1 As shown, the original image is a normal, non-deformed grading ring image taken in the substation patrol system; (2) Smoothing the inner and outer contours of each balancing ring mask, removing noise points, and obtaining a denoised mask; (3) Use the refinement algorithm to extract the skeleton of the de-dried mask and obtain the pressure-equalizing ring skeleton; (4) Perform ellipse fitting on the skeleton of the pressure equalizing ring to obtain the major axis, minor axis, center point coordinates and rotation angle of the ellipse in the image coordinate system; (5) Calculate the shortest distance from each pixel on the pressure equalizing ring skeleton to the ellipse and normalize the distance to obtain the set of shortest distances between the pressure equalizing ring skeleton and the ellipse; (6) Count the number N of the shortest distances between the pressure equalizing ring skeleton and the ellipse that are greater than the distance threshold T1. If N exceeds the number threshold T2, it is considered that the pressure equalizing ring is deformed.

[0015] In step (1), the instance segmentation network adopts any one of the Mask-RCNN network, HTC network, and YOLOv8-seg network. Image instance segmentation is a task in computer vision, and its goal is to identify each independent object in the image and generate an independent pixel-level mask for each object. Unlike image classification and object detection, instance segmentation not only identifies the category of the object, but also distinguishes different instances of the same category. The current method is mainly based on the deep learning method, which collects and annotates data from the inspection site and realizes instance segmentation by training the network. The instance segmentation methods mainly include Mask-RCNN, HTC, YOLOv8-seg, etc. The present invention uses YOLOv8-seg to perform instance segmentation on the equalizing ring.

[0016] The step (2) specifically refers to: using a Gaussian filter to smooth the contour curves of the inner and outer contours of each balancing ring mask to remove noise and burrs, and then binarizing the blurred result through a binarization algorithm; the window size of the Gaussian filter is set to 3, σ is set to 2; the threshold of the binarization algorithm is set to 1, the foreground is set to 255, and the background is set to 0.

[0017] The step (3) specifically refers to: using the Zhang-Suen thinning algorithm to extract the skeleton of the denoised mask to obtain a pressure-equalizing ring skeleton with a width of 1 pixel. The thinning algorithm is two alternating iterative processes, which gradually remove redundant foreground pixels and maintain the original shape after convergence.

[0018] The step (4) specifically refers to: using an ellipse fitting algorithm based on the least squares method, fitting the coordinates of the ellipse center, the lengths of the major axis and the minor axis, and the rotation angle according to the skeleton of the pressure equalizing ring, and drawing an ellipse with a foreground of 255 on an image with a background of 0.

[0019] The step (5) specifically refers to: traversing each pixel on the equalizing ring skeleton, using Euclidean distance to calculate the shortest distance from the pixel to the elliptical foreground, and normalizing the distance value, the numerator of the distance value is the Euclidean distance, the denominator of the distance value is the size of the original image, and the normalized maximum and minimum interval is 0 to 255.

[0020] The step (6) specifically refers to: counting the number N of the shortest distances between the equalizing ring skeleton and the ellipse that is greater than the distance threshold T1, and judging whether the equalizing ring is deformed based on whether the number N is greater than T2, the distance threshold T1 is selected as 20, and the number threshold T2 is selected as 10% of the sampling points of the equalizing ring skeleton.

[0021] like Figure 2 As shown, the lower left part of the pressure equalizing ring is deformed.

[0022] like Figure 3 As shown, the distance map is normalized according to the input image size. The larger the distance, the redder the pixel, and the smaller the distance, the bluer the image, and it is superimposed on the original image. Figure 3 The middle pressure equalizing ring is generally blue, which means that the pressure equalizing ring is similar to an ellipse and has not been deformed.

[0023] Figure 4 The lower left corner of the middle pressure equalizing ring appears red, indicating that the lower left corner of the pressure equalizing ring has been deformed. The redder the color, the greater the difference from the ellipse and the more severe the deformation.

[0024] like Figure 5 As shown, the horizontal axis is the pixel corresponding to each equalizing ring skeleton, the vertical axis is the distance from the pixel to the nearest ellipse, and the red dotted line is the set distance threshold T1. In this figure, no distance exceeds the threshold, so the equalizing ring is not deformed.

[0025] Figure 6 There are multiple distances exceeding T1, and the number exceeding T1 is greater than T2, and T2 is 10% of the value on the x-axis. Figure 6 In the figure, the value of the x-axis is about 2000, so it is judged that the pressure equalizing ring is deformed. On the contrary, if the number exceeding T1 is less than or equal to T2, it is judged that the pressure equalizing ring is not deformed.

[0026] To sum up, the present invention greatly improves the efficiency and accuracy of inspection work by automatically analyzing whether there is deformation of the equalizing ring in the intelligent inspection image of the substation, reduces the work pressure of the inspectors, and effectively reduces omissions caused by human factors; the position where the equalizing ring is deformed can be automatically identified to help engineering personnel quickly locate the problem and reduce troubleshooting time and cost; the present invention can quantify the degree of deformation and identify subtle deformations, so as to timely discover potential safety hazards of equipment, avoid major losses caused by equipment failure, and provide a scientific basis for equipment maintenance; the present invention provides a visualization method for the deformation of the equalizing ring, which allows inspectors to intuitively and clearly see the location and degree of deformation through a thermal map, helping decision makers make faster and more accurate maintenance decisions, and effectively ensuring the safe and stable operation of the power system.

Claims

1. A method for identifying deformation of a grading ring in electric power intelligent inspection, characterized in that: The method comprises the following steps in order: (1) Use the instance segmentation network to detect and segment the original image collected by the inspection equipment to obtain the mask of each balancing ring; (2) Smoothing the inner and outer contours of each balancing ring mask, removing noise points, and obtaining a denoised mask; (3) Use the refinement algorithm to extract the skeleton of the de-dried mask and obtain the pressure-equalizing ring skeleton; (4) Perform ellipse fitting on the skeleton of the pressure equalizing ring to obtain the major axis, minor axis, center point coordinates and rotation angle of the ellipse in the image coordinate system; (5) Calculate the shortest distance from each pixel on the pressure equalizing ring skeleton to the ellipse and normalize the distance to obtain the set of shortest distances between the pressure equalizing ring skeleton and the ellipse; (6) Count the number N of the shortest distances between the pressure equalizing ring skeleton and the ellipse that are greater than the distance threshold T1. If N exceeds the number threshold T2, it is considered that the pressure equalizing ring is deformed.

2. The method for identifying deformation of a grading ring in electric power intelligent inspection according to claim 1 is characterized in that: In step (1), the instance segmentation network adopts any one of the Mask-RCNN network, the HTC network, and the YOLOv8-seg network.

3. The method for identifying deformation of a grading ring in electric power intelligent inspection according to claim 1 is characterized in that: The step (2) specifically refers to: using a Gaussian filter to smooth the contour curves of the inner and outer contours of each balancing ring mask to remove noise and burrs, and then binarizing the blurred result through a binarization algorithm; the window size of the Gaussian filter is set to 3, σ is set to 2; the threshold of the binarization algorithm is set to 1, the foreground is set to 255, and the background is set to 0.

4. The method for identifying deformation of a grading ring in electric power intelligent inspection according to claim 1 is characterized in that: The step (3) specifically refers to: using a thinning algorithm to extract the skeleton of the denoised mask to obtain a pressure equalizing ring skeleton with a width of 1 pixel.

5. The method for identifying deformation of a grading ring in electric power intelligent inspection according to claim 1 is characterized in that: The step (4) specifically refers to: using an ellipse fitting algorithm based on the least squares method, fitting the coordinates of the ellipse center, the lengths of the major axis and the minor axis, and the rotation angle according to the skeleton of the pressure equalizing ring, and drawing an ellipse with a foreground of 255 on an image with a background of 0.

6. The method for identifying deformation of a grading ring in electric power intelligent inspection according to claim 1 is characterized in that: The step (5) specifically refers to: traversing each pixel on the equalizing ring skeleton, using Euclidean distance to calculate the shortest distance from the pixel to the elliptical foreground, and normalizing the distance value, the numerator of the distance value is the Euclidean distance, the denominator of the distance value is the size of the original image, and the normalized maximum and minimum interval is 0 to 255.

7. The method for identifying deformation of a grading ring in electric power intelligent inspection according to claim 1 is characterized in that: The step (6) specifically refers to: counting the number N of the shortest distances between the equalizing ring skeleton and the ellipse that is greater than the distance threshold T1, and judging whether the equalizing ring is deformed based on whether the number N is greater than T2, the distance threshold T1 is selected as 20, and the number threshold T2 is selected as 10% of the sampling points of the equalizing ring skeleton.

Images