A method and system for defect recognition of a cable intermediate joint

Abstract

The application discloses a kind of defect identification method and system of cable intermediate joint, construct the electric field distribution graph of multiple cable intermediate joints to be judged and carry out analysis, according to the analysis result from multiple cable intermediate joints to be judged determine sampling cable intermediate joint, obtain the sampling image of each sampling cable intermediate joint and carry out image preprocessing, generate target contrast image set, adopt target contrast image set feedback optimization target intermediate joint defect identification model, wherein, target intermediate joint defect identification model is generated by preset model training process, and the defect of the image of the cable intermediate joint to be identified is detected by target defect identification model, and the defect result of the cable intermediate joint to be identified is output;It has reached the beneficial effect that the light, angle and other factors of the image acquisition environment are influenced by overcoming prior art, and inaccurate defect identification result is prone to occur.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a method and system for identifying defects in cable joints. Background Technology

[0002] With the increasing prevalence of power grid cabling, the use of intermediate joints, as a core accessory for intermediate connections in cross-linked cables, is on the rise. Research on power system accidents shows that most cable line faults are caused by improper operation during construction. On the one hand, excessive mechanical traction during cable laying can damage the cable, or excessive bending can cause cable damage. On the other hand, improper operation during the installation of intermediate joints can lead to construction defects at the joint. Construction defects in cable intermediate joints are the most insidious, and they account for the largest proportion of cable accidents. The presence of construction defects directly reduces the insulation performance of the joint, shortens its service life, and makes it prone to breakdown after installation.

[0003] Currently, most existing methods for identifying construction defects in power cable joints rely on manual identification, supplemented by image recognition technology. However, due to factors such as lighting and angle in the image acquisition environment, inaccurate defect identification results are prone to occur. Summary of the Invention

[0004] This invention provides a method and system for identifying defects in cable intermediate joints, which solves the technical problem that existing methods for identifying construction defects in power cable intermediate joints mostly rely on manual identification and image recognition technology for assistance, but are prone to inaccurate defect identification results due to factors such as light and angle in the image acquisition environment.

[0005] The first aspect of this invention provides a method for defect identification of cable joints, comprising:

[0006] Construct and analyze the electric field distribution diagrams of multiple cable intermediate joints to be judged, and determine the sampling cable intermediate joints from the multiple cable intermediate joints to be judged based on the analysis results;

[0007] Acquire sampling images of each of the sampling cable intermediate joints and perform image preprocessing to generate a target comparison image set;

[0008] The target intermediate joint defect identification model is optimized using the target comparison image set, wherein the target intermediate joint defect identification model is generated through a preset model training process;

[0009] The target defect identification model is used to detect defects in the image of the intermediate joint of the cable to be identified, and the defect results of the intermediate joint of the cable to be identified are output.

[0010] Optionally, the step of constructing and analyzing the electric field distribution maps of multiple cable joints to be judged, and determining the sampling cable joint from the multiple cable joints to be judged based on the analysis results, includes:

[0011] A three-dimensional simulation model was established using the structural parameters of multiple cable intermediate joints to be judged.

[0012] The three-dimensional simulation model is meshed using a parametric structure language, and the electric field strength of the intermediate joint of the cable to be judged is calculated based on the finite element method.

[0013] Based on the electric field strength and the three-dimensional simulation model, multiple electric field distribution maps corresponding to the intermediate joints of the cables to be judged are generated.

[0014] Compare each of the described electric field distribution diagrams with the preset cable axial symmetry condition;

[0015] If the electric field distribution map satisfies the preset cable axial symmetry condition, then the electric field distribution map is analyzed, and sampled cable intermediate joints without abnormalities are selected from multiple intermediate joints of the cables to be judged.

[0016] Optionally, the step of analyzing the electric field distribution map and selecting abnormal sampled cable joints from multiple cable joints to be judged if the electric field distribution map satisfies the preset cable axial symmetry condition includes:

[0017] If the electric field distribution map satisfies the preset cable axial symmetry condition, then multiple corresponding two-dimensional field strength variation curves are extracted based on each electric field distribution map;

[0018] Calculate the slope values ​​of multiple corresponding initial change curves based on the two-dimensional field strength change curves described above.

[0019] Select multiple corresponding target change curve slope values ​​from multiple initial change curve slope values ​​according to a preset interval;

[0020] Calculate the change between the slope values ​​of two adjacent target change curves to generate multiple target change values;

[0021] Compare the changes in each target with the changes in the preset standard;

[0022] If all the target changes are less than the preset standard changes, then the intermediate joint of the cable to be judged associated with the electric field distribution map is determined to be normal, and the intermediate joint of the cable to be judged is taken as the intermediate joint of the sampling cable.

[0023] Optionally, the step of acquiring sampled images of each of the sampled cable intermediate joints and performing image preprocessing to generate a target comparison image set includes:

[0024] Acquire sampling images of each of the sampling cable intermediate joints and perform clustering to generate multiple cluster sets;

[0025] Feature extraction is performed on the multiple sampled images associated with each of the cluster sets to generate multiple initial pixel feature data;

[0026] Multiple initial pixel feature data that have similar features are selected from the multiple initial pixel feature data as the first fused feature data;

[0027] Imaging and denoising operations are performed on the initial pixel feature data that do not have similar features to generate corresponding target pixel feature data;

[0028] Calculate the average pixel value corresponding to the target pixel feature data, and use the average pixel value as the second fused feature data;

[0029] The first fused feature data and the second fused feature data are stitched together to generate a target comparison image set.

[0030] Optionally, the model training process includes:

[0031] Acquire images of cable joints in various power scenarios;

[0032] The cable joint image is preprocessed to generate a training image sample set;

[0033] The training image sample set is used as input to a preset initial intermediate joint defect identification model for training, and corresponding training indicators are generated based on the training results.

[0034] Calculate the training loss value between the training metric and the associated standard metric;

[0035] Compare the training loss value with the preset standard loss value;

[0036] If the training loss value is less than or equal to the preset standard loss value, then training is stopped and a target intermediate joint defect identification model is generated.

[0037] If the training loss value is greater than the preset standard loss value, then the network parameters of the preset initial intermediate joint defect identification model are adjusted according to the preset gradient.

[0038] The process jumps to the step of training the preset initial intermediate joint defect identification model using the training image sample set as input, generating corresponding training indicators based on the training results, until the training loss value is less than or equal to the preset standard loss value, and then generates the target intermediate joint defect identification model.

[0039] A second aspect of the present invention provides a defect identification system for cable joints, comprising:

[0040] The analysis module is used to construct and analyze the electric field distribution diagrams of multiple cable intermediate joints to be judged, and to determine the sampling cable intermediate joint from the multiple cable intermediate joints to be judged based on the analysis results.

[0041] The image preprocessing module is used to acquire sampled images of each of the sampling cable intermediate joints and perform image preprocessing to generate a target comparison image set;

[0042] The model optimization module is used to optimize the target intermediate joint defect identification model using the target comparison image set, wherein the target intermediate joint defect identification model is generated by a preset model training module;

[0043] The defect detection module is used to perform defect detection on the image of the intermediate joint of the cable to be identified through the target defect identification model, and output the defect results of the intermediate joint of the cable to be identified.

[0044] Optionally, the parsing module includes:

[0045] The three-dimensional simulation model submodule is used to establish a corresponding three-dimensional simulation model using the structural parameters of multiple cable intermediate joints to be judged.

[0046] The electric field strength submodule is used to call the parametric structure language to mesh the three-dimensional simulation model and calculate the electric field strength of the intermediate joint of the cable to be judged based on the finite element method.

[0047] The electric field distribution map submodule is used to generate multiple electric field distribution maps corresponding to the intermediate joints of the cables to be judged based on the electric field strength and the three-dimensional simulation model.

[0048] The symmetry condition comparison submodule is used to compare each electric field distribution diagram with the preset cable axial symmetry condition.

[0049] The sampling cable intermediate joint submodule is used to analyze the electric field distribution map if the electric field distribution map meets the preset cable axial symmetry condition, and then select the sampling cable intermediate joint without abnormality from multiple cable intermediate joints to be judged.

[0050] Optionally, the sampling cable intermediate connector submodule includes:

[0051] The two-dimensional field strength variation curve unit is used to extract multiple corresponding two-dimensional field strength variation curves based on each electric field distribution map if the electric field distribution map satisfies the preset cable axial symmetry condition.

[0052] The initial change curve slope value unit is used to calculate multiple corresponding initial change curve slope values ​​based on each of the two-dimensional field strength change curves.

[0053] The target change curve slope value unit is used to select multiple corresponding target change curve slope values ​​from multiple initial change curve slope values ​​according to a preset interval;

[0054] The target change unit is used to calculate the change between the slope values ​​of two adjacent target change curves and generate multiple target change units.

[0055] The change comparison unit is used to compare the change of each target with the change of a preset standard.

[0056] The determination unit is used to determine that the intermediate joint of the cable to be judged associated with the electric field distribution map is not abnormal if all the target changes are less than the preset standard changes, and to use the intermediate joint of the cable to be judged as the intermediate joint of the sampling cable.

[0057] Optionally, the image preprocessing module includes:

[0058] The clustering set submodule is used to acquire the sampled images of each of the sampled cable intermediate joints and perform clustering to generate multiple cluster sets;

[0059] The initial pixel feature data submodule is used to extract features from multiple sampled images associated with each of the cluster sets to generate multiple initial pixel feature data.

[0060] The first fusion feature data submodule is used to select multiple initial pixel feature data that have similar features from multiple initial pixel feature data as the first fusion feature data;

[0061] The target pixel feature data submodule is used to perform imaging and denoising operations on the initial pixel feature data that do not have similar features, and generate corresponding target pixel feature data.

[0062] The second fusion feature data submodule is used to calculate the average pixel value corresponding to the target pixel feature data and use the average pixel value as the second fusion feature data.

[0063] The target comparison image set submodule is used to stitch together the first fused feature data and the second fused feature data to generate a target comparison image set.

[0064] Optionally, it also includes:

[0065] The cable joint image submodule is used to acquire images of cable joints in various power scenarios.

[0066] The training image sample set submodule is used to perform image preprocessing on the cable intermediate joint image to generate a training image sample set.

[0067] The training index submodule is used to train the preset initial intermediate joint defect identification model by inputting the training image sample set, and to generate corresponding training indexes based on the training results.

[0068] The training loss value submodule is used to calculate the training loss value between the training metric and the associated standard metric.

[0069] The loss value comparison submodule is used to compare the training loss value with the preset standard loss value;

[0070] The first data processing submodule is used to stop training and generate a target intermediate joint defect identification model if the training loss value is less than or equal to the preset standard loss value.

[0071] The second data processing submodule is used to adjust the network parameters of the preset initial intermediate joint defect identification model according to a preset gradient if the training loss value is greater than the preset standard loss value.

[0072] The jump-rotor module is used to jump to execute the steps of training the preset initial intermediate joint defect identification model by inputting the training image sample set, generating corresponding training indicators based on the training results, until the training loss value is less than or equal to the preset standard loss value, and generating the target intermediate joint defect identification model.

[0073] As can be seen from the above technical solutions, the present invention has the following advantages:

[0074] This paper constructs and analyzes electric field distribution maps of multiple cable joints to be judged. Based on the analysis results, it identifies sampled cable joints from these joints, acquires sampling images of each joint, performs image preprocessing, and generates a target reference image set. The target reference image set is used to optimize the defect identification model for the cable joints. This model is generated through a pre-defined model training process. The model detects defects in the images of the cable joints to be identified and outputs the defect results. This approach addresses the problem that existing methods for identifying construction defects in power cable joints often rely on manual identification with image recognition technology as an aid. However, these methods are susceptible to inaccurate results due to factors such as lighting and angle in the image acquisition environment. This paper overcomes the limitations of existing technologies, which are prone to inaccurate defect identification due to these environmental factors. Attached Figure Description

[0075] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0076] Figure 1 A flowchart illustrating the steps of a defect identification method for cable intermediate joints provided in Embodiment 1 of the present invention;

[0077] Figure 2 This is a flowchart illustrating the steps of a defect identification method for cable intermediate joints provided in Embodiment 2 of the present invention.

[0078] Figure 3 This is a structural block diagram of a defect identification system for cable intermediate joints provided in Embodiment 3 of the present invention. Detailed Implementation

[0079] This invention provides a method and system for identifying defects in cable intermediate joints, which addresses the problem that existing methods for identifying construction defects in power cable intermediate joints mostly rely on manual identification, supplemented by image recognition technology. However, these methods are often limited by factors such as lighting and angle in the image acquisition environment, leading to inaccurate defect identification results.

[0080] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0081] Please see Figure 1 , Figure 1 The flowchart illustrates the steps of a defect identification method for cable intermediate joints provided in Embodiment 1 of the present invention.

[0082] This invention provides a method for defect identification of cable joints, comprising:

[0083] Step 101: Construct and analyze the electric field distribution diagrams of multiple cable intermediate joints to be judged, and determine the sampling cable intermediate joints from the multiple cable intermediate joints to be judged based on the analysis results.

[0084] The cable joint to be judged refers to the cable joint used for screening based on its electric field conditions.

[0085] An electric field distribution diagram refers to the distribution diagram formed by the electric field lines around the intermediate joint of the cable to be judged, which is used to show the direction and magnitude of the electric field of the intermediate joint of the cable to be judged.

[0086] Sampling cable joints refer to defect-free cable joints selected for model optimization based on their electric field characteristics.

[0087] In this embodiment of the invention, electric field distribution diagrams of multiple cable intermediate joints to be judged are obtained for electric field analysis. Based on the analysis of the electric field distribution diagrams, it is determined whether the conditions of cable axial symmetry are met and whether distortion occurs. If the conditions of cable axial symmetry are not met or distortion occurs, the cable intermediate joint to be judged is determined to have a defect or abnormality. From multiple cable intermediate joints to be judged, the cable intermediate joints that meet the conditions of cable axial symmetry and do not show distortion are selected as the sampling cable intermediate joints.

[0088] Step 102: Obtain the sampling images of each sampling cable intermediate joint and perform image preprocessing to generate a target comparison image set.

[0089] The use of images refers to sampling a full-view image of the cable mid-joint.

[0090] Image preprocessing refers to the process of fusing features in sampled images.

[0091] The target reference image set refers to the reference images generated after image preprocessing that are input into the trained target joint defect recognition model for optimization.

[0092] In this embodiment of the invention, sampling images of each sampling cable intermediate joint are acquired, and feature fusion is performed on the images to generate a target comparison image set.

[0093] Step 103: Use the target comparison image set to optimize the target intermediate joint defect identification model. The target intermediate joint defect identification model is generated through a preset model training process.

[0094] In this embodiment of the invention, the target comparison image set generated after image preprocessing is used as the training set and input into the already trained target intermediate joint defect identification model to optimize and update the parameters of the target intermediate joint defect identification model.

[0095] Step 104: Use the target defect recognition model to perform defect detection on the image of the intermediate joint of the cable to be identified, and output the defect results of the intermediate joint of the cable to be identified.

[0096] In this embodiment of the invention, a target defect identification model is used to detect defects in the image of the cable intermediate joint to be identified. If no significant defect border is identified in the image of the cable intermediate joint to be identified, it is determined that the cable intermediate joint associated with the image of the cable intermediate joint to be identified has no abnormality. If a significant defect border is identified in the image of the cable intermediate joint to be identified, an associated identification method is selected for analysis based on the type of defect appearing within the significant defect border, thereby determining the target defect of the cable intermediate joint to be identified.

[0097] In this invention, electric field distribution maps of multiple cable joints to be judged are constructed and analyzed. Based on the analysis results, sampled cable joints are determined from the multiple cable joints to be judged. Sampling images of each sampled cable joint are acquired and preprocessed to generate a target reference image set. The target reference image set is used to optimize the target joint defect identification model. The target joint defect identification model is generated through a pre-set model training process. The target defect identification model is used to detect defects in the images of the cable joints to be identified and outputs the defect results of the cable joints to be identified. This invention solves the problem that existing methods for identifying construction defects in power cable joints mostly rely on manual identification and image recognition technology for assistance. However, these methods are easily affected by factors such as light and angle in the image acquisition environment, leading to inaccurate defect identification results. This invention overcomes the problem of inaccurate defect identification results caused by factors such as light and angle in the image acquisition environment in existing technologies.

[0098] Please see Figure 2 , Figure 2 The flowchart illustrates the steps of a defect identification method for a cable intermediate joint provided in Embodiment 2 of the present invention.

[0099] This invention provides a method for defect identification of cable joints, comprising:

[0100] Step 201: Use the structural parameters of multiple intermediate joints of the cables to be judged to establish corresponding three-dimensional simulation models.

[0101] In this embodiment of the invention, a three-dimensional simulation model corresponding to each cable intermediate joint to be judged is constructed based on the structural parameters of multiple cable intermediate joints to be judged.

[0102] The specific implementation method refers to the actual 10kV single-core cold-shrink connector generated by the cable manufacturer. A three-dimensional simulation model is constructed according to the structural parameters of the cable intermediate joint at a 1:1 scale. The structural parameters are shown in Table 1 below:

[0103] Table 1. Structural Parameters

[0104]

[0105]

[0106] Step 202: Use parametric structure language to mesh the 3D simulation model and calculate the electric field strength of the cable joint to be judged based on the finite element method.

[0107] In this embodiment of the invention, the three-dimensional simulation model is meshed using the parametric structural language APDL, and the electric field strength of the cable is calculated using the finite element method.

[0108] Specifically, when meshing the model using the parametric structural language APDL in ANSYS software, for volumes with chamfers, the lines are first segmented proportionally before volume meshing; for volumes without chamfers, a 4-level SmartSize intelligent meshing method is used. For copper mesh and silicone grease models, which are essentially shell-shaped with a large aspect ratio, a mapped meshing method is used to improve mesh quality while reducing mesh size. Then, smaller mesh angles are selected for local optimization to ensure the accuracy of the calculation results. When applying boundary conditions, values ​​are applied to the conductors... The voltage is zero potential applied to the copper shielding layer outside the connector. In the simulation software, based on the required material size and parameters, the element type is defined as plane121, electric field analysis is specified, an axisymmetric model is established, the mesh size level is set to 5, the specified mesh element is triangular, and finally, the necessary node elements are generated on the surface, and the nodes are automatically numbered.

[0109] When calculating electric field strength, the electromagnetic wave wavelength of power frequency AC voltage is 3 × 10⁶ m, which is much larger than the insulation distance of common high-voltage electrical equipment. The voltage change frequency of power frequency equipment is relatively low; therefore, the transient electric field caused by this low-frequency voltage can be analyzed as an electrostatic field. Solving the electrostatic field problem can be reduced to solving a set of linear ordinary differential equations, or transformed into solving a functional extremum problem using the variational method; both methods yield consistent results. The finite element method is a numerical calculation method developed based on the variational principle. It expresses the electrostatic field energy as a specific potential function and the integral expression of the derivative of this function. Specifically, a mathematical model and an electric field model are established. Based on boundary conditions (the potential satisfies the first kind of boundary conditions on the boundaries), the domain is divided into a finite number of discrete polygonal subdomains. Then, an interpolation function is used to approximate the function to be solved. The magnitude of the field strength and energy distribution in each subdomain are calculated based on the known potential distribution and the electric field energy functional formula.

[0110] The expression for the first type of boundary condition is:

[0111]

[0112] The energy functional formula is expressed as follows:

[0113]

[0114] In the formula, Let E represent the electric field energy density, and let ρ be the charge density, ε be the permittivity, and δ be the variational sign in a planar domain D with boundary c. Let E be the unit electric field strength. It is a Hamiltonian operator.

[0115] By dividing the region of interest into several discrete small cells, if these cells are sufficiently finely divided, the continuous electrostatic energy of the complete region can be approximated by a finite number of discrete nodal potential functions. Through this operation, the problem of finding the extrema of the electrostatic energy functional is transformed into solving for the extrema of a set of multivariate functions. Generally, such systems of equations have distinct characteristics, and the solution to the system can be obtained using appropriate methods, i.e., the values ​​of the nodal potentials in the grid. Based on the values ​​of the nodal potentials, the electric field strength corresponding to each cable joint to be judged can be generated.

[0116] Step 203: Based on the electric field strength and the three-dimensional simulation model, generate electric field distribution diagrams corresponding to multiple intermediate joints of the cables to be judged.

[0117] In this embodiment of the invention, based on the electric field strength corresponding to each intermediate joint of the cable to be judged, and combined with the three-dimensional simulation model corresponding to each intermediate joint of the cable to be judged, multiple electric field distribution maps corresponding to the intermediate joints of the cable to be judged are generated.

[0118] Step 204: Compare each electric field distribution diagram with the preset cable axial symmetry condition.

[0119] The preset cable axial symmetry condition refers to one of the conditions used to determine whether each intermediate joint of the cable to be judged is abnormal, that is, whether the electric field distribution diagram corresponding to each intermediate joint of the cable to be judged coincides symmetrically along the cable axis.

[0120] In this embodiment of the invention, the electric field distribution diagram corresponding to each intermediate joint of the cable to be judged is compared with the preset axial symmetry condition of the cable.

[0121] Step 205: If the electric field distribution diagram meets the preset cable axial symmetry condition, then analyze the electric field distribution diagram and select the sampled cable intermediate joints without abnormalities from multiple cable intermediate joints to be judged.

[0122] Furthermore, step 205 may include the following sub-steps:

[0123] S11. If the electric field distribution diagram satisfies the preset cable axial symmetry condition, then extract multiple corresponding two-dimensional field strength variation curves based on each electric field distribution diagram.

[0124] A two-dimensional electric field strength variation curve refers to the electric field strength at the corresponding position in the electric field distribution diagram obtained from the center of the cable shaft in the direction perpendicular to the cable shaft, generating a two-dimensional electric field strength variation trend curve. The x-axis of the two-dimensional electric field strength variation trend curve is the distance from the center of the cable shaft, and the y-axis of the two-dimensional electric field strength variation trend curve is the electric field strength at that distance.

[0125] In this embodiment of the invention, if the electric field distribution map corresponding to the intermediate joint of the cable to be judged coincides symmetrically along the cable axis, it is determined that the electric field distribution map corresponding to the intermediate joint of the cable to be judged satisfies the preset cable axis symmetry condition, and further, multiple corresponding two-dimensional field strength change curves are extracted from the electric field distribution map that satisfies the preset cable axis symmetry condition.

[0126] It is worth mentioning that if the electric field distribution diagram corresponding to the intermediate joint of the cable to be judged does not coincide symmetrically along the cable axis, the intermediate joint of the cable to be judged is directly determined to have a defect and is thus screened out.

[0127] S12. Calculate the slope values ​​of multiple corresponding initial change curves based on each two-dimensional field strength change curve.

[0128] In this embodiment of the invention, the x-axis of the two-dimensional field strength change trend curve is the distance from the center of the cable shaft, and the y-axis of the two-dimensional field strength change trend curve is the magnitude of the field strength at the distance. The slope value of the initial change curve corresponding to each two-dimensional field strength change curve is calculated. That is, a two-dimensional field strength change curve contains multiple initial change curve slope values.

[0129] S13. Select multiple corresponding target change curve slope values ​​from multiple initial change curve slope values ​​according to the preset interval.

[0130] The preset interval refers to any interval selected from the two-dimensional field strength variation curve.

[0131] In this embodiment of the invention, a preset interval is arbitrarily selected on the x-axis, thereby using the slope values ​​of multiple initial change curves within the preset interval on the x-axis as the slope values ​​of the target change curve.

[0132] S14. Calculate the change between the slope values ​​of two adjacent target change curves to generate multiple target change values.

[0133] In this embodiment of the invention, the change in slope between two adjacent target change curves within a preset interval is calculated, thereby generating multiple target change values.

[0134] S15. Compare the changes in each target with the changes in the preset standard.

[0135] In this embodiment of the invention, the change in each target is compared with the change in a preset standard.

[0136] S16. If all target changes are less than the preset standard changes, the intermediate joint of the cable to be judged associated with the electric field distribution map is determined to be normal, and the intermediate joint of the cable to be judged is taken as the intermediate joint of the sampling cable.

[0137] In this embodiment of the invention, if all target changes are less than the preset standard changes, it indicates that the trend of the cable intermediate joint to be judged is not distorted. Therefore, the cable intermediate joint to be judged associated with the electric field distribution map is determined to be normal, and the cable intermediate joint to be judged is used as the sampling cable intermediate joint.

[0138] It is worth mentioning that if any target change is greater than or equal to the preset standard change, it indicates that the change trend of the intermediate joint of the cable to be judged is distorted, and the intermediate joint of the cable to be judged associated with the electric field distribution map is determined to have a defect.

[0139] Step 206: Obtain the sampling images of each sampling cable intermediate joint and perform image preprocessing to generate a target comparison image set.

[0140] Furthermore, step 206 may include the following sub-steps:

[0141] S21. Obtain the sampling images of each sampling cable intermediate joint and perform clustering to generate multiple cluster sets.

[0142] In this embodiment of the invention, based on the electric field distribution characteristics, cluster analysis is performed on the sampled images of each sampled cable intermediate joint to generate multiple cluster sets.

[0143] It is worth mentioning that the electric field distribution of a normal, defect-free cable joint should be basically symmetrical, and the change in electric field intensity should not have obvious distortions such as sudden decreases or drops.

[0144] S22. Extract features from multiple sampled images associated with each cluster set to generate multiple initial pixel feature data.

[0145] In this embodiment of the invention, pixel features are extracted from the sampled images in each cluster set to generate multiple initial pixel feature data.

[0146] It is worth mentioning that the initial pixel feature data includes second-order moments and contrast.

[0147] In practical implementation, to facilitate the method's implementation, the above process can be converted into a formulaic encapsulation. The expression for the second moment is as follows:

[0148]

[0149] In the formula, f represents the second moment, L represents the gray level of the sampled image, i represents the first gray value, and j represents the second gray value.

[0150] It is worth mentioning that the grayscale level of the sampled image is 256.

[0151] In practical implementation, to simplify the method, the above process can be converted into a formula encapsulated in the form of a comparison, as shown below:

[0152]

[0153] In the formula, y represents the contrast ratio, and P d Gray-level co-occurrence matrix, The normalized gray-level co-occurrence matrix, where m is the number of sets outside the intersection, and n represents the regular variables from 1 to L-1.

[0154] It is worth mentioning that the gray-level co-occurrence matrix is ​​used to calculate the number of times the same texture appears. For example, the value of p(1,1) refers to the number of times the gray values ​​are arranged in (1,1).

[0155] S23. Select multiple initial pixel feature data that have similar features from multiple initial pixel feature data as the first fused feature data.

[0156] In this embodiment of the invention, multiple initial pixel feature data that have similar features are selected from multiple initial pixel feature data as the first fused feature data.

[0157] In the specific implementation, the intersection of the initial pixel feature data of all sampled images in the cluster set is used as the first fusion feature. The intersection refers to the intersection of the pixel features of multiple sampled images. The sampled images are the same cable joints, and there are similar pixel features between them. These features are extracted into a set, and the initial pixel feature data in the set is used as the first fusion feature data.

[0158] S24. Perform imaging and denoising operations on the initial pixel feature data that do not have similar features to generate the corresponding target pixel feature data.

[0159] In this embodiment of the invention, imaging and denoising operations are performed on initial pixel feature data that do not have similar features to generate corresponding target pixel feature data.

[0160] In the specific implementation, the initial pixel feature data other than the intersection are processed by imaging, as well as image pre-filtering, noise reduction and other processing to generate the corresponding target pixel feature data.

[0161] S25. Calculate the average pixel value corresponding to the target pixel feature data, and use the average pixel value as the second fused feature data.

[0162] In this embodiment of the invention, the average pixel value corresponding to the target pixel feature data is calculated, and the average pixel value is used as the second fused feature data.

[0163] In the specific implementation, the initial pixel feature data other than the intersection are processed by imaging, image pre-filtering, noise reduction and other processing. The pixel features are summed and divided by the total number to obtain the pixel average value, which is then used as the second fused feature data.

[0164] In practical implementation, to simplify the method, the above process can be converted into a formula, and the average pixel value can be expressed as follows:

[0165]

[0166] In the formula, f represents the average pixel value. i Let represent the i-th second moment.

[0167] It is worth mentioning that f i Different i represent different sets.

[0168] S26. The first fusion feature data and the second fusion feature data are spliced ​​together to generate a target comparison image set.

[0169] In this embodiment of the invention, the first fusion feature data and the second fusion feature data are spliced ​​together to obtain a spliced ​​image containing fusion features, and all spliced ​​images are merged into a target comparison image set.

[0170] It is worth mentioning that the first fusion feature is the intersection of pixel features of all sampled images in the cluster set;

[0171] The second fusion feature is the average value of pixel features in the cluster set other than the first fusion feature, calculated after preprocessing.

[0172] Clustering – Using morphological operators to cluster and merge neighboring similar classification regions.

[0173] To ensure optimization, multiple sampled cable joints are typically identified, and features from multiple sampled images of these joints are fused. Since unnecessary interference may occur during image acquisition, feature fusion of multiple sampled images from various cable joints is necessary to eliminate the influence of factors such as lighting and angle in the image acquisition environment. First, the sampled images are clustered based on the changing trend of the electric field distribution. This ensures that the electric field distribution of the sampled images in each cluster is similar, facilitating unified feature fusion.

[0174] In this embodiment, feature fusion refers to taking the intersection of pixel features from all sampled images in the cluster set as the first fusion feature, and then calculating the average value of the other pixel features outside the intersection after preprocessing, using this average value as the second fusion feature. Taking the intersection of pixel features from the sampled images means extracting the common parts from all sampled images, thereby avoiding the influence of different image acquisition environments on the recognition results. The second fusion feature is the average fusion of other pixel features, which minimizes the impact of image noise unrelated to the cable joint on the recognition results.

[0175] Step 207: Use the target comparison image set to optimize the target intermediate joint defect identification model. The target intermediate joint defect identification model is generated through a preset model training process.

[0176] In this embodiment of the invention, the acquired images are fused to obtain a stitched image containing fused features. This stitched image with fused features is then used as a training set and input into the target intermediate joint defect identification model to optimize and update the parameters of the model. In this embodiment, the target intermediate joint defect identification model can be constructed using a Convolutional Neural Network (CNN) and feature fusion enhancement technology to improve the accuracy of cable intermediate joint detection under different scales, backgrounds, and lighting conditions. The updated parameters are the weight coefficients between nodes at each level of the CNN model.

[0177] The above process is based on the integration of features from multiple defect-free sampled images to optimize the image recognition model. This optimized model can adaptively update and optimize to the influencing factors of the current sampling environment, thereby improving the accuracy of defect recognition.

[0178] In practical implementation, based on the three-dimensional simulation model built using ANSYS software, the electric field distribution of the following seven defect conditions can be analyzed:

[0179] (1) When silicone grease is not applied to the interface between the main insulation and silicone rubber, or when silicone grease is applied unevenly, the electric field will be severely distorted compared to normal conditions. The electric field strength is significantly increased at areas with uneven silicone grease application compared to normal conditions. When the grease is unevenly applied, the electric field distortion is most severe when the air gap is located around the stress cone, and the electric field strength gradually decreases as the air gap moves away from the stress cone. During the cable joint fabrication process, it is crucial to check whether the silicone grease is applied evenly at the stress cone. Furthermore, the thinner the air gap at areas with uneven grease application, the greater the electric field strength around it. When silicone grease contains impurities in the main insulation, the electric field strength at the impurity defects is reduced by 47% compared to when silicone grease is not applied and impurities are present. Numerical calculations show that applying silicone grease can significantly reduce the degree of electric field distortion when potential impurity defects exist at the interface, and has a good optimization effect on the electric field at the impurity defects.

[0180] (2) When semiconducting tape is omitted or insulating tape is misused, the air and insulating tape present between the silicone rubber and the connecting pipe will cause the electric field at that point to increase.

[0181] (3) When the main insulation is scratched, the electric field strength at the scratch location is relatively large when the air gap is 0mm away from the cut position of the outer semiconductor in the axial direction, which is 3.2 times the air breakdown field strength. However, as the scratch location moves away from the cut position of the outer semiconductor, the field strength gradually decreases. Within a distance of 0mm-1.5mm from the cut position, the electric field is significantly distorted compared to normal conditions. The electric field is more prone to distortion when the scratch location is far from the cut position of the outer semiconductor; therefore, careful inspection for scratches should be conducted during joint fabrication.

[0182] (4) When the joint is damp, the electric field distortion is most severe when the water film is near the stress cone, and the electric field of the joint is less affected when the water film is far away from the stress cone.

[0183] (5) When sand particles are mixed into the joint, the electric field is severely distorted from the outer semiconducting chamfer to the stress cone chamfer of the cable. When the impurity is located between the stress cone chamfer and the semiconducting tape, the electric field strength at the impurity gradually decreases.

[0184] (6) When the crimping of the connecting pipe is not smooth and has burrs, the burrs are located close to the main insulation of the cable, and the electric field distortion around the burrs is more severe. Moreover, the electric field strength around the burrs gradually increases as the distance from the main insulation of the cable decreases.

[0185] (7) When applying silicone grease, if the outer semiconducting layer is peeled too long or too short, or if the chamfer is not made, the electric field strength will increase slightly compared to the normal situation, but the effect on the electric field will not be significant.

[0186] As described above, the electric field distribution of a normal, defect-free cable joint should be basically symmetrical, with no significant distortion such as a sudden decrease or drop in electric field intensity. Therefore, the sampled cable joints whose variation trend shows no distortion, as analyzed above, can be used as a standard control group. In this embodiment, the sampled cable joints determined by electric field analysis are used to optimize the defect identification model of the target joint, thereby improving the accuracy of the image recognition model.

[0187] Furthermore, the model training process includes:

[0188] A1. Obtain images of cable intermediate joints in various power scenarios.

[0189] In this embodiment of the invention, cable joint images in various power scenarios can be acquired, such as cable joint images under different lighting conditions, different heights, or different angles in power operation scenarios.

[0190] A2. Perform image preprocessing on the cable intermediate joint image to generate a training image sample set.

[0191] In this embodiment of the invention, the cable intermediate joint image is preprocessed as shown in steps S21-S26 to generate a training image sample set.

[0192] A3. Use the training image sample set as input to train the preset initial intermediate joint defect recognition model, and generate corresponding training indicators based on the training results.

[0193] In this embodiment of the invention, a training image sample set is used as input to a preset initial intermediate joint defect identification model for training, and corresponding training indicators are generated based on the training results.

[0194] A4. Calculate the training loss value between the training metric and the associated standard metric.

[0195] In this embodiment of the invention, the training loss value between the training metric and the associated standard metric is calculated.

[0196] A5. Compare the training loss value with the preset standard loss value.

[0197] In this embodiment of the invention, the training loss value is compared with the preset standard loss value.

[0198] A6. If the training loss value is less than or equal to the preset standard loss value, stop training and generate the target intermediate joint defect identification model.

[0199] In this embodiment of the invention, if the training loss value is less than or equal to the preset standard loss value, training is stopped and a target intermediate joint defect identification model is generated.

[0200] A7. If the training loss value is greater than the preset standard loss value, adjust the network parameters of the preset initial intermediate joint defect identification model according to the preset gradient.

[0201] In this embodiment of the invention, if the training loss value is greater than the preset standard loss value, the network parameters of the preset initial intermediate joint defect identification model are adjusted according to the preset gradient.

[0202] A8. Jump to execute the step of using the training image sample set as input to train the preset initial intermediate joint defect identification model, and generating the corresponding training index according to the training results, until the training loss value is less than or equal to the preset standard loss value, and generate the target intermediate joint defect identification model.

[0203] In this embodiment of the invention, the step of jumping to execute the training of a preset initial intermediate joint defect identification model by inputting a training image sample set, generating corresponding training indicators based on the training results, until the training loss value is less than or equal to the preset standard loss value, and generating the target intermediate joint defect identification model.

[0204] Step 208: Use the target defect recognition model to perform defect detection on the image of the cable intermediate joint to be identified, and output the defect results of the cable intermediate joint to be identified.

[0205] In this embodiment of the invention, a target defect identification model is used to detect defects in the image of the cable intermediate joint to be identified. If no significant defect border is identified in the image of the cable intermediate joint to be identified, it is determined that the cable intermediate joint associated with the image of the cable intermediate joint to be identified is not abnormal. If a significant defect border is identified in the image of the cable intermediate joint to be identified, an associated identification method is selected for analysis based on the type of defect appearing within the significant defect border. For example, for scratches and stains on the main insulation, adjusting the detection threshold can make them clearly visible; for burrs on the connecting pipe, quantitative analysis is performed through edge extrema; for uneven cutting of the main insulation, quantitative analysis is performed through edge variance; thereby determining the target defect of the cable intermediate joint to be identified.

[0206] In this invention, electric field distribution maps of multiple cable joints to be judged are constructed and analyzed. Based on the analysis results, sampled cable joints are determined from the multiple cable joints to be judged. Sampling images of each sampled cable joint are acquired and preprocessed to generate a target reference image set. The target reference image set is used to optimize the target joint defect identification model. The target joint defect identification model is generated through a pre-set model training process. The target defect identification model is used to detect defects in the images of the cable joints to be identified and outputs the defect results of the cable joints to be identified. This invention solves the problem that existing methods for identifying construction defects in power cable joints mostly rely on manual identification and image recognition technology for assistance. However, these methods are easily affected by factors such as light and angle in the image acquisition environment, leading to inaccurate defect identification results. This invention overcomes the problem of inaccurate defect identification results caused by factors such as light and angle in the image acquisition environment in existing technologies.

[0207] Please see Figure 3 , Figure 3 This is a structural block diagram of a defect identification system for cable intermediate joints provided in Embodiment 3 of the present invention.

[0208] This invention provides a defect identification system for cable joints, comprising:

[0209] The analysis module 301 is used to construct and analyze the electric field distribution diagrams of multiple cable intermediate joints to be judged, and to determine the sampling cable intermediate joint from the multiple cable intermediate joints to be judged based on the analysis results.

[0210] Image preprocessing module 302 is used to acquire sampled images of each sampling cable intermediate joint and perform image preprocessing to generate a target comparison image set.

[0211] The model optimization module 303 is used to optimize the target intermediate joint defect identification model by using the target comparison image set as feedback. The target intermediate joint defect identification model is generated by the preset model training module 304.

[0212] The defect detection module 305 is used to perform defect detection on the image of the intermediate joint of the cable to be identified through the target defect recognition model, and output the defect results of the intermediate joint of the cable to be identified.

[0213] Furthermore, the parsing module 301 includes:

[0214] The 3D simulation model submodule is used to establish corresponding 3D simulation models using the structural parameters of multiple cable intermediate joints to be judged.

[0215] The electric field strength submodule is used to call the parametric structure language to mesh the three-dimensional simulation model and calculate the electric field strength of the cable joint to be judged based on the finite element method.

[0216] The electric field distribution map submodule is used to generate electric field distribution maps for multiple cable intermediate joints to be judged based on the electric field strength and the three-dimensional simulation model.

[0217] The symmetry condition comparison submodule is used to compare each electric field distribution diagram with the preset cable axial symmetry condition.

[0218] The sampling cable intermediate joint submodule is used to analyze the electric field distribution diagram if the electric field distribution diagram meets the preset cable axial symmetry condition, and then select the sampling cable intermediate joint without abnormality from multiple cable intermediate joints to be judged.

[0219] Furthermore, the sampling cable intermediate connector submodule includes:

[0220] The two-dimensional field strength variation curve unit is used to extract multiple corresponding two-dimensional field strength variation curves based on each electric field distribution map if the electric field distribution map satisfies the preset cable axial symmetry condition.

[0221] The initial change curve slope value unit is used to calculate multiple corresponding initial change curve slope values ​​based on each two-dimensional field strength change curve.

[0222] The target change curve slope value unit is used to select multiple corresponding target change curve slope values ​​from multiple initial change curve slope values ​​according to a preset interval.

[0223] The target change unit is used to calculate the change between the slope values ​​of two adjacent target change curves, generating multiple target change units.

[0224] The change comparison unit is used to compare the change of each target with the change of a preset standard.

[0225] The judgment unit is used to determine that if all target changes are less than the preset standard changes, the intermediate joint of the cable to be judged associated with the electric field distribution map is not abnormal, and the intermediate joint of the cable to be judged is taken as the intermediate joint of the sampling cable.

[0226] Furthermore, the image preprocessing module 302 includes:

[0227] The clustering set submodule is used to acquire the sampled images of each sampled cable intermediate joint and perform clustering to generate multiple cluster sets.

[0228] The initial pixel feature data submodule is used to extract features from multiple sampled images associated with each cluster set and generate multiple initial pixel feature data.

[0229] The first fusion feature data submodule is used to select multiple initial pixel feature data that have similar features from multiple initial pixel feature data as the first fusion feature data.

[0230] The target pixel feature data submodule is used to perform imaging and denoising operations on initial pixel feature data that do not have similar features, and generate corresponding target pixel feature data.

[0231] The second fusion feature data submodule is used to calculate the average pixel value corresponding to the target pixel feature data and use the average pixel value as the second fusion feature data.

[0232] The target comparison image set submodule is used to stitch together the first fused feature data and the second fused feature data to generate a target comparison image set.

[0233] Furthermore, the model training module 304 includes:

[0234] The cable joint image submodule is used to acquire images of cable joints in various power scenarios.

[0235] The training image sample set submodule is used to preprocess images of cable joints to generate a training image sample set.

[0236] The training metrics submodule is used to train the preset initial intermediate joint defect recognition model by inputting the training image sample set, and to generate corresponding training metrics based on the training results.

[0237] The training loss value submodule is used to calculate the training loss value between the training metric and the associated standard metric.

[0238] The loss comparison submodule is used to compare the training loss value with the preset standard loss value.

[0239] The first data processing submodule is used to stop training and generate a target intermediate joint defect identification model if the training loss value is less than or equal to the preset standard loss value.

[0240] The second data processing submodule is used to adjust the network parameters of the preset initial intermediate joint defect identification model according to the preset gradient if the training loss value is greater than the preset standard loss value.

[0241] The jump rotor module is used to jump to the execution of the steps of training the preset initial intermediate joint defect identification model by inputting the training image sample set, generating the corresponding training index according to the training results, until the training loss value is less than or equal to the preset standard loss value, and generating the target intermediate joint defect identification model.

[0242] In this invention, electric field distribution maps of multiple cable joints to be judged are constructed and analyzed. Based on the analysis results, sampled cable joints are determined from the multiple cable joints to be judged. Sampling images of each sampled cable joint are acquired and preprocessed to generate a target reference image set. The target reference image set is used to optimize the target joint defect identification model. The target joint defect identification model is generated through a pre-set model training process. The target defect identification model is used to detect defects in the images of the cable joints to be identified and outputs the defect results of the cable joints to be identified. This invention solves the problem that existing methods for identifying construction defects in power cable joints mostly rely on manual identification and image recognition technology for assistance. However, these methods are easily affected by factors such as light and angle in the image acquisition environment, leading to inaccurate defect identification results. This invention overcomes the problem of inaccurate defect identification results caused by factors such as light and angle in the image acquisition environment in existing technologies.

[0243] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0244] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0245] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for defect identification of cable joints, characterized in that, include: Construct and analyze the electric field distribution diagrams of multiple cable intermediate joints to be judged, and determine the sampling cable intermediate joints from the multiple cable intermediate joints to be judged based on the analysis results; Acquire sampling images of each of the sampling cable intermediate joints and perform image preprocessing to generate a target comparison image set; The step of acquiring sampled images of each of the sampled cable intermediate joints and performing image preprocessing to generate a target comparison image set includes: Acquire sampling images of each of the sampling cable intermediate joints and perform clustering to generate multiple cluster sets; Feature extraction is performed on the multiple sampled images associated with each of the cluster sets to generate multiple initial pixel feature data; Multiple initial pixel feature data that have similar features are selected from the multiple initial pixel feature data as the first fused feature data; Imaging and denoising operations are performed on the initial pixel feature data that do not have similar features to generate corresponding target pixel feature data; Calculate the average pixel value corresponding to the target pixel feature data, and use the average pixel value as the second fused feature data; The first fused feature data and the second fused feature data are stitched together to generate a target comparison image set; The target intermediate joint defect identification model is optimized using the target comparison image set, wherein the target intermediate joint defect identification model is generated through a preset model training process; The target defect identification model is used to detect defects in the image of the intermediate joint of the cable to be identified, and the defect results of the intermediate joint of the cable to be identified are output.

2. The defect identification method for cable intermediate joints according to claim 1, characterized in that, The step of constructing and analyzing the electric field distribution maps of multiple cable joints to be judged, and determining the sampling cable joint from the multiple cable joints to be judged based on the analysis results, includes: A three-dimensional simulation model was established using the structural parameters of multiple cable intermediate joints to be judged. The three-dimensional simulation model is meshed using a parametric structure language, and the electric field strength of the intermediate joint of the cable to be judged is calculated based on the finite element method. Based on the electric field strength and the three-dimensional simulation model, multiple electric field distribution maps corresponding to the intermediate joints of the cables to be judged are generated. Compare each of the described electric field distribution diagrams with the preset cable axial symmetry condition; If the electric field distribution map satisfies the preset cable axial symmetry condition, then the electric field distribution map is analyzed, and sampled cable intermediate joints without abnormalities are selected from multiple intermediate joints of the cables to be judged.

3. The defect identification method for cable intermediate joints according to claim 2, characterized in that, The step of analyzing the electric field distribution map and selecting abnormal sampled cable joints from multiple cable joints to be judged if the electric field distribution map satisfies the preset cable axial symmetry condition includes: If the electric field distribution map satisfies the preset cable axial symmetry condition, then multiple corresponding two-dimensional field strength variation curves are extracted based on each electric field distribution map; Calculate the slope values ​​of multiple corresponding initial change curves based on the two-dimensional field strength change curves described above. Select multiple corresponding target change curve slope values ​​from multiple initial change curve slope values ​​according to a preset interval; Calculate the change between the slope values ​​of two adjacent target change curves to generate multiple target change values; Compare the changes in each target with the changes in the preset standard; If all the target changes are less than the preset standard changes, then the intermediate joint of the cable to be judged associated with the electric field distribution map is determined to be normal, and the intermediate joint of the cable to be judged is taken as the intermediate joint of the sampling cable.

4. The defect identification method for cable intermediate joints according to claim 1, characterized in that, The model training process includes: Acquire images of cable joints in various power scenarios; The cable joint image is preprocessed to generate a training image sample set; The training image sample set is used as input to a preset initial intermediate joint defect identification model for training, and corresponding training indicators are generated based on the training results. Calculate the training loss value between the training metric and the associated standard metric; Compare the training loss value with the preset standard loss value; If the training loss value is less than or equal to the preset standard loss value, then training is stopped and a target intermediate joint defect identification model is generated. If the training loss value is greater than the preset standard loss value, then the network parameters of the preset initial intermediate joint defect identification model are adjusted according to the preset gradient. The process jumps to the step of training the preset initial intermediate joint defect identification model using the training image sample set as input, generating corresponding training indicators based on the training results, until the training loss value is less than or equal to the preset standard loss value, and then generates the target intermediate joint defect identification model.

5. A defect identification system for cable joints, characterized in that, include: The analysis module is used to construct and analyze the electric field distribution diagrams of multiple cable intermediate joints to be judged, and to determine the sampling cable intermediate joint from the multiple cable intermediate joints to be judged based on the analysis results. The image preprocessing module is used to acquire sampled images of each of the sampling cable intermediate joints and perform image preprocessing to generate a target comparison image set; The image preprocessing module includes: The clustering set submodule is used to acquire the sampled images of each of the sampled cable intermediate joints and perform clustering to generate multiple cluster sets; The initial pixel feature data submodule is used to extract features from multiple sampled images associated with each of the cluster sets to generate multiple initial pixel feature data. The first fusion feature data submodule is used to select multiple initial pixel feature data that have similar features from multiple initial pixel feature data as the first fusion feature data; The target pixel feature data submodule is used to perform imaging and denoising operations on the initial pixel feature data that do not have similar features, and generate corresponding target pixel feature data. The second fusion feature data submodule is used to calculate the average pixel value corresponding to the target pixel feature data and use the average pixel value as the second fusion feature data. The target comparison image set submodule is used to stitch together the first fused feature data and the second fused feature data to generate a target comparison image set; The model optimization module is used to optimize the target intermediate joint defect identification model using the target comparison image set, wherein the target intermediate joint defect identification model is generated by a preset model training module; The defect detection module is used to perform defect detection on the image of the intermediate joint of the cable to be identified through the target defect identification model, and output the defect results of the intermediate joint of the cable to be identified.

6. The defect identification system for cable joints according to claim 5, characterized in that, The parsing module includes: The three-dimensional simulation model submodule is used to establish a corresponding three-dimensional simulation model using the structural parameters of multiple cable intermediate joints to be judged. The electric field strength submodule is used to call the parametric structure language to mesh the three-dimensional simulation model and calculate the electric field strength of the intermediate joint of the cable to be judged based on the finite element method. The electric field distribution map submodule is used to generate multiple electric field distribution maps corresponding to the intermediate joints of the cables to be judged based on the electric field strength and the three-dimensional simulation model. The symmetry condition comparison submodule is used to compare each electric field distribution diagram with the preset cable axial symmetry condition. The sampling cable intermediate joint submodule is used to analyze the electric field distribution map if the electric field distribution map meets the preset cable axial symmetry condition, and then select the sampling cable intermediate joints without abnormalities from multiple cable intermediate joints to be judged.

7. The defect identification system for cable joints according to claim 6, characterized in that, The sampling cable intermediate connector submodule includes: The two-dimensional field strength variation curve unit is used to extract multiple corresponding two-dimensional field strength variation curves based on each electric field distribution map if the electric field distribution map satisfies the preset cable axial symmetry condition. The initial change curve slope value unit is used to calculate multiple corresponding initial change curve slope values ​​based on each of the two-dimensional field strength change curves. The target change curve slope value unit is used to select multiple corresponding target change curve slope values ​​from multiple initial change curve slope values ​​according to a preset interval; The target change unit is used to calculate the change between the slope values ​​of two adjacent target change curves and generate multiple target change units. The change comparison unit is used to compare the change of each target with the change of a preset standard. The determination unit is used to determine that the intermediate joint of the cable to be judged associated with the electric field distribution map is not abnormal if all the target changes are less than the preset standard changes, and to use the intermediate joint of the cable to be judged as the intermediate joint of the sampling cable.

8. The defect identification system for cable joints according to claim 5, characterized in that, The model training module includes: The cable joint image submodule is used to acquire images of cable joints in various power scenarios. The training image sample set submodule is used to perform image preprocessing on the cable intermediate joint image to generate a training image sample set. The training index submodule is used to train the preset initial intermediate joint defect identification model by inputting the training image sample set, and to generate corresponding training indexes based on the training results. The training loss value submodule is used to calculate the training loss value between the training metric and the associated standard metric. The loss value comparison submodule is used to compare the training loss value with the preset standard loss value; The first data processing submodule is used to stop training and generate a target intermediate joint defect identification model if the training loss value is less than or equal to the preset standard loss value. The second data processing submodule is used to adjust the network parameters of the preset initial intermediate joint defect identification model according to a preset gradient if the training loss value is greater than the preset standard loss value. The jump-rotor module is used to jump to execute the steps of training the preset initial intermediate joint defect identification model by inputting the training image sample set, generating corresponding training indicators based on the training results, until the training loss value is less than or equal to the preset standard loss value, and generating the target intermediate joint defect identification model.

Images