Classification method and system for identifying GIS equipment part based on X-ray detection image

Abstract

The invention discloses a classification method and system for identifying GIS equipment parts based on X-ray detection images. The classification method comprises the following steps: respectively establishing contour sets M1, M2... M6 of a circuit breaker, an isolating switch, a grounding switch, a mutual inductor, a lightning arrester and a bus structure which form GIS equipment; the method comprises the following steps: inputting and identifying an original image IOriginal, and carrying out preprocessing including graying and denoising on the original image IOriginal to obtain a target image Itarget; a step of extracting an edge contour m0 in the target image Itarget based on a Canny edge detection algorithm, normalizing the contour, and comparing whether the contour m '0 normalized by the step STP500 belongs to contour sets M' 1, M '2... M' 6 by adopting similarity measurement; according to the method provided by the invention, the contour set M for each part (such as a circuit breaker, an isolating switch, a grounding switch, a mutual inductor, a lightning arrester and a bus) of the GIS equipment is established, so that the different parts can be accurately classified. Compared with a traditional image recognition technology, the classification method can effectively reduce the possibility of misrecognition and improve the classification accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment detection, in particular to the technical field of power detection image post-processing, and specifically to a classification method and system for identifying GIS equipment parts based on X-ray detection images. Background Art

[0002] Gas-insulated electrical equipment (GIS) is a new type of power equipment, mainly used in high-voltage power systems. Its structure is mainly composed of insulating gas (such as sulfur hexafluoride SF6), metal casing, switchgear, sensors and control systems. In GIS, the insulating gas not only plays the role of electrical insulation, but also effectively prevents arcs and electric shocks, improving the safety and reliability of the equipment. The compact design of GIS makes it occupy a small area and is suitable for use in space-constrained areas such as cities. Compared with traditional air-insulated equipment, GIS has better anti-pollution ability and lower maintenance requirements. Due to its excellent performance, GIS is widely used in key links such as substations, distribution networks and renewable energy access in the power industry, which improves the stability and economy of the power system and promotes the development and construction of smart grids. At the same time, the long life and high reliability of GIS also make it an indispensable and important equipment in modern power engineering.

[0003] Defect detection of gas-insulated electrical equipment (GIS) is an important part of ensuring its safe and stable operation. At present, GIS defect detection technology mainly includes two methods: online monitoring and offline detection. Online monitoring uses various sensors and monitoring systems to monitor parameters such as gas pressure, temperature, gas composition and partial discharge inside GIS in real time, and can timely detect potential fault hazards. However, due to the complex structure of GIS, some defects (such as aging of insulating materials, corrosion of metal parts, etc.) are often difficult to directly identify through conventional monitoring methods. Offline detection methods, such as infrared thermal imaging and ultrasonic detection, are mainly used for comprehensive inspection of equipment during regular maintenance. Although these detection methods can effectively detect equipment defects, they have certain limitations, such as long detection cycle and high labor cost. Therefore, GIS defect detection still needs to be further improved, combined with advanced artificial intelligence and big data analysis technologies to achieve more intelligent and automated detection methods and improve detection efficiency and accuracy.

[0004] Classification and management of inspection images of different parts of gas-insulated electrical equipment (GIS) has an important role and significance. First, by classifying the inspection images of various parts of GIS, systematic fault analysis can be achieved, helping engineers to quickly identify and locate problems and reduce fault handling time. For example, by classifying images of different parts such as electrical connectors, insulators and metal shells, different inspection and maintenance strategies can be applied in a targeted manner. Secondly, classification management helps to establish a health record of GIS equipment, making equipment maintenance and management more scientific and efficient. By accumulating and analyzing historical data, it can provide decision support for the operation and maintenance of equipment, predict potential faults, and improve the overall reliability of equipment. In addition, combined with advanced technologies such as machine learning, classification management can also realize automated fault detection and diagnosis, further improving the safety and operation efficiency of GIS. Therefore, classification management of inspection images of different parts of GIS not only improves the accuracy of detection, but also provides strong technical support for smart grids and sustainable development. However, in the existing technology, a complete archive system for GIS inspection status in the power industry has not yet been established. To solve this system problem, the first thing to solve is to classify the inspection images of different structures of GIS equipment for subsequent management. Summary of the invention

[0005] In order to solve the technical problem of identifying and classifying the structures of different parts of GIS equipment, the present application provides a classification method for identifying the parts of GIS equipment based on X-ray detection images, which is used to analyze the detection images containing different structural types of GIS equipment and classify them according to the structural type, thereby realizing automatic fault detection and diagnosis, and further improving the safety and operation efficiency of GIS.

[0006] In order to solve the problem of image classification according to different GIS equipment parts, the technical solution adopted by this application is:

[0007] The present invention provides a classification method for identifying GIS equipment parts based on X-ray detection images, comprising the following steps:

[0008] Step STP100, respectively establishing the contour sets M1, M2...M6 of the circuit breaker, disconnector, grounding switch, transformer, lightning arrester and busbar structure constituting the GIS equipment;

[0009] Step STP200, input and recognize the original image I Original , the original image I Original Perform preprocessing including grayscale and denoising to obtain the target image I target ;

[0010] Step STP300, extracting the target image I based on the Canny edge detection algorithmtarget The edge contour m0 in the figure, the pixels that make up the contour are represented by E(x, y)

[0011]

[0012] Among them, E(x, y) represents a function, represents the function operator, T is the preset threshold;

[0013] Step STP400, establish HOG features: calculate the gradient direction of each pixel and form a feature vector

[0014]

[0015] Among them, HOG(x, y) is the histogram of directional gradients, g(x, y) is the pixel gradient, θ k is the separated angle interval; then use principal component analysis PCA to calculate the main direction of the contour;

[0016] Step STP500, normalize the contour: move the centroid of the contour to the origin, then scale the contour to the unit scale, rotate the contour through principal component analysis PCA, align the main direction with the X-axis to obtain the normalized contour m′0; at the same time, normalize the contour sets M1, M2…M6 in step STP100 to obtain contour sets M1′, M2′…M6′;

[0017] Step STP600, using similarity measurement to compare whether the contour m'0 normalized in step STP500 belongs to the contour set M1', M2'...M6';

[0018] If m′0∈M1′, define the current target image I target The structure of the part belongs to a circuit breaker;

[0019] If m′0∈M2′, define the current target image I target The structure of the part belongs to isolating switch;

[0020] If m′0∈M3′, define the current target image I target The structure of the part belongs to the grounding switch;

[0021] If m′0∈M4′, define the current target image I target The structure of the part belongs to a mutual inductor;

[0022] If m′0∈M5′, define the current target image I target The structure of the part belongs to the arrester;

[0023] If m′0∈M6′, define the current target image I target The structure of the part is busbar.

[0024] In order to more accurately and efficiently identify and compare, and complete the classification of the structure of the parts in the judgment object, preferably, the contour sets M1, M2...M6 of the circuit breaker, disconnector, grounding switch, mutual inductor, lightning arrester and busbar structure in step STP100 are obtained by the following method:

[0025] Step STP111, setting up a shooting environment, installing a fixed X-ray device and a detector for receiving X-ray imaging in pairs, and completing debugging;

[0026] Step STP112, collect image samples, place the circuit breaker product or sample in front of the detector, change the angle each time you shoot, repeat shooting N times, and obtain N detection images; N∈[1,+∞];

[0027] Step STP113, performing contour extraction on the N detection images obtained in step STP112 to obtain N contour data;

[0028] Step STP114, the contour data obtained in step STP113 is expressed by a parameterized equation as

[0029] C(θ)=(r(θ)cos(θ),r(θ)sin(θ)

[0030] Wherein, r(θ) represents the contour radius of the circuit breaker at the angle θ; or, an implicit function is used to represent the contour, specifically: F(x, y, θ) = 0, where (x, y) represents the coordinates on the image plane, and θ represents the angle at which the current detection image is taken;

[0031] Step STP115, establish a circuit breaker profile set function M1, where:

[0032] The set expression established by parameterized equation is C = {C(θ)|θ∈[0,2π)},

[0033] The implicit function expression is C = {(x, y) | F (x, y, θ) = 0, θ ∈ [0, 2π)}, C = M1;

[0034] Step STP116, repeating steps STP112 to STP115, photographing the disconnector, grounding switch, transformer, lightning arrester and busbar respectively and obtaining corresponding contour sets M2, M3...M6.

[0035] In order to improve the accuracy and efficiency of the comparison, preferably, in step STP600, whether the contour m′0 belongs to one of the contour sets M1′, M2′…M6′ is determined by using the Hausdorff distance, specifically:

[0036]

[0037] Where d(x,y) is the distance between points;

[0038] Set a similarity threshold μ, if d H (m′0,M j ′)<μ, then determine m′0∈M j ′,j=1,2…6.

[0039] The present invention also provides a classification system for identifying GIS equipment parts based on X-ray detection images, including a processing unit and a storage module, wherein the storage module stores a software program for implementing the method described above, and the processing unit is used to execute the software program to realize X-ray detection image recognition and classification according to GIS equipment parts.

[0040] Beneficial effects:

[0041] 1. The method provided by the present invention can accurately classify different parts by establishing a contour set M for each part of GIS equipment (such as circuit breakers, disconnectors, grounding switches, mutual inductors, lightning arresters and busbars). Compared with traditional image recognition technology, this classification method can effectively reduce the possibility of misidentification and improve the accuracy of classification.

[0042] 2. The present invention provides a special preprocessing process. The original image is preprocessed by grayscale and denoising, which can significantly enhance the clarity and contrast of the image. This preprocessing step helps to extract clearer and more accurate features, thereby providing a better basis for subsequent classification and improving the overall recognition effect.

[0043] 3. The method provided by the present invention adopts the Hausdorff distance to judge the attribution of contours, which can realize automatic classification. Compared with the prior art, this method reduces the complexity of manual operation and the subjective error caused by manual judgment, making the classification process more intelligent.

[0044] 4. The present invention ensures the standardization and consistency of data collection by building a special shooting environment and shooting different parts separately. This feature makes the method highly adaptable and can be widely used in different types of GIS equipment detection, thus improving the universality of the technology.

[0045] 5. By effectively classifying the parts of GIS equipment, this method can help maintenance personnel quickly identify and locate problem parts, thereby achieving more efficient inspection and maintenance. This efficiency is particularly important in the power industry, because the rapid recovery of equipment is directly related to the stability of power supply.

[0046] 6. The implementation of the method provided by the present invention will generate a large amount of detection image data. By accumulating and analyzing these data, it can provide valuable reference for subsequent equipment maintenance and fault prediction, and lay the foundation for establishing the GIS atlas library of the industry. This data-driven method can continuously optimize maintenance strategies and improve the scientific nature of equipment management. DETAILED DESCRIPTION

[0047] 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.

[0048] Embodiment 1:

[0049] The present invention provides a classification method for identifying GIS equipment parts based on X-ray detection images, comprising the following steps:

[0050] In step STP100, contour sets M1, M2...M6 of circuit breakers, disconnectors, grounding switches, transformers, lightning arresters and busbar structures constituting the GIS equipment are established respectively; in order to more accurately and efficiently identify and compare, and complete the classification of the part structures in the judgment object, in this embodiment, the contour sets M1, M2...M6 of circuit breakers, disconnectors, grounding switches, transformers, lightning arresters and busbar structures in step STP100 are obtained by the following method:

[0051] Step STP111, setting up a shooting environment, installing a fixed X-ray device and a detector for receiving X-ray imaging in pairs, and completing debugging;

[0052] Step STP112, collect image samples, place the circuit breaker product or sample in front of the detector, change the angle each time you shoot, repeat shooting N times, and obtain N detection images; N∈[1,+∞];

[0053] Step STP113, performing contour extraction on the N detection images obtained in step STP112 to obtain N contour data;

[0054] Step STP114, the contour data obtained in step STP113 is expressed by a parameterized equation as

[0055] C(θ)=(r(θ)cos(θ),r(θ)sin(θ)

[0056] Wherein, r(θ) represents the contour radius of the circuit breaker at the angle θ; or, an implicit function is used to represent the contour, specifically: F(x, y, θ) = 0, where (x, y) represents the coordinates on the image plane, and θ represents the angle at which the current detection image is taken;

[0057] Step STP115, establish a circuit breaker profile set function M1, where:

[0058] The set expression established by parameterized equation is C = {C(θ)|θ∈[0,2π)},

[0059] The implicit function expression is C = {(x, y) | F (x, y, θ) = 0, θ ∈ [0, 2π)}, C = M1;

[0060] Step STP116, repeat steps STP112 to STP115, respectively photograph the disconnector, grounding switch, transformer, lightning arrester and busbar and obtain corresponding contour sets M2, M3...M6.

[0061] Step STP200, input and recognize the original image I Original , the original image I Original Perform preprocessing including grayscale and denoising to obtain the target image I target ;

[0062] Step STP300, extracting the target image I based on the Canny edge detection algorithm target The edge contour m0 in the figure, the pixels that make up the contour are represented by E(x, y)

[0063]

[0064] Among them, E(x, y) represents a function, represents the function operator, T is the preset threshold;

[0065] Step STP400, establish HOG features: calculate the gradient direction of each pixel and form a feature vector

[0066]

[0067] Among them, HOG(x, y) is the histogram of directional gradients, g(x, y) is the pixel gradient, θ k is the separated angle interval; then use principal component analysis PCA to calculate the main direction of the contour;

[0068] Step STP500, normalize the contour: move the centroid of the contour to the origin, then scale the contour to the unit scale, rotate the contour through principal component analysis PCA, align the main direction with the X-axis to obtain the normalized contour m′0; at the same time, normalize the contour sets M1, M2…M6 in step STP100 to obtain contour sets M1′, M2′…M6′;

[0069] In step STP600, a similarity metric is used to compare whether the contour m′0 normalized in step STP500 belongs to the contour set M1′, M2′…M6′. In order to improve the accuracy and efficiency of the comparison, in this step, whether the contour m′0 belongs to one of the contour sets M1′, M2′…M6′ is determined by using the Hausdorff distance, which is specifically:

[0070]

[0071] Where d(x,y) is the distance between points;

[0072] Set a similarity threshold μ, if d H (m′0,M j ′)<μ, then determine m′0∈M j ′,j=1,2…6.

[0073] If m′0∈M1′, define the current target image I target The structure of the part belongs to a circuit breaker;

[0074] If m′0∈M2′, define the current target image I target The structure of the part belongs to isolating switch;

[0075] If m′0∈M3′, define the current target image I target The structure of the part belongs to the grounding switch;

[0076] If m′0∈M4′, define the current target image I target The structure of the part belongs to a mutual inductor;

[0077] If m′0∈M5′, define the current target image I target The structure of the part belongs to the arrester;

[0078] If m′0∈M6′, define the current target image I target The structure of the part is busbar.

[0079] It is worth noting that in real detection work, the same detection image may contain two structures or even more than two structures at the same time. Although the probability of this happening is not high, it does exist for special detection scenarios. In this case, the system will simultaneously define the target image I target The part structure is multiple, for example, including a lightning rod and a busbar at the same time, or including an isolating switch and a grounding switch at the same time.

[0080] Embodiment 2:

[0081] The present invention also provides a classification system for identifying GIS equipment parts based on X-ray detection images, including a processing unit and a storage module, wherein the storage module stores a software program for implementing the method described above, and the processing unit is used to execute the software program to realize X-ray detection image recognition and classification according to GIS equipment parts.

[0082] 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 classification method for identifying GIS equipment parts based on X-ray detection images, characterized in that: The steps include: Step STP100, respectively establishing the contour sets M1, M2...M6 of the circuit breaker, disconnector, grounding switch, transformer, lightning arrester and busbar structure constituting the GIS equipment; Step STP200, input and recognize the original image I Original , the original image I Original Perform preprocessing including grayscale and denoising to obtain the target image I target ; Step STP300, extracting the target image I based on the Canny edge detection algorithm target The edge contour m0 in the figure, the pixels that make up the contour are represented by E(x, y) Among them, E(x, y) represents the function, represents the function operator, T is the preset threshold; Step STP400, establish HOG features: calculate the gradient direction of each pixel and form a feature vector Among them, HOG(x, y) is the histogram of directional gradients, g(x, y) is the pixel gradient, θ k is the separated angle interval; then use principal component analysis PCA to calculate the main direction of the contour; Step STP500, normalize the contour: move the centroid of the contour to the origin, then scale the contour to unit scale, rotate the contour through principal component analysis PCA, align the main direction with the X-axis to obtain the normalized contour m ′ 0; at the same time, normalize the contour sets M1, M2…M6 in step STP100 to obtain contour set M1 ′ ,M2 ′ …M6 ′ ; Step STP600, using a similarity metric to compare the contours m normalized in step STP500 ′ 0 belongs to the contour set M1 ′ ,M2 ′ …M6 ′ ; If m ′ 0∈M1 ′ , define the current target image I target The structure of the part belongs to a circuit breaker; If m ′ 0∈M2 ′ , define the current target image I target The structure of the part belongs to isolating switch; If m ′ 0∈M3 ′ , define the current target image I target The structure of the part belongs to the grounding switch; If m ′ 0∈M4 ′ , define the current target image I target The structure of the part belongs to a mutual inductor; If m ′ 0∈M5 ′ , define the current target image I target The structure of the part is a lightning arrester; If m ′ 0∈M6 ′ , define the current target image I target The structure of the part is busbar.

2. The classification method for identifying GIS equipment parts based on X-ray detection images according to claim 1 is characterized in that: In step STP100, the contour sets M1, M2…M6 of the circuit breaker, disconnector, earthing switch, transformer, lightning arrester and busbar structure are obtained by the following method: Step STP111, setting up a shooting environment, installing a fixed X-ray device and a detector for receiving X-ray imaging in pairs, and completing debugging; Step STP112, collect image samples, place the circuit breaker product or sample in front of the detector, change the angle each time you shoot, repeat shooting N times, and obtain N detection images; N∈[1,+∞]; Step STP113, performing contour extraction on the N detection images obtained in step STP112 to obtain N contour data; Step STP114, the contour data obtained in step STP113 is expressed by a parameterized equation as C(θ)=(r(θ)cos(θ),r(θ)sin(θ) Wherein, r(θ) represents the contour radius of the circuit breaker at the angle θ; or, an implicit function is used to represent the contour, specifically: F(x, y, θ) = 0, where (x, y) represents the coordinates on the image plane, and θ represents the angle at which the current detection image is taken; Step STP115, establish a circuit breaker profile set function M1, where: The set expression established by parameterized equation is C = {C(θ)|θ∈[0,2π)}, The implicit function expression is C = {(x, y) | F (x, y, θ) = 0, θ ∈ [0, 2π)}, C = M1; Step STP116, repeating steps STP112 to STP115, photographing the disconnector, grounding switch, transformer, lightning arrester and busbar respectively and obtaining corresponding contour sets M2, M3...M6.

3. The classification method for identifying GIS equipment parts based on X-ray detection images according to claim 1 is characterized in that: Step STP600 in contour m ′ 0 belongs to the contour set M1 ′ ,M2 ′ …M6 ′ One method is to use Hausdorff distance judgment, specifically: Where d(x,y) is the distance between points; Set a similarity threshold μ, if d H (m ′ 0,M j ′ )<μ, then determine m ′ 0∈M j ′ ,j=1,2…6.

4. A classification system for identifying GIS equipment parts based on X-ray detection images, including a processing unit and a storage module, characterized in that: The storage module stores a software program for implementing the method described in any one of claims 1 to 3, and the corresponding processing unit is used to execute the corresponding software program to realize X-ray detection image recognition and classification according to GIS equipment parts.

Images