Power grid equipment defect identification system based on visual large model

Abstract

The invention discloses a power grid equipment defect identification system based on a visual large model, and the system comprises the following parts: an image collection module which employs a high-definition industrial camera and an infrared thermal imager to collect an appearance image and a temperature image of power grid equipment; the image preprocessing module is used for preprocessing the appearance image and the temperature image; the defect recognition module is used for recognizing defects, including cracks, corrosion, deformation and discharge traces, on the power grid equipment through a deep learning model and image processing according to the appearance image and the temperature image; and the defect analysis report generation module is used for scoring the defects and generating a defect analysis report according to a scoring result, and the defect analysis report comprises defect types and defect severity. According to the method, more detailed and accurate power grid equipment image information is acquired, so that the accuracy of power grid equipment defect identification is improved.

Description

Technical Field

[0001] The present invention relates to the field of power grid equipment defect recognition, and in particular to a power grid equipment defect recognition system based on a large visual model. Background Art

[0002] During the operation of power grid equipment, various defects are prone to appear on the surface of the equipment due to long-term exposure to complex environmental conditions (such as high temperature, high humidity, wind and sand, rain and snow, etc.). If these defects are not discovered and repaired in time, they may lead to equipment performance degradation, energy efficiency loss and even serious power grid accidents. However, traditional detection methods can only simply determine whether there are defects in the equipment, but cannot accurately classify the defect type or assess its severity. Maintenance personnel cannot formulate priority treatment plans based on the specific defect information of the equipment. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a power grid equipment defect recognition system based on a large visual model.

[0004] The present invention provides a power grid equipment defect recognition system based on a visual large model, which is characterized by comprising the following:

[0005] Image acquisition module: uses high-definition industrial cameras and infrared thermal imagers to collect appearance images and temperature images of power grid equipment respectively;

[0006] Image preprocessing module: preprocessing the appearance image and the temperature image;

[0007] Defect recognition module: based on the appearance image and the temperature image, through deep learning model and image processing, identify defects on the power grid equipment, including cracks, rust, deformation and discharge marks;

[0008] Defect analysis report generation module: Score defects and generate defect analysis reports based on the scoring results, including defect type and defect severity.

[0009] Preferably, the image preprocessing module includes:

[0010] De-noising processing unit: using Gaussian filtering to perform denoising processing on the appearance image and the temperature image to remove noise and interference in the image;

[0011] Enhancement processing unit: performing enhancement processing on the appearance image and the temperature image through an adaptive enhancement algorithm;

[0012] A resolution enhancement unit: using an interpolation algorithm to enhance the resolution of the appearance image and the temperature image;

[0013] Distortion correction unit: performs distortion correction on the appearance image and the temperature image through camera calibration and distortion correction algorithm to eliminate image distortion caused by camera lens distortion;

[0014] A normalization processing unit: performing normalization processing on the appearance image and the temperature image;

[0015] Adjustment function setting unit: used to set the parameter adjustment function of image preprocessing. Users can adjust the parameters of image preprocessing according to actual needs.

[0016] Preferably, the defect identification module comprises:

[0017] Surface defect feature recognition submodule: uses a convolutional neural network to extract features from the appearance image and identify surface defect features;

[0018] Temperature defect feature recognition submodule: Combined with the temperature image collected by the infrared thermal imager, it detects the abnormal temperature distribution on the surface of the equipment and identifies the temperature defect features;

[0019] Data fusion submodule: The surface defect features and the temperature defect features are fused through a multimodal data fusion algorithm, and the types of defects, including cracks, rust, deformation and discharge marks, are identified based on the fused defect features.

[0020] Preferably, the surface defect feature recognition submodule includes:

[0021] Feature map generation unit: uses a convolutional neural network to extract features from the appearance image and generate a multi-level feature map;

[0022] Abnormal area positioning unit: analyzes the feature map layer by layer to locate the defective area

[0023] Defect area segmentation unit: Combine edge detection algorithm and segmentation network to segment the surface defect area;

[0024] Classification unit: Classify the surface defect area to obtain the surface defect type, including cracks, rust, deformation and discharge marks.

[0025] Preferably, the temperature defect feature identification submodule includes:

[0026] Thermal map segmentation unit: Use image segmentation algorithm to divide the image into different temperature areas;

[0027] Abnormal area detection unit: used to preset a temperature threshold and mark abnormal areas with a temperature greater than the threshold;

[0028] Feature extraction unit: extracts features of the detected abnormal area, including maximum temperature, temperature gradient, area size, shape and temperature difference distribution;

[0029] Temperature defect classification unit: classifies the extracted temperature defect features to obtain defect types, including hot spots, overheating areas, and poor heat dissipation.

[0030] Preferably, the data fusion submodule includes:

[0031] Feature alignment unit: principal component analysis is used to align appearance features and temperature features;

[0032] Feature fusion unit: Use weighted fusion to jointly analyze appearance features and temperature features to generate comprehensive defect features;

[0033] Regularization processing unit: performing regularization processing on the comprehensive defect features;

[0034] Feature matching unit: matches the predefined defect template library according to the comprehensive defect feature to identify the defect type.

[0035] Preferably, the defect report generating module comprises:

[0036] Scoring calculation unit: using a fully connected neural network to map the comprehensive defect characteristics into a single scoring value, wherein the scoring value ranges from 0 to 100 points, 0 represents no defect, and 100 represents the most serious defect;

[0037] Grading unit: Based on the scoring results, defects are classified into different severity levels, including minor, moderate and severe;

[0038] Defect analysis report generation unit: used to generate defect analysis reports, including defect types, severity scores, and related images;

[0039] Visualization interface setting unit: used to set the visualization interface, display the defect distribution map of the power grid equipment, and use different colors to mark defects of different severity;

[0040] Communication unit: used to send the defect analysis report to relevant managers or maintenance teams.

[0041] The present invention provides a power grid equipment defect recognition system based on a visual large model. Through the combined use of a high-definition industrial camera and an infrared thermal imager, more detailed and accurate image information of the power grid equipment can be obtained, thereby improving the accuracy of identifying power grid equipment defects, especially for those tiny defects or temperature anomalies that are difficult to detect with the naked eye. By automatically acquiring images, preprocessing images, identifying defects and generating analysis reports, the need for manual intervention is greatly reduced, work efficiency is improved, and a wider range of detection work can be completed in a shorter time. The detection work can be completed without contacting the equipment, thereby reducing the safety risks of the operators. By scoring defects and generating detailed analysis reports, it can help power companies better plan maintenance work, give priority to serious defects, reasonably arrange resources, reduce the number of unnecessary power outages for maintenance, and improve power supply reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 An architecture diagram of a power grid equipment defect recognition system based on a visual large model provided by the present invention;

[0044] Figure 2 The architecture diagram of the defect identification module provided by the present invention;

[0045] Figure 3 A unit structure diagram of a surface defect feature recognition submodule in a defect recognition module provided by the present invention;

[0046] Figure 4 A unit structure diagram of a temperature defect feature recognition submodule in a defect recognition module provided by the present invention;

[0047] Figure 5 This is a unit structure diagram of the data fusion submodule in the defect identification module provided by the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0050] refer to Figure 1 The present invention provides a power grid equipment defect recognition system based on a visual large model, which is characterized by comprising the following:

[0051] Image acquisition module 11: uses a high-definition industrial camera and an infrared thermal imager to respectively collect appearance images and temperature images of power grid equipment, the power grid equipment including transmission lines, transformers, distribution transformers, and box-type substations;

[0052] Image preprocessing module 12: preprocessing the appearance image and the temperature image;

[0053] Defect recognition module 13: Based on the appearance image and temperature image, the defects on the power grid equipment, including cracks, rust, deformation and discharge marks, are identified through deep learning models and image processing;

[0054] Defect analysis report generation module 14: Scores defects and generates a defect analysis report based on the scoring results, including defect type and defect severity.

[0055] The power grid equipment defect recognition system based on visual large model provided by the present invention can obtain more detailed and accurate image information of power grid equipment through the combined use of high-definition industrial cameras and infrared thermal imagers, thereby improving the accuracy of power grid equipment defect recognition, especially for those tiny defects or temperature anomalies that are difficult to detect with the naked eye; by automatically acquiring images, preprocessing images, identifying defects and generating analysis reports, the need for manual intervention is greatly reduced, work efficiency is improved, and a wider range of detection work can be completed in a shorter time; and the detection work can be completed without contacting the equipment, thereby reducing the safety risks of operators; by scoring defects and generating detailed analysis reports, it can help power companies better plan maintenance work, give priority to serious defects, reasonably arrange resources, reduce the number of unnecessary power outages and maintenance, and improve power supply reliability.

[0056] In a preferred embodiment, the image preprocessing module 12 includes,

[0057] Denoising unit: Use Gaussian filtering to perform denoising on the appearance image and temperature image to remove noise and interference in the image;

[0058] Enhancement processing unit: enhances the appearance image and temperature image through adaptive enhancement algorithm to improve the clarity and contrast of the image;

[0059] Resolution enhancement unit: uses interpolation algorithms to enhance the resolution of appearance images and temperature images to ensure the clarity and accuracy of image details;

[0060] Distortion correction unit: performs distortion correction on the appearance image and temperature image through camera calibration and distortion correction algorithm to eliminate image distortion caused by camera lens distortion;

[0061] Normalization processing unit: normalizes the appearance image and temperature image to ensure that the image data is within a uniform scale range;

[0062] Adjustment function setting unit: used to set the parameter adjustment function of image preprocessing. Users can adjust the parameters of image preprocessing according to actual needs.

[0063] refer to Figure 2 In a preferred embodiment, the defect identification module 13 includes:

[0064] Surface defect feature recognition submodule 131: uses a convolutional neural network to extract features from the appearance image and recognize surface defect features;

[0065] Temperature defect feature recognition submodule 132: detects abnormal temperature distribution on the surface of the equipment and identifies temperature defect features in combination with the temperature image collected by the infrared thermal imager;

[0066] Data fusion submodule 133: Surface defect features and temperature defect features are fused through a multimodal data fusion algorithm, and the types of defects, including cracks, rust, deformation and discharge marks, are identified based on the fused defect features.

[0067] refer to Figure 3 In a preferred embodiment, the surface defect feature recognition submodule 131 includes:

[0068] Feature map generating unit 1311: extracting features from the appearance image using a convolutional neural network to generate a multi-level feature map;

[0069] Abnormal area positioning unit 1312: Analyze the feature map layer by layer to locate the defective area;

[0070] Defect area segmentation unit 1313: combining edge detection algorithm and segmentation network to segment the surface defect area;

[0071] Classification unit 1314: classifies the defect area to obtain defect types, including cracks, rust, deformation and discharge marks.

[0072] refer to Figure 4 In a preferred embodiment, the temperature defect feature identification submodule 132 includes:

[0073] The thermal image segmentation unit 1321: uses an image segmentation algorithm such as threshold-based segmentation, region growing method or edge detection-based segmentation to divide the image into different temperature regions;

[0074] Abnormal area detection unit 1322: used to preset a temperature threshold and mark abnormal areas with a temperature greater than the temperature threshold;

[0075] Feature extraction unit 1323: extracts features of the detected abnormal area, including maximum temperature, temperature gradient, area size, shape and temperature difference distribution;

[0076] The temperature defect classification unit 1324 classifies the extracted temperature defect features to obtain defect types, including hot spots, overheated areas, and poor heat dissipation.

[0077] refer to Figure 5 In a preferred embodiment, the data fusion submodule 133 includes

[0078] Feature alignment unit 1331: aligning appearance features and temperature features using principal component analysis;

[0079] Feature fusion unit 1332: using weighted fusion to jointly analyze appearance features and temperature features to generate comprehensive defect features;

[0080] Specifically, by analyzing the contribution of appearance and temperature images to defect recognition, weights are dynamically assigned, for example, the weight of appearance features is 0.6, the weight of temperature features is 0.4, and the two types of features are weighted and summed in the feature space. This method can effectively balance the impact of the two modes on the final result and avoid over-emphasis or neglect of a certain feature mode.

[0081] Regularization processing unit 1333: performs regularization processing on the comprehensive defect features to ensure consistency and accuracy between the features.

[0082] Feature matching unit 1334: matches the pre-defined defect template library according to the comprehensive defect features to identify the type of defect.

[0083] In a preferred embodiment, the defect report generating module 14 includes:

[0084] Scoring calculation unit: A fully connected neural network is used to map the comprehensive defect features into a single scoring value, where the scoring value ranges from 0 to 100, with 0 indicating no defect and 100 indicating the most serious defect;

[0085] Grading unit: Based on the scoring results, defects are classified into different severity levels, including minor, moderate and severe

[0086] Specifically, defects with a score of 0-30 are minor; defects with a score of 31-70 are moderate; and defects with a score of 71-100 are severe.

[0087] Defect analysis report generation unit: used to generate defect analysis reports, including defect types, severity scores, and related images;

[0088] Visualization interface setting unit: used to set the visualization interface, display the defect distribution map of the power grid equipment, and use different colors to mark defects of different severity, so that users can intuitively understand the overall status of the equipment;

[0089] Specifically, yellow marks minor defects; blue marks medium defects; and red marks severe defects. Users can zoom in and out of the distribution map to view the overall picture or details. Appearance images and temperature images are superimposed on the distribution map to intuitively display the defect distribution and thermal anomaly areas of the equipment. The score value and defect type are displayed in each defect area.

[0090] Communication unit: used to send defect analysis reports to relevant managers or maintenance teams.

[0091] When a serious defect is detected, that is, a score higher than 70 points, the system can send real-time alerts via email, SMS or enterprise management system, including key information such as defect type, score, equipment location, etc.; the report supports multiple formats such as PDF, HTML or Excel, and users can customize the report content and layout to meet different needs.

Claims

1. A power grid equipment defect recognition system based on a visual large model, characterized in that: These include: Image acquisition module: uses high-definition industrial cameras and infrared thermal imagers to collect appearance images and temperature images of power grid equipment respectively; Image preprocessing module: preprocessing the appearance image and the temperature image; Defect recognition module: based on the appearance image and the temperature image, through deep learning model and image processing, identify defects on the power grid equipment, including cracks, rust, deformation and discharge marks; Defect analysis report generation module: Score defects and generate defect analysis reports based on the scoring results, including defect type and defect severity.

2. A power grid equipment defect recognition system based on a visual large model according to claim 1, characterized in that: The image preprocessing module includes: De-noising processing unit: using Gaussian filtering to perform denoising processing on the appearance image and the temperature image to remove noise and interference in the image; Enhancement processing unit: performing enhancement processing on the appearance image and the temperature image through an adaptive enhancement algorithm; A resolution enhancement unit: using an interpolation algorithm to enhance the resolution of the appearance image and the temperature image; Distortion correction unit: performs distortion correction on the appearance image and the temperature image through camera calibration and distortion correction algorithm to eliminate image distortion caused by camera lens distortion; A normalization processing unit: performing normalization processing on the appearance image and the temperature image; Adjustment function setting unit: used to set the parameter adjustment function of image preprocessing. Users can adjust the parameters of image preprocessing according to actual needs.

3. A power grid equipment defect recognition system based on a visual large model according to claim 1, characterized in that: The defect identification module includes: Surface defect feature recognition submodule: uses a convolutional neural network to extract features from the appearance image and identify surface defect features; Temperature defect feature recognition submodule: Combined with the temperature image collected by the infrared thermal imager, it detects the abnormal temperature distribution on the surface of the equipment and identifies the temperature defect features; Data fusion submodule: The surface defect features and the temperature defect features are fused through a multimodal data fusion algorithm, and the types of defects, including cracks, rust, deformation and discharge marks, are identified based on the fused defect features.

4. A power grid equipment defect recognition system based on a visual large model according to claim 3, characterized in that: The surface defect feature recognition submodule includes: Feature map generation unit: uses a convolutional neural network to extract features from the appearance image and generate a multi-level feature map; Abnormal area positioning unit: analyzes the feature map layer by layer to locate the defective area Defect area segmentation unit: Combine edge detection algorithm and segmentation network to segment the surface defect area; Classification unit: Classify the surface defect area to obtain the surface defect type, including cracks, rust, deformation and discharge marks.

5. A power grid equipment defect recognition system based on visual large model according to claim 4, characterized in that: The temperature defect feature recognition submodule includes: Thermal map segmentation unit: Use image segmentation algorithm to divide the image into different temperature areas; Abnormal area detection unit: used to preset a temperature threshold and mark abnormal areas with a temperature greater than the threshold; Feature extraction unit: extracts features of the detected abnormal area, including maximum temperature, temperature gradient, area size, shape and temperature difference distribution; Temperature defect classification unit: classifies the extracted temperature defect features to obtain defect types, including hot spots, overheating areas, and poor heat dissipation.

6. A power grid equipment defect recognition system based on visual large model according to claim 5, characterized in that: The data fusion submodule includes: Feature alignment unit: uses principal component analysis to align appearance features and temperature features; Feature fusion unit: Use weighted fusion to jointly analyze appearance features and temperature features to generate comprehensive defect features; Regularization processing unit: performing regularization processing on the comprehensive defect features; Feature matching unit: matches the predefined defect template library according to the comprehensive defect feature to identify the defect type.

7. A power grid equipment defect recognition system based on a visual large model according to claim 6, characterized in that: The defect report generating module includes: Scoring calculation unit: using a fully connected neural network to map the comprehensive defect characteristics into a single scoring value, wherein the scoring value ranges from 0 to 100, where 0 represents no defect and 100 represents the most serious defect; Grading unit: Based on the scoring results, defects are classified into different severity levels, including minor, moderate and severe; Defect analysis report generation unit: used to generate defect analysis reports, including defect types, severity scores, and related images; Visualization interface setting unit: used to set the visualization interface, display the defect distribution map of the power grid equipment, and use different colors to mark defects of different severity; Communication unit: used to send the defect analysis report to relevant managers or maintenance teams.

Images