GIS defect detection method and device

Abstract

The invention discloses a GIS defect detection method and device, and the method comprises the steps: carrying out the graying processing of an original image IOriginal of the X-ray detection of GIS equipment, obtaining a gray image Igrqy, carrying out the Gaussian blur processing of the gray image Igrqy, removing the noise of the gray image Igrqy, obtaining a blurred image Iblue, calculating the levelness and verticality of the blurred image Iblue through a Sobel operator, and carrying out the detection of the defects of the GIS equipment. And finally, carrying out pixel value adjustment processing on the blurred image Iblue to obtain a target image Itarget. The method comprises the following steps: carrying out edge extraction on a target image Itarget by adopting a Canny edge detection algorithm, carrying out paired pixel center line fitting by adopting a least square method, and determining a current detection defect type by judging a relationship among a curvature k (x), a curvature change rate ki and a bifurcation number P. According to the method, analysis is carried out according to the existing possible defect types of the GIS equipment, the method of combining the curvature of the center line, the curvature change and the bifurcation is pointedly put forward to recognize the defect types, and compared with manual look-up recognition and judgment one by one image, the efficiency is greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of GIS equipment detection technology, in particular to the field of image post-processing technology of GIS equipment based on X-ray detection, specifically to the field of detection method technology based on X-ray detection image analysis to determine whether equipment defects exist and the defect types, specifically to a GIS defect detection method and device. Background Art

[0002] In recent years, X-ray inspection-based defect inspection technology for gas-insulated electrical equipment (GIS) has been widely used. GIS is widely used in high-voltage power systems due to its excellent insulation performance and compact design. However, due to the complex structure and working environment of the equipment, various defects may occur in GIS during long-term operation, such as gas leakage, aging of insulation materials, and mechanical damage. As a non-destructive testing (NDT) technology, X-ray inspection can effectively see through the inside of the equipment and provide high-resolution images, thereby realizing early identification of potential defects. At present, many power companies and inspection agencies have integrated X-ray inspection into routine maintenance processes to improve the reliability and safety of equipment through regular inspections. In addition, with the development of image processing technology, automated defect recognition systems have gradually emerged, improving detection efficiency and accuracy. However, despite the continuous advancement of technology, X-ray inspection still faces some challenges in practical applications, such as radiation safety, detection equipment costs, and professional training of operators, which have affected its widespread promotion and application.

[0003] Analyzing the defect types in X-ray inspection images is of great significance to ensure the safety and reliability of GIS equipment. By identifying and classifying different types of defects, such as bubbles, cracks, corrosion, and material inhomogeneity, operation and maintenance personnel can formulate targeted maintenance strategies to reduce equipment failure rates and extend equipment service life. In addition, the analysis of defect types can also provide a reference for equipment design improvements and material selection. However, existing technologies still face many challenges in defect analysis. First, the influence of image noise and artifacts may lead to a decrease in the accuracy of defect identification, which brings difficulties to subsequent analysis. Secondly, current automated defect identification systems often rely on traditional image processing algorithms, which are difficult to cope with complex defect types and diverse equipment structures. In addition, the lack of sufficient annotated data and effective machine learning models also limits the scope of application of intelligent analysis technology. Therefore, improving the accuracy and robustness of image processing and analysis technology is a key direction for future research.

[0004] Close prior art: Patent document, announcement number CN117740819B, subject name is a GIS tank defect detection system and detection method, which provides a specific system for GIS tank defect detection and a method for judging the severity of defects. This method can only detect GIS tanks, and at the same time, it can only perform qualitative detection on defects, and cannot further judge the defect type. In order to quickly and accurately analyze defects on X-ray detection images of GIS equipment, the present invention is specially completed. Summary of the invention

[0005] In order to solve the problems of defect detection, identification and classification in the post-processing of X-ray detection images of GIS equipment, the present application provides a GIS defect detection method and device for automatically identifying and analyzing X-ray detection images and qualitatively identifying the defect types, which can save a lot of time for manual viewing and manual screening, and greatly improve the efficiency of post-detection processing.

[0006] The present invention can directly input the original X-ray image obtained by detection without image preprocessing, and directly perform data judgment through the detection method provided by the present invention. The defect judgment efficiency is high and can meet the needs of large-scale image processing.

[0007] The present invention adopts unique curvature and bifurcation discrimination, and has extremely high classification accuracy for the defect types that often occur in GIS equipment defects, with extremely low probability of misjudgment and missed judgment.

[0008] In order to achieve the above purpose, the technical solution adopted in this application is:

[0009] The present invention provides a GIS defect detection method, which includes the following detection steps:

[0010] Step STP100, the original image I of the X-ray detection of the GIS equipment Original Grayscale processing is performed to obtain a grayscale image I grqy , the grayscale image I grqy Use Gaussian blur to remove the grayscale image I grqy The noise is used to obtain the blurred image I blur , and then use the Sobel operator to calculate the blurred image I blur The horizontal and vertical degrees of the image will be blurred. blur Perform pixel value adjustment to obtain target image I target ;

[0011] Step STP200, using the Canny edge detection algorithm to detect the target image I target Perform edge extraction to obtain multiple edge objects B1, B2…B n , where n≥1;

[0012] Step STP300: each edge object B1, B2, ... B obtained in step STP200 is n The least square method is used to fit the center lines of paired pixels to obtain the edge objects B1, B2…B n The corresponding multiple center lines y(x)1,y(x)2…y(x) n , where x, y are the target image I target The pixel coordinates of

[0013] Step STP400, each center line y(x)1, y(x)2…y(x) obtained in step STP300 is n Calculate the curvature k(x) and the curvature change rate;

[0014] Step STP500, by judging the curvature k(x) and setting the bifurcation threshold T k Determine the current center line y(x) n The number of bifurcations P is determined by judging the curvature k(x) and the curvature change rate k i And the relationship between the bifurcation number P determines the current detection defect type:

[0015] The defect is hollowing: the curvature k(x) < the preset value K, and the actual number of bifurcations P = 0; K is the preset curvature value;

[0016] Defect is impurity: curvature change rate △k i <ξ1 and i∈n;

[0017] The defect is a fatigue crack: the curvature change rate △k i >ξ1, And P∈[1,3], i∈n; where ξ1 is the threshold of curvature change rate, and ξ2 is the threshold of curvature cumulative change;

[0018] The defect is a fracture: the curvature k(x)>K is satisfied, and the actual number of bifurcations P≥3.

[0019] In order to improve the accuracy of subsequent edge recognition of defective parts, preferably, in step STP100, the Sobel operator is used to calculate the blurred image I blur The horizontal and vertical method is to use the blurred image I after Gaussian filtering blur The two convolution kernels of the Sobel operator are used to calculate the gradients in the horizontal and vertical directions respectively. x It is expressed as:

[0020]

[0021] Vertical Sobel operator Gy It is expressed as:

[0022]

[0023] Combine each convolution kernel with the input blurred image I blur Perform convolution and calculate the gradients in the horizontal and vertical directions as follows:

[0024] Horizontal gradient G x (i, j) is expressed as: Vertical gradient G y (i, j) is expressed as: Where I(i,j) is the input blurred image I blur The pixel value at (i, j); also includes the gradient magnitude and direction calculated by horizontal and vertical gradients, as follows:

[0025] The gradient magnitude G is expressed as G(i, j)=G x (i, j) 2 +G y (i, j) 2 ;

[0026] The gradient direction θ is expressed as

[0027] In order to further improve the accuracy of defect recognition and extraction, preferably, in step STP100, the blurred image I blur Perform pixel value adjustment to obtain target image I target The process includes the steps of adjusting the gradient amplitude of the pixel value and pixel classification, specifically including:

[0028] Adjustment of gradient amplitude: Convert the gradient direction θ(i, j) of any pixel into discrete direction angles of 0° / 45° / 90° / 135°, and then compare the gradient amplitude G(i, j) of adjacent pixels according to the direction of each pixel. If G(i, j) is greater than the value of the adjacent pixel, retain the gradient amplitude of the current pixel, otherwise, set it to 0, and traverse the entire blurred image I blur Pixels;

[0029] Pixel classification: Set the high gradient threshold T1 and the low gradient threshold T2, and determine the edge type based on the relationship between the current pixel gradient amplitude G(i, j) and the high gradient threshold T1 and the low gradient threshold T2:

[0030] If G(i, j)>T1, G(i, j) is classified as a strong edge;

[0031] If T1>G(i, j)>T2, G(i, j) is classified as a weak edge;

[0032] If G(i, j)<T2, G(i, j) is classified as a non-edge image pixel.

[0033] The present invention also provides a detection device, including a processor and a memory, wherein the memory stores a program for executing the GIS defect detection method described above, and the processor is used to call the program in the memory to execute the GIS defect detection method.

[0034] Beneficial effects:

[0035] 1. The present invention analyzes the existing possible defect types of GIS equipment and specifically proposes a method that combines centerline curvature, curvature change and bifurcation to identify the defect type. Compared with manual image review, identification and judgment, the efficiency is greatly improved, and it can be competent for the defect identification task of massive detection images, providing a good solution for the image post-processing task of X-ray detection of GIS equipment.

[0036] 2. Before identifying defects in the detection image, the present invention will perform pixel-level processing on the portrait, especially quantitative identification and processing of pixel gradients and directions, which can greatly improve the efficiency and accuracy of defect processing. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0038] Embodiment 1:

[0039] This embodiment provides a GIS defect detection method, including the following detection steps:

[0040] Step STP100, the original image I of the X-ray detection of the GIS equipment Original Grayscale processing is performed to obtain a grayscale image I grqy , the grayscale image I grqy Use Gaussian blur to remove the grayscale image I grqy The noise is used to obtain the blurred image I blur , and then use the Sobel operator to calculate the blurred image I blur The horizontal and vertical degrees of the image will be blurred. blur Perform pixel value adjustment to obtain target image I target ;

[0041] Step STP200, using the Canny edge detection algorithm to detect the target image I target Perform edge extraction to obtain multiple edge objects B1, B2…Bn , where n≥1;

[0042] Step STP300: each edge object B1, B2, ... B obtained in step STP200 is n The least square method is used to fit the center lines of paired pixels to obtain the edge objects B1, B2…B n The corresponding multiple center lines y(x)1,y(x)2…y(x) n , where x, y are the target image I target The pixel coordinates of

[0043] Step STP400, each center line y(x)1, y(x)2…y(x) obtained in step STP300 is n Calculate the curvature k(x) and the curvature change rate;

[0044] Step STP500, by judging the curvature k(x) and setting the bifurcation threshold T k Determine the current center line y(x) n The number of bifurcations P is determined by judging the curvature k(x) and the curvature change rate k i And the relationship between the bifurcation number P determines the current detection defect type:

[0045] The defect is hollowing: the curvature k(x) < the preset value K, and the actual number of bifurcations P = 0; K is the preset curvature value;

[0046] Defect is impurity: curvature change rate △k i <ξ1 and i∈n;

[0047] The defect is a fatigue crack: the curvature change rate △k i >ξ1, And P∈[1,3], i∈n; where ξ1 is the threshold of curvature change rate, and ξ2 is the threshold of curvature cumulative change;

[0048] The defect is a fracture: the curvature k(x)>K is satisfied, and the actual number of bifurcations P≥3.

[0049] In this embodiment, step STP300-step STP500 is a particularly important part of the present invention, which has the advantages of high efficiency and high accuracy compared to the existing manual image viewing or computer recognition. Specifically, in this embodiment, the curvature, curvature change rate / amount, whether it is bifurcated, and the number of bifurcations are combined to make a comprehensive judgment. Compared with the traditional defect recognition method, which mainly relies on threshold-based segmentation technology and template matching, these methods are often sensitive to image noise and lighting changes, resulting in low recognition accuracy. This embodiment uses edge extraction technology, especially edge detection based on Canny or Sobel operators, which can effectively capture the detailed information of the defect, especially in a complex background. By accurately identifying the edges of the defect, the subsequent centerline fitting can more accurately reflect the shape characteristics of the defect, thereby improving the overall recognition accuracy. Furthermore, the existing defect recognition methods usually focus on simple features, such as the area or length of the defect, which limits the in-depth analysis of the nature of the defect. By fitting the centerline of the defect and analyzing its curvature, curvature change and bifurcation, richer geometric features can be extracted. These characteristics can not only help identify the type of defects, but also reflect the generation mechanism and development trend of defects, providing an important basis for subsequent fault prediction and equipment maintenance. For example, fatigue cracks will extend with the use time of the equipment and its working conditions. Based on the existing defect conditions and types, it is possible to scientifically infer the time when the failure may occur. In this way, it is possible to predict and prejudge in advance, and do a good job of equipment maintenance or replacement before the failure occurs, so as to avoid power outages caused by sudden equipment failures, thereby saving unnecessary power outage losses and avoiding power outage accidents.

[0050] In the prior art, although there are some software for computer classification, these recognition software mostly rely on machine learning models, usually require a large amount of labeled data for training, and the classification effect is limited by the rationality of feature selection. On the other hand, many model trainings are only theoretical. In practical applications, especially for the detection image training of GIS equipment, it is almost impossible to accumulate image samples that can train a mature model; therefore, limited by the number of samples, even if the training mode, mechanism and parameters have been adjusted to the optimal, it is almost impossible to obtain a mature model that can independently judge whether GIS equipment defects exist and what type of defects they belong to. However, this embodiment can construct a more sophisticated classification standard by analyzing the curvature and number of bifurcations of the defect centerline. For example, different types of cracks, hollows or impurities show obvious differences in curvature characteristics, which can be effectively classified by setting thresholds. This method not only reduces the dependence on training data, but also can adapt to new defect types more quickly in practical applications. In practical applications, there may be multiple defects in GIS equipment at the same time, and the morphology and distribution of these defects may affect each other. Traditional methods are often unable to cope with such complex defects, which easily leads to recognition errors or omissions. By combining edge extraction and centerline fitting, multiple defects can be effectively distinguished and analyzed. During the analysis process, the relationship between different defects (such as the number and location of bifurcation points) can be detected to determine their mutual influence, thereby providing a more comprehensive fault analysis. Currently, many defect detection methods rely on manual experience for defect identification and classification, which are easily affected by subjective factors. Through the automated processing of edge extraction and centerline fitting, the need for manual intervention can be greatly reduced, and the detection efficiency and consistency can be improved. This not only improves the reliability of detection, but also reduces the consumption of human resources during large-scale detection and improves overall work efficiency.

[0051] Finally, the method using edge extraction and centerline fitting has good scalability and can adapt to different types of GIS equipment and defect characteristics. In the prior art, many methods have a strong dependence on specific equipment or defect types, while the detection method provided in this embodiment can be easily migrated to other fields or other types of equipment by extracting geometric features, and has strong compatibility and application potential. This makes this method have broad application prospects in future equipment maintenance and detection. Based on the implementation of the present invention, a foundation is laid for establishing a map library of the GIS equipment status of the entire industry. Since the present invention can quickly and autonomously classify the defects of the detection image, it is possible to perform cluster analysis on massive data, and the detection images of all GIS equipment in the country can be imported into the map library. The current status of the in-service GIS equipment in the entire industry can be previewed in real time, which has an unspeakable huge role and practical significance in maintaining the entire power industry.

[0052] Embodiment 2:

[0053] This embodiment refines the inspection image before defect recognition based on the embodiment 1, aiming to further improve the accuracy and efficiency of defect recognition, analysis and classification. Specifically, in this embodiment, the Sobel operator is used to calculate the fuzzy image I in step STP100. blur The horizontal and vertical method is to use the blurred image I after Gaussian filtering blur The two convolution kernels of the Sobel operator are used to calculate the gradients in the horizontal and vertical directions respectively. x It is expressed as:

[0054]

[0055] Vertical Sobel operator G y It is expressed as:

[0056]

[0057] Combine each convolution kernel with the input blurred image I blur Perform convolution and calculate the gradients in the horizontal and vertical directions as follows:

[0058] Horizontal gradient G x (i, j) is expressed as: Vertical gradient G y (i, j) is expressed as: Where I(i,j) is the input blurred image I blur The pixel value at (i, j); also includes the gradient magnitude and direction calculated by horizontal and vertical gradients, as follows:

[0059] The gradient magnitude G is expressed as G(i, j)=G x (i, j) 2 +G y (i, j) 2 ;

[0060] The gradient direction θ is expressed as

[0061] At the same time, in step STP100, the blurred image I blur Perform pixel value adjustment to obtain target image I targetThe process includes the adjustment of the gradient amplitude of the pixel value and the pixel classification step, specifically including: Gradient amplitude adjustment: convert the gradient direction θ(i, j) of any pixel into a discrete direction angle of 0° / 45° / 90° / 135°, and then compare the gradient amplitude G(i, j) of the adjacent pixels according to the direction of each pixel. If G(i, j) is greater than the value of the adjacent pixel, the gradient amplitude of the current pixel is retained, otherwise, it is set to 0, and the entire blurred image I is traversed. blur Pixels;

[0062] Pixel classification: Set the high gradient threshold T1 and the low gradient threshold T2, and determine the edge type based on the relationship between the current pixel gradient amplitude G(i, j) and the high gradient threshold T1 and the low gradient threshold T2:

[0063] If G(i, j)>T1, G(i, j) is classified as a strong edge;

[0064] If T1>G(i, j)>T2, G(i, j) is classified as a weak edge;

[0065] If G(i, j)<T2, G(i, j) is classified as a non-edge image pixel.

[0066] Of course, the method provided in this embodiment is only one of the preferred methods and is not exclusive. Those skilled in the art can still extend or combine the method within the framework of the inventive concept disclosed in the present invention.

[0067] Embodiment 3:

[0068] This embodiment provides a detection device, including a processor and a memory, wherein the memory stores a program for executing the GIS defect detection method described in any of the above embodiments, and the processor is used to call the program in the memory to execute the GIS defect detection method.

[0069] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A GIS defect detection method, characterized in that: The following detection steps are included: Step STP100, the original image I of the X-ray detection of the GIS equipment Original Grayscale processing is performed to obtain a grayscale image I grqy , the grayscale image I grqy Use Gaussian blur to remove the grayscale image I grqy The noise is used to obtain the blurred image I blur , and then use the Sobel operator to calculate the blurred image I blur The horizontal and vertical degrees of the image will be blurred. blur Perform pixel value adjustment to obtain target image I target ; Step STP200, using the Canny edge detection algorithm to detect the target image I target Perform edge extraction to obtain multiple edge objects B1, B2…B n , where n≥1; Step STP300: each edge object B1, B2, ... B obtained in step STP200 is n The least square method is used to fit the center lines of paired pixels to obtain the edge objects B1, B2…B n The corresponding multiple center lines y(x)1,y(x)2…y(x) n , where x, y are the target image I target The pixel coordinates of Step STP400, each center line y(x)1, y(x)2…y(x) obtained in step STP300 is n Calculate the curvature k(x) and the curvature change rate; Step STP500, by judging the curvature k(x) and setting the bifurcation threshold T k Determine the current center line y(x) n The number of bifurcations P is determined by judging the curvature k(x) and the curvature change rate k i And the relationship between the bifurcation number P determines the current detection defect type: The defect is hollowing: the curvature k(x) < the preset value K, and the actual number of bifurcations P = 0; K is the preset curvature value; Defect is an impurity: Curvature change rate △k i <ξ1 and ∑1 n k i <ξ2, i ∈ n; The defect is a fatigue crack: the curvature change rate △k i >ξ1, And P∈[1,3], i∈n; where ξ1 is the threshold of curvature change rate, and ξ2 is the threshold of curvature cumulative change; The defect is a fracture: the curvature k(x)>K is satisfied, and the actual number of bifurcations P≥3.

2. The GIS defect detection method according to claim 1, characterized in that: In step STP100, the Sobel operator is used to calculate the blurred image I blur The horizontal and vertical method is to use the blurred image I after Gaussian filtering blur The two convolution kernels of the Sobel operator are used to calculate the gradients in the horizontal and vertical directions respectively. The specific horizontal Sobel operator G x It is expressed as: Vertical Sobel operator G y It is expressed as: Combine each convolution kernel with the input blurred image I blur Perform convolution and calculate the gradients in the horizontal and vertical directions as follows: Horizontal gradient G x (i, j) is expressed as: Vertical gradient G y (i, j) is expressed as: Where I(i,j) is the input blurred image I blur The pixel value at (i, j); also includes the gradient magnitude and direction calculated by horizontal and vertical gradients, as follows: The gradient magnitude G is expressed as G(i, j)=G x (i, j) 2 +G y (i, j) 2 ; The gradient direction θ is expressed as 3. The GIS defect detection method according to claim 2, characterized in that: In step STP100, the blurred image I blur Perform pixel value adjustment to obtain target image I target The process includes the steps of adjusting the gradient amplitude of the pixel value and pixel classification, specifically including: Adjustment of gradient amplitude: Convert the gradient direction θ(i, j) of any pixel into discrete direction angles of 0° / 45° / 90° / 135°, and then compare the gradient amplitude G(i, j) of adjacent pixels according to the direction of each pixel. If G(i, j) is greater than the value of the adjacent pixel, retain the gradient amplitude of the current pixel, otherwise, set it to 0, and traverse the entire blurred image I blur Pixels; Pixel classification: Set the high gradient threshold T1 and the low gradient threshold T2, and determine the edge type based on the relationship between the current pixel gradient amplitude G(i, j) and the high gradient threshold T1 and the low gradient threshold T2: If G(i, j)>T1, G(i, j) is classified as a strong edge; If T1>G(i, j)>T2, G(i, j) is classified as a weak edge; If G(i, j)<T2, G(i, j) is classified as a non-edge image pixel.

4. A detection device, comprising a processor and a memory, characterized in that: The memory stores a program for executing the GIS defect detection method according to any one of claims 1 to 3, and the processor is used to call the program in the memory to execute the GIS defect detection method.

Images