Electrical equipment defect detection system based on infrared three-dimensional imaging

Abstract

The invention discloses a power equipment defect detection system based on infrared three-dimensional imaging, and relates to the technical field of equipment testing, and the system comprises an imaging slicing module which carries out the slicing processing of an infrared three-dimensional image of target power equipment, and obtains a target infrared image; the layering processing module introduces a predetermined layering mechanism to generate a first graph layer with a first label; the hierarchical detection module matches a corresponding defect detection unit based on the first label to obtain a first defect detection result; the significant defect determination module identifies the most significant defect according to a significance strategy and marks the most significant defect as a first-level defect; the defect degree analysis module analyzes the first-level defect degree in combination with the detection result to obtain a first-level defect index; and when the defect index does not meet the preset threshold value, the defect handling module calls an emergency scheme for handling. Therefore, the technical effects of improving the defect detection accuracy and adaptability and improving the detection efficiency are achieved.

Description

Technical Field

[0001] The invention relates to the technical field of equipment testing, and in particular to an electric power equipment defect detection system based on infrared three-dimensional imaging. Background Art

[0002] With the development of the power industry, the reliability and safety of power equipment have gradually become the focus of operation and maintenance management. Once a failure occurs in high-voltage equipment, power transmission and distribution equipment, etc., it may cause significant economic losses and safety hazards. At present, infrared imaging technology is mainly used for temperature anomaly monitoring in defect detection of power equipment. Combined with infrared image analysis technology, it can preliminarily identify hot spots and judge overheating or aging problems of equipment, but it is difficult to provide comprehensive equipment defect information. There are technical problems such as difficulty in dealing with complex defects, low recognition accuracy and detection efficiency. Summary of the invention

[0003] The present invention provides an electric power equipment defect detection system based on infrared three-dimensional imaging to solve the technical problems in the prior art that it is difficult to deal with complex defects and that the recognition accuracy and detection efficiency are low, thereby achieving the technical effect of improving the accuracy and adaptability of defect detection and improving the detection efficiency.

[0004] The power equipment defect detection system based on infrared three-dimensional imaging provided by the present invention comprises:

[0005] An imaging slicing module, wherein the imaging slicing module is used to slice the target infrared three-dimensional imaging to obtain the target infrared image, wherein the target infrared three-dimensional imaging refers to the infrared three-dimensional imaging of the target power equipment;

[0006] A hierarchical processing module, the hierarchical processing module is used to introduce a predetermined hierarchical mechanism to perform hierarchical processing on the target infrared image to obtain a target hierarchical result, wherein the target hierarchical result includes a first layer marked with a first label;

[0007] A hierarchical detection module, the hierarchical detection module is used to match a first defect detection unit corresponding to the first layer based on the first label, and obtain a first defect detection result through the first defect detection unit;

[0008] a significant defect determination module, the significant defect determination module being used to determine a first most significant defect in the first defect detection result according to a graph significance strategy, and to use the first most significant defect as a primary defect;

[0009] a defect degree analysis module, the defect degree analysis module being used to perform defect degree analysis on the primary defect in combination with the first defect detection result to obtain a primary defect index, and use the primary defect index as a first defect index of the first layer;

[0010] A defect handling module is used to call a first predetermined emergency plan to handle the defect of the target power equipment when the first defect index does not meet the first predetermined defect threshold corresponding to the first tag.

[0011] In a feasible implementation, the execution steps of the system further include:

[0012] Obtaining historical defect records of similar electrical equipment to the target electrical equipment;

[0013] Extract a first historical record from the historical defect record, wherein the first historical record includes a first historical defect type, and the first historical defect type has an identification of a first defect infrared feature; perform cluster analysis on the first historical defect type using the first defect infrared feature as a clustering constraint to obtain a defect clustering library; generate multiple defect clustering layers based on multiple defect clustering clusters in the defect clustering library, and set the predetermined hierarchical mechanism based on the multiple defect clustering layers.

[0014] In a feasible implementation, the execution steps of the system further include:

[0015] Extract a second historical record from the historical defect record, wherein the second historical defect type includes a second historical defect type, wherein the second historical defect type is different from the first historical defect type, and the second historical defect type has an identification of a second defect infrared feature; obtain a first infrared indicator, and match the first defect infrared feature with the second defect infrared feature, respectively, to obtain a first indicator parameter and a second indicator parameter corresponding to the first infrared indicator, respectively; calculate a first indicator deviation rate based on the first indicator parameter and the second indicator parameter; determine whether the first indicator deviation rate is within a predetermined deviation threshold; if so, mark the first infrared indicator similarly, and if not, mark the first infrared indicator differently to obtain a first marking result; statistically analyze the first marking result to obtain a first similarity between the first historical defect type and the second historical defect type; cluster the defect types according to the first similarity to obtain the defect clustering library.

[0016] In a feasible implementation, the execution steps of the system further include:

[0017] The first defect infrared feature and the first historical defect type form a first data group; the second defect infrared feature and the second historical defect type form a second data group; the first data group is subjected to supervised learning training and tested through the second data group to obtain the first defect detection unit.

[0018] In a feasible implementation, the execution steps of the system further include:

[0019] A first defect visual graph is generated based on the first defect monitoring result, the first defect visual graph includes multiple defects with defect range identifications; a first defect among the multiple defects with defect range identifications is extracted, the first defect corresponds to a first defect range; features of the first defect range are collected according to the graph significance strategy to obtain first defect feature parameters; a first defect feature parameter set of the first defect visual graph is constructed based on the first defect feature parameters; the first defect feature parameter set is screened to obtain a first most significant defect feature parameter, and the defect corresponding to the first most significant defect feature parameter is recorded as the first most significant defect.

[0020] In a feasible implementation, the graph saliency strategy includes a graph color saliency strategy and a graph structure saliency strategy, including:

[0021] According to the image color significance strategy, first color characteristic parameters of the first defect range are obtained, wherein the first color characteristic parameters include a first red parameter, a first green parameter and a first blue parameter; according to the image structure significance strategy, first structural characteristic parameters of the first defect range are obtained; and the first defect characteristic parameters are composed based on the first red parameter, the first green parameter, the first blue parameter and the first structural characteristic parameter.

[0022] In a feasible implementation, the execution steps of the system further include:

[0023] Taking the position of the primary defect as the defect center, determining the secondary defect threshold of the defect center in the first defect visual graph; randomly sampling the defects in the secondary defect threshold based on the random sampling principle to form secondary defects; weighting the defect characteristic parameters of each defect in the secondary defects to obtain a feedback calibration coefficient; calibrating the first defect characteristic parameter of the primary defect according to the feedback calibration coefficient to obtain a first calibration result; and using the first calibration result as the primary defect index.

[0024] In a feasible implementation, a defect degree evaluation function is introduced to analyze the first calibration result to obtain the first-level defect index, wherein the first calibration result includes a first red calibration parameter, a first green calibration parameter, a first blue calibration parameter and a first structural feature calibration parameter.

[0025] The present invention discloses a power equipment defect detection system based on infrared three-dimensional imaging, comprising: an imaging slicing module for slicing infrared three-dimensional imaging of a target power equipment to generate a target infrared image; a hierarchical processing module for hierarchical processing of the target infrared image through a predetermined hierarchical mechanism to obtain a hierarchical result including a first layer with a first label identification; a hierarchical detection module for matching a first defect detection unit corresponding to the first layer based on the first label, and generating a first defect detection result through the detection unit; a significant defect determination module for determining the most significant defect in the first defect detection result based on a graph significance strategy, which is called a first-level defect; a defect degree analysis module for analyzing the defect degree of the first-level defect in combination with the first defect detection result, generating a first-level defect index, and using it as the defect index of the first layer; a defect handling module for calling a corresponding predetermined emergency plan to handle the defect of the target power equipment when the first defect index is lower than a predetermined defect threshold corresponding to the first label. The power equipment defect detection system based on infrared three-dimensional imaging disclosed by the present invention solves the technical problems of difficulty in dealing with complex defects, low recognition accuracy and detection efficiency, and achieves the technical effects of improving the accuracy and adaptability of defect detection and improving detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a structural schematic diagram of the power equipment defect detection system based on infrared three-dimensional imaging of the present invention;

[0027] Figure 2 It is a flow chart of the steps executed in the electric power equipment defect detection system based on infrared three-dimensional imaging of the present invention.

[0028] Explanation of the reference numerals: imaging slicing module 11 , hierarchical processing module 12 , hierarchical detection module 13 , significant defect determination module 14 , defect degree analysis module 15 , defect handling module 16 . DETAILED DESCRIPTION

[0029] The technical solution provided in the embodiments of the present invention is to solve the technical problems of the prior art that it is difficult to deal with complex defects, and the recognition accuracy and detection efficiency are low. The overall idea adopted is as follows:

[0030] First, the imaging slicing module slices the target infrared three-dimensional imaging to obtain the target infrared image, which is the image of the target power equipment; then, the hierarchical processing module introduces a predetermined hierarchical mechanism to perform hierarchical processing on the target infrared image to obtain a target hierarchical result, which includes a first layer identifier with a first label; then, the hierarchical detection module matches the first defect detection unit corresponding to the layer based on the first label, and obtains the first defect detection result through the detection unit; then, the significant defect determination module determines the first most significant defect in the first defect detection result according to the graph significance strategy, and marks it as a first-level defect; then, the defect degree analysis module combines the first defect detection result to perform defect degree analysis on the first-level defect, generates a first-level defect index, and uses it as the defect index of the first layer; finally, when the defect index does not reach the predetermined defect threshold of the first label, the defect handling module calls the first predetermined emergency plan to perform defect handling on the target power equipment.

[0031] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention are shown in the drawings, rather than all of them.

[0032] Embodiment 1

[0033] Figure 1 The present invention is a flow chart of a power equipment defect detection system based on infrared three-dimensional imaging, wherein the system comprises:

[0034] The imaging slicing module 11 is used to slice the target infrared three-dimensional imaging to obtain the target infrared image, where the target infrared three-dimensional imaging refers to the infrared three-dimensional imaging of the target power equipment.

[0035] Specifically, target infrared three-dimensional imaging refers to infrared image information obtained by three-dimensional imaging of target power equipment (such as transformers, transmission lines, etc.) through infrared detectors. The target infrared three-dimensional imaging not only includes the temperature distribution of the equipment surface, but also includes the temperature conditions of its various layers of structure, thereby forming a three-dimensional thermal distribution model. Exemplarily, an infrared thermal imager is used to perform multi-angle scanning or combined acquisition of multiple camera positions to obtain surface and internal temperature information of the object and generate three-dimensional thermal distribution data of the equipment.

[0036] Specifically, the target infrared three-dimensional imaging is sliced ​​to extract one or more key sections of the three-dimensional temperature distribution model and convert it into a two-dimensional infrared image, thereby providing a more intuitive thermal distribution map, which is convenient for analyzing the temperature anomaly in a specific area. Exemplarily, the longitudinal, transverse or cross-sectional area of ​​the device is selected for slicing to extract the temperature distribution of the device in a specific area or direction.

[0037] Optionally, based on a preset slicing step or slicing position, the target infrared three-dimensional imaging is sliced ​​to obtain a two-dimensional target infrared image, which reflects the infrared image information of the target power equipment on a certain section, and each pixel in the image corresponds to the temperature information on the section.

[0038] Through the above-mentioned slicing process, on the one hand, the temperature distribution of different planes is made clearer, which facilitates the precise location of temperature anomalies inside the equipment or in specific areas. On the other hand, slicing the three-dimensional imaging data into two-dimensional images reduces the amount of data, improves the efficiency of storage and processing, and facilitates real-time monitoring and analysis.

[0039] The hierarchical processing module 12 is used to introduce a predetermined hierarchical mechanism to perform hierarchical processing on the target infrared image to obtain a target hierarchical result, wherein the target hierarchical result includes a first layer marked with a first label.

[0040] Specifically, based on a preset hierarchical mechanism, the target infrared image is divided into multiple levels to facilitate targeted detection and identification of defects of different categories, thereby better meeting the requirements for high accuracy and high efficiency in power equipment detection. The preset hierarchical mechanism includes multiple defined levels, each of which has a corresponding category label.

[0041] Exemplarily, first, infrared features related to defects such as temperature distribution, texture features or shape features are extracted from the infrared image to determine the characteristic patterns of different defects. For example, an area with too high temperature may correspond to a hot spot defect, while an area with abnormal texture may correspond to a surface crack defect. Then, according to the preset defect characteristics, the infrared image is decomposed into multiple layers, each layer corresponding to a specific defect type. Each layer after layer division is marked with a first label to facilitate distinction in subsequent processing.

[0042] In some embodiments, before introducing a predetermined hierarchical mechanism to perform hierarchical processing on the target infrared image, the execution steps of the system further include:

[0043] Obtain historical defect records of similar power equipment to the target power equipment; extract a first historical record from the historical defect records, wherein the first historical record includes a first historical defect type, and the first historical defect type has an identification of a first defect infrared feature; using the first defect infrared feature as a clustering constraint, perform cluster analysis on the first historical defect type to obtain a defect clustering library; generate multiple defect clustering layers based on multiple defect clustering clusters in the defect clustering library, and set the predetermined hierarchical mechanism based on the multiple defect clustering layers.

[0044] Specifically, the predetermined hierarchical mechanism is used to divide the infrared image according to specific infrared features (such as temperature gradient, texture similarity) to form multiple hierarchical layers. Each layer is dedicated to detecting a specific type of defect, thereby avoiding mutual interference between different types of defect features and ensuring the specificity and accuracy of detection.

[0045] Specifically, a predetermined hierarchical mechanism is generated. First, historical defect data of similar power equipment is extracted from a database or an archive system. The record content includes the defect type that occurred in the past, the corresponding infrared characteristics (such as temperature anomaly, morphological characteristics) and the specific defect location. Exemplarily, the historical defect record can come from the operation and maintenance history of similar power equipment, or the diagnostic log in the fault diagnosis system of the same type of equipment.

[0046] Specifically, a typical single defect record is extracted from the historical defect record and set as a first historical record, which includes a first historical defect type and its corresponding infrared feature identifier. Among them, the defect type generally includes hot spots, cracks, corrosion, etc. These defect types are manifested as specific infrared feature identifiers (such as temperature distribution patterns or texture features) in infrared images.

[0047] Specifically, cluster analysis is performed using the first defect infrared feature as a clustering constraint condition, and defect types similar to this feature in all historical defect records are grouped together to form clusters with the same or similar infrared features. The clusters (clustering results) are stored in a defect clustering library. The defect clustering library contains multiple clusters corresponding to different defect types, and each cluster contains defect instances of the same type and their infrared feature data.

[0048] Furthermore, according to the defect clustering library, a corresponding defect clustering layer is generated for each defect cluster, and each clustering layer is used to separate and identify the corresponding defect features. Through the layered detection strategy, the infrared features of each type of defect are clearer, reducing the false alarms or missed alarms caused by feature confusion, which helps to improve the accuracy and efficiency of power equipment fault detection.

[0049] Exemplarily, first, based on each cluster in the defect cluster library, feature information related to the category of defects corresponding to the cluster is extracted. Then, the defect features are trained using a machine learning model to form a corresponding defect clustering layer to identify the infrared features of this type of defect. Next, based on multiple defect clustering layers, each clustering layer is set as an independent detection level, and a predetermined hierarchical mechanism is set so that each detection level only processes the defect type corresponding to the level. For example, the temperature anomaly detection level only detects hot spots, and the texture anomaly detection level only detects cracks, etc.

[0050] Exemplarily, during the detection process, the hierarchical mechanism applies the model of each defect clustering layer in turn for discrimination. For example, the infrared image is first input into the temperature detection layer, and after the detection is completed, it enters the texture detection layer, and then enters the shape detection layer. Each defect clustering layer performs independent discrimination in its respective detection stage to ensure accurate identification of the defect features in the clustering layer. Through the above-mentioned predetermined hierarchical mechanism, the interference problem of multiple defect features being mixed in the same layer for processing is avoided, and the accuracy of defect recognition is improved.

[0051] In some implementations, the system further includes:

[0052] Extract a second historical record from the historical defect record, wherein the second historical defect type includes a second historical defect type, wherein the second historical defect type is different from the first historical defect type, and the second historical defect type has an identification of a second defect infrared feature; obtain a first infrared indicator, and match the first defect infrared feature with the second defect infrared feature, respectively, to obtain a first indicator parameter and a second indicator parameter corresponding to the first infrared indicator, respectively; calculate a first indicator deviation rate based on the first indicator parameter and the second indicator parameter; determine whether the first indicator deviation rate is within a predetermined deviation threshold; if so, mark the first infrared indicator similarly, and if not, mark the first infrared indicator differently to obtain a first marking result; statistically analyze the first marking result to obtain a first similarity between the first historical defect type and the second historical defect type; cluster the defect types according to the first similarity to obtain the defect clustering library.

[0053] Specifically, a second historical record is extracted from the historical defect record, the second historical record and the first historical record belong to a defect type different from each other, and the second historical record also includes an identifier of a corresponding second defect infrared feature.

[0054] Specifically, the first infrared index is a key index extracted from the infrared feature, which is used to measure the characteristic values ​​of different defect types in terms of temperature, shape or texture. The index can be a specific temperature value, edge clarity, texture roughness, etc. By matching the first infrared index in the first defect infrared feature and the second defect infrared feature respectively, the corresponding values ​​are obtained, which are called the first index parameter and the second index parameter.

[0055] Specifically, the deviation rate of the first indicator parameter and the second indicator parameter is calculated. The deviation rate is used to measure the degree of difference between the two types of defect characteristics corresponding to the second historical record and the first historical record in the first infrared indicator, and then evaluate the similarity of the two types of defects. If the first indicator deviation rate is less than or equal to the threshold, it can be considered that the first infrared indicator is similar between the two defect types and is marked as "similar". Conversely, if the deviation rate exceeds the threshold, it is marked as "different". Among them, the predetermined deviation threshold is a threshold value set according to demand. If the first indicator deviation rate is lower than the threshold, it is considered that the indicator has similarity between the two types of defects.

[0056] Furthermore, the proportion of similar marking results in the total marking results is counted to obtain the first similarity of the two types of defects, indicating the similarity of the two types of defects in various infrared features. For example, if 80% of the infrared indicators are marked as similar, the similarity of the two types of defects is 80%. Then, according to the similarity analysis results, all defect types are clustered to form a defect clustering library. This library stores the clustering relationship and feature information of different defect types in a structured manner, providing data support for subsequent defect identification.

[0057] The hierarchical detection module 13 is used to match the first defect detection unit corresponding to the first layer based on the first label, and obtain a first defect detection result through the first defect detection unit.

[0058] Specifically, the first defect detection unit is a monitoring functional unit used to specifically identify the defect category corresponding to the first label. The first defect detection unit is constructed based on a machine learning model and is used to detect and identify whether the target power equipment has a defect corresponding to the first label based on the infrared features in the first layer.

[0059] In some embodiments, the system performs the following steps:

[0060] The first defect infrared feature and the first historical defect type form a first data group; the second defect infrared feature and the second historical defect type form a second data group; the first data group is subjected to supervised learning training and tested through the second data group to obtain the first defect detection unit.

[0061] Specifically, the first data group consists of a first defect infrared feature and a first historical defect type, characterizing the detection target case of the first defect detection unit. The second data group consists of a second defect infrared feature and a second historical defect type. The second defect infrared feature is different from the first group of features, representing another defect type, and is used as control data during model training.

[0062] Specifically, first, the features of the first data group are used as input, and the first historical defect type is used as the target output. The selected machine learning model (such as decision tree, support vector machine, random forest, etc.) is trained for multiple iterations, and the model parameters are adjusted to reduce the training error. At the same time, the cross-validation method is used to divide the data set into multiple subsets, and part of the data is used for training and part of the data for verification in turn to evaluate the generalization ability of the model. Then, the trained model is verified using the second data group to evaluate the performance of the model on unseen data. The inspection process is used to verify the specificity of the first defect detection unit, that is, the first defect detection unit can only identify the first defect type, but cannot deal with the second defect type. The model generated after training and verification is the first defect detection unit, which can be used to detect the first defect type in power equipment in real time.

[0063] The above execution steps enable the selected model to learn the key features of the first defect type through supervised learning training. At the same time, through the inspection of the second data group, the model is ensured to be specific and specific in identifying the defect category, thereby improving the accuracy of defect detection.

[0064] The significant defect determination module 14 is used to determine the first most significant defect in the first defect detection result according to the image significance strategy, and use the first most significant defect as a primary defect.

[0065] In some embodiments, the first most significant defect in the first defect detection result is determined according to the image significance strategy, and the execution step of the system further includes:

[0066] A first defect visual graph is generated based on the first defect monitoring result, the first defect visual graph includes multiple defects with defect range identifications; a first defect among the multiple defects with defect range identifications is extracted, the first defect corresponds to a first defect range; features of the first defect range are collected according to the graph significance strategy to obtain first defect feature parameters; a first defect feature parameter set of the first defect visual graph is constructed based on the first defect feature parameters; the first defect feature parameter set is screened to obtain a first most significant defect feature parameter, and the defect corresponding to the first most significant defect feature parameter is recorded as the first most significant defect.

[0067] Specifically, first, based on the first defect detection result, the defect feature ranges and defect feature positions corresponding to multiple defect points are extracted and visualized to generate a first defect visual map of the first defect detection result, which contains multiple defect identifiers to intuitively display the position and range of each defect. Then, the first defect is randomly selected from the multiple defects, and the first defect range is feature collected through the image significance strategy, wherein the image significance strategy is used to identify the most prominent part of the image. The acquired first defect feature parameters are used to characterize the significance level of the first defect, such as the significance of the pixel color value or the significance of the shape and size.

[0068] Furthermore, multiple defects with identifications of defect ranges in the first defect detection results are traversed to obtain multiple defect feature parameters, which are stored as a first defect feature parameter set of the first defect visual graph, and the multiple defect feature parameters in the first defect feature parameter set of the first defect visual graph are sorted to obtain a significance sequence of multiple defects with identifications of defect ranges, and then the defect corresponding to the defect feature parameter with the highest significance is extracted as the first most significant defect of the first defect visual graph.

[0069] Through the above steps, the system can effectively identify and analyze defects in the image and determine the most significant defects, thereby ensuring that the most serious defects are given priority, improving the accuracy and efficiency of defect detection, and providing a reliable basis for subsequent repair and maintenance.

[0070] In some implementations, the graph saliency strategy includes a graph color saliency strategy and a graph structure saliency strategy, including:

[0071] According to the image color significance strategy, first color characteristic parameters of the first defect range are obtained, wherein the first color characteristic parameters include a first red parameter, a first green parameter and a first blue parameter; according to the image structure significance strategy, first structural characteristic parameters of the first defect range are obtained; and the first defect characteristic parameters are composed based on the first red parameter, the first green parameter, the first blue parameter and the first structural characteristic parameter.

[0072] Optionally, the image saliency strategy is used to identify the most prominent feature area in the image, including the image color saliency strategy and the image structure saliency strategy, wherein the image color saliency strategy is used to extract the first color feature parameter of the first defect range according to the image color saliency strategy, and the first color feature parameter includes the hue and intensity of the red parameter, the green parameter, and the blue parameter. It involves color space conversion (such as RGB to HSV) for color analysis, calculation of color histogram, or use of color clustering algorithm to extract color features.

[0073] Optionally, the graph structure saliency strategy is used to extract first structural feature parameters of the first defect range according to the graph structure saliency strategy, and the first structural feature parameters include edges, textures, etc. Exemplarily, the graph structure saliency strategy involves identifying edges in an image through an edge detection algorithm (such as Canny or Sobel) and extracting texture features through a texture analysis method (such as a Gabor filter).

[0074] Furthermore, the first red parameter, the first green parameter, the first blue parameter and the first structural characteristic parameter are combined to form a first defect characteristic parameter. The first defect characteristic parameter is characterized as a characteristic vector, which is used to provide a comprehensive defect description and provide a richer information basis for subsequent defect processing.

[0075] The defect degree analysis module 15 is used to perform defect degree analysis on the first defect in combination with the first defect detection result to obtain a first defect index, and use the first defect index as the first defect index of the first layer.

[0076] In some embodiments, the execution steps of the system further include:

[0077] Taking the position of the primary defect as the defect center, determining the secondary defect threshold of the defect center in the first defect visual graph; randomly sampling the defects in the secondary defect threshold based on the random sampling principle to form secondary defects; weighting the defect characteristic parameters of each defect in the secondary defects to obtain a feedback calibration coefficient; calibrating the first defect characteristic parameter of the primary defect according to the feedback calibration coefficient to obtain a first calibration result; and using the first calibration result as the primary defect index.

[0078] Specifically, a secondary defect threshold is set in the surrounding area of ​​the primary defect according to a certain distance or spatial coordinate system, and the threshold is used to define the detection range of the secondary defect in order to identify possible related defects. Then, a random sampling method is used to select samples from the potential defects within the secondary defect threshold to obtain secondary defects, which represent smaller range or less significant related defects that may be related to the primary defect in the first defect visual map. In other words, the secondary defect is an additional defect that may be caused by the primary defect.

[0079] Specifically, each secondary defect has corresponding characteristic parameters, including temperature value, infrared radiation intensity, shape characteristics, etc. Then, a corresponding weight is assigned to the characteristic parameters of each secondary defect, and weighting is performed to highlight the influence of key features in the overall detection, and the feedback calibration coefficient is calculated. Optionally, the weight of the weighting is determined based on the importance of the secondary defect, the distance from the primary defect, etc.

[0080] Furthermore, the characteristic parameters of the primary defect are adjusted using the feedback calibration coefficient to obtain the first calibration result, ensuring that the assessment of the primary defect is more accurate and reflects the impact of the secondary defect. The calibrated result (first calibration result) is the final primary defect index, which is used to provide a basis for decision-making to evaluate the health status of the power equipment. Exemplarily, the first calibration result = (1 + feedback calibration coefficient) × first defect characteristic parameter.

[0081] In some implementations, the system further includes: introducing a defect degree assessment function to analyze the first calibration result to obtain the first-level defect index, wherein the first calibration result includes a first red calibration parameter, a first green calibration parameter, a first blue calibration parameter and a first structural feature calibration parameter.

[0082] Optionally, a defect degree assessment function performs analysis and calculation based on the first calibration result to obtain a primary defect index of the first defect, wherein the defect degree assessment function includes a mathematical regression model constructed based on historical defect assessment records; a model constructed based on the empirical formula of an expert system or professional technicians, and based on machine learning methods such as decision trees and support vector machines, which is used to provide a quantitative evaluation standard to make operation and maintenance decisions more scientific and reasonable.

[0083] Specifically, the first-level defect index is a quantitative value that indicates the severity of the defect. It is used to monitor the status of target power equipment in real time and help operation and maintenance personnel identify and deal with potential problems in a timely manner.

[0084] Specifically, the first calibration result includes: a first red calibration parameter, a first green calibration parameter, a first blue calibration parameter and a first structural feature calibration parameter. Among them, the first red calibration parameter represents the intensity of the heat source in the infrared image, corresponding to the abnormal temperature of the device; the first green calibration parameter represents the reference value under environmental factors or normal working conditions, helping to identify deviations from the expected state. The first blue calibration parameter reflects the part of the device with a lower temperature or a temperature close to the ambient temperature (such as a mechanical structure, etc.), which is used to assist in judging physical damage.

[0085] The first structural characteristic calibration parameter: describes the stability of the overall structure of the device and whether it has defects.

[0086] The defect handling module 16 is used to call a first predetermined emergency plan to handle the defect of the target power equipment when the first defect index does not meet the first predetermined defect threshold corresponding to the first tag.

[0087] Optionally, the first predetermined defect threshold is a multi-level threshold, and different levels of thresholds correspond to different predetermined emergency plans, and an association relationship is established between different predetermined emergency plans and the first predetermined defect threshold through a plan mapping matrix. Exemplarily, each row in the plan mapping matrix corresponds to a level, and each column in the row corresponds to a predetermined emergency plan corresponding to each type of defect at the level.

[0088] Through the above settings, the system provides flexible emergency plan options for defect detection and disposal of target power equipment, ensuring that appropriate measures can be taken for different situations. It ensures effective response to different defect situations and improves the operational safety and stability of power equipment.

[0089] In summary, the power equipment defect detection system based on infrared three-dimensional imaging provided by the present invention has the following technical effects:

[0090] The imaging slice module is used to slice the infrared three-dimensional imaging of the target power equipment to generate the target infrared image; the hierarchical processing module hierarchically processes the target infrared image through a predetermined hierarchical mechanism to obtain a hierarchical result including a first layer with a first label; the hierarchical detection module matches the first defect detection unit corresponding to the first layer based on the first label, and generates a first defect detection result through the detection unit; the significant defect determination module determines the most significant defect in the first defect detection result based on the graph significance strategy, which is called the first defect; the defect degree analysis module analyzes the defect degree of the first defect in combination with the first defect detection result, generates a first defect index, and uses it as the defect index of the first layer; the defect handling module calls the corresponding predetermined emergency plan to handle the defect of the target power equipment when the first defect index is lower than the predetermined defect threshold corresponding to the first label. Thus, the technical effect of improving the accuracy and adaptability of defect detection and improving detection efficiency is achieved.

[0091] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the above-mentioned embodiments. It should be understood that those skilled in the art can still modify the technical solutions recorded in the above-mentioned embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. The power equipment defect detection system based on infrared three-dimensional imaging is characterized by: include: An imaging slicing module, wherein the imaging slicing module is used to slice the target infrared three-dimensional imaging to obtain the target infrared image, wherein the target infrared three-dimensional imaging refers to the infrared three-dimensional imaging of the target power equipment; A hierarchical processing module, the hierarchical processing module is used to introduce a predetermined hierarchical mechanism to perform hierarchical processing on the target infrared image to obtain a target hierarchical result, wherein the target hierarchical result includes a first layer marked with a first label; A hierarchical detection module, the hierarchical detection module is used to match a first defect detection unit corresponding to the first layer based on the first label, and obtain a first defect detection result through the first defect detection unit; a significant defect determination module, the significant defect determination module being used to determine a first most significant defect in the first defect detection result according to a graph significance strategy, and to use the first most significant defect as a primary defect; a defect degree analysis module, the defect degree analysis module being used to perform defect degree analysis on the primary defect in combination with the first defect detection result to obtain a primary defect index, and use the primary defect index as a first defect index of the first layer; A defect handling module is used to call a first predetermined emergency plan to handle the defect of the target power equipment when the first defect index does not meet the first predetermined defect threshold corresponding to the first tag.

2. The power equipment defect detection system based on infrared three-dimensional imaging according to claim 1 is characterized in that: The execution steps of the hierarchical processing module include: Obtaining historical defect records of similar electrical equipment to the target electrical equipment; Extracting a first historical record from the historical defect record, wherein the first historical record includes a first historical defect type, and the first historical defect type has an identification of a first defect infrared feature; Taking the first defect infrared feature as a clustering constraint, performing cluster analysis on the first historical defect type to obtain a defect cluster library; A plurality of defect clustering layers are generated according to a plurality of defect clustering clusters in the defect clustering library, and the predetermined hierarchical mechanism is set based on the plurality of defect clustering layers.

3. The power equipment defect detection system based on infrared three-dimensional imaging according to claim 2 is characterized in that: The execution steps of the hierarchical processing module also include: Extracting a second historical record from the historical defect record, wherein the second historical record includes a second historical defect type, wherein the second historical defect type is different from the first historical defect type, and the second historical defect type has an identification of a second defect infrared feature; Acquire a first infrared index, and respectively match the first defect infrared feature with the second defect infrared feature to obtain a first index parameter and a second index parameter corresponding to the first infrared index; Calculate a first indicator deviation rate according to the first indicator parameter and the second indicator parameter; Determining whether the first indicator deviation rate is within a predetermined deviation threshold; If it is, marking the first infrared indicator similarly; if it is not, marking the first infrared indicator differently to obtain a first marking result; Counting the first marking results to obtain a first similarity between the first historical defect type and the second historical defect type; Defect types are clustered according to the first similarity level to obtain the defect cluster library.

4. The power equipment defect detection system based on infrared three-dimensional imaging according to claim 3 is characterized in that: The execution steps of the hierarchical detection module include: The first defect infrared feature and the first historical defect type form a first data group; The second defect infrared feature and the second historical defect type form a second data group; The first defect detection unit is obtained by performing supervised learning training on the first data set and verifying it through the second data set.

5. The power equipment defect detection system based on infrared three-dimensional imaging according to claim 1 is characterized in that: The execution steps of the significant defect determination module include: generating a first defect visual graph based on the first defect monitoring result, wherein the first defect visual graph includes a plurality of defects with identifications of defect ranges; Extracting a first defect from the plurality of defects having identifications of defect ranges, the first defect corresponding to a first defect range; Performing feature collection on the first defect range according to the image significance strategy to obtain first defect feature parameters; Building a first defect characteristic parameter set of the first defect visual graph based on the first defect characteristic parameter; The first defect characteristic parameter set is screened to obtain a first most significant defect characteristic parameter, and the defect corresponding to the first most significant defect characteristic parameter is recorded as the first most significant defect.

6. The power equipment defect detection system based on infrared three-dimensional imaging according to claim 5 is characterized in that: The image saliency strategy includes an image color saliency strategy and an image structure saliency strategy. The execution steps of the significant defect determination module also include: Acquire first color characteristic parameters of the first defect range according to the image color significance strategy, wherein the first color characteristic parameters include a first red parameter, a first green parameter, and a first blue parameter; Acquire a first structural feature parameter of the first defect range according to the graph structure significance strategy; The first defect characteristic parameter is composed based on the first red parameter, the first green parameter, the first blue parameter and the first structural characteristic parameter.

7. The power equipment defect detection system based on infrared three-dimensional imaging according to claim 6 is characterized in that: The execution steps of the defect degree analysis module include: Taking the position of the primary defect as the defect center, determining the secondary defect threshold of the defect center in the first defect visual graph; Randomly sampling defects in the secondary defect threshold based on a random sampling principle to form secondary defects; Weighting the defect characteristic parameters of each defect in the secondary defects to obtain a feedback calibration coefficient; Calibrate the first defect characteristic parameter of the primary defect according to the feedback calibration coefficient to obtain a first calibration result; The first calibration result is used as the primary defect index.

8. The power equipment defect detection system based on infrared three-dimensional imaging according to claim 7 is characterized in that: The execution steps of the defect degree analysis module also include: introducing a defect degree evaluation function to analyze the first calibration result to obtain the first-level defect index, wherein the first calibration result includes a first red calibration parameter, a first green calibration parameter, a first blue calibration parameter and a first structural feature calibration parameter.

Images