Power Fault Diagnosis and Early Warning System Based on Infrared Imagery

Abstract

This invention discloses a power fault diagnosis and early warning system based on infrared imagery, relating to the field of image data processing. The system includes: performing infrared scanning monitoring on power equipment areas to determine infrared scan images; filtering, denoising, and grayscale processing on the infrared scan images to determine the power infrared image; using corner windows determined by pixel structure tensors for local image localization; and supervising the training of a power fault diagnosis module through transfer learning; calculating the structure tensor pixel by pixel in the power infrared image using the power fault diagnosis module to locate corner window images and determine fault features to identify fault diagnosis orders; traversing the fault diagnosis orders, marking faults in the power infrared image, and displaying early warnings of power faults on the terminal interface. This invention solves the technical problem in existing power fault detection systems that cannot identify and prioritize faults in real time and accurately, leading to untimely maintenance responses and increased system operational risks.

Description

Technical Field

[0001] This application relates to the field of image data processing, and in particular to a power fault diagnosis and early warning system based on infrared spectra. Background Technology

[0002] With the continuous expansion of power system scale and the increasing complexity of power equipment, ensuring the reliable operation of power equipment has become particularly important. Traditional power fault detection methods mainly rely on manual inspections and basic monitoring techniques, which have drawbacks such as untimely detection, low accuracy, and inability to provide real-time early warnings. Therefore, there is an urgent need for an intelligent early warning system that can monitor and diagnose power equipment faults in real time and effectively prioritize fault handling.

[0003] Therefore, in the existing technology, power fault detection systems cannot identify and prioritize faults in real time and accurately, resulting in untimely maintenance response and increased system operation risks. Summary of the Invention

[0004] This application provides a power fault diagnosis and early warning system based on infrared imaging, solving the technical problem that existing power fault detection systems cannot identify and prioritize faults in real time and accurately, leading to untimely maintenance responses and increased system operational risks. It achieves the technical effect of enabling real-time and accurate detection and classification of power faults, and through fault feature clustering and prioritization, realizing efficient maintenance responses and improving the operational reliability and safety of the power system.

[0005] This application provides a power fault diagnosis and early warning system based on infrared imagery. The system includes: a scan image acquisition module for infrared scanning monitoring of power equipment areas to determine infrared scan images; an infrared image acquisition module for filtering, noise reduction, and grayscale processing of the infrared scan images based on a power data interface to determine power infrared images; a fault analysis module for determining local image localization using corner windows determined by pixel structure tensors, and supervising the training of the power fault diagnosis module through transfer learning; a power fault diagnosis module for transmitting the power infrared images back to the power fault diagnosis module, performing structure tensor calculations pixel by pixel, locating corner window images, and determining fault features to determine fault diagnosis orders; and a display and early warning module for traversing the fault diagnosis orders, marking faults in the power infrared images, and displaying power fault warnings on a terminal interface.

[0006] In a possible implementation, the fault analysis module is further configured to: determine the critical temperature value under standard power operation for the power equipment area and generate an infrared reference spectrum; establish a mapping between the infrared reference spectrum and the power equipment area, and determine the power fault characteristics of the mapped power structure based on the mapping relationship; and perform sample-driven training based on the infrared reference spectrum and the power fault characteristics to generate the power fault diagnosis module.

[0007] In a possible implementation, the fault analysis module is further used to: determine a large diagnostic model by performing sample-supervised training based on power fault records; introduce a functional relationship between gray values ​​and temperature values, perform transfer calls on the large diagnostic model, add a corner reconstruction layer and perform supervised training until convergence, and determine the power fault diagnosis module.

[0008] In a possible implementation, the power fault diagnosis module is further configured to: transmit the power infrared image to the power fault diagnosis module, perform corner window determination based on pixel structure tensor, and reconstruct and determine the corner window image; identify the corner window image, and, based on the functional relationship between grayscale value and temperature value, perform temperature value exceeding the limit determination based on infrared reference spectrum to locate abnormal temperature areas; for the abnormal temperature areas, perform power fault decision based on mapping relationship, and integrate and output the fault diagnosis report.

[0009] In a possible implementation, the power fault diagnosis module is further configured to: traverse the power infrared image and identify a first pixel, wherein the first pixel is any pixel within the power infrared image; for the first pixel, combine neighboring pixels to identify and determine a first pixel value matrix; perform corner window determination on the first pixel value matrix, and if it is a corner window, retain the first pixel value matrix; if it is not a corner window, filter out the first pixel value matrix.

[0010] In a possible implementation, the power fault diagnosis module is further configured to: determine a preset gradient value, wherein the preset gradient value is a gradient threshold between pixels; traverse the first pixel value matrix and calculate the gradient features of neighboring pixels as the structure tensor of the first pixel, wherein the structure tensor contains four terms for quantifying the gradient change features of the pixel, and the gradient features of neighboring pixels include gradient coefficients and gradient directions; based on the preset gradient value, determine a corner window by judging the structure tensor; wherein if at least one term of the structure tensor is greater than the preset gradient value, the first pixel value matrix is ​​used as the corner window.

[0011] In a possible implementation, the power fault diagnosis module is further configured to: determine corner points pixel by pixel in the power infrared image to identify N corner point windows; and, based on the relative distribution of the image, discretely stitch the N corner point windows to determine the corner point window image.

[0012] In a possible implementation, the display and warning module is further configured to: traverse the fault diagnosis sheet, determine fault features based on fault level and fault propagation characteristics, wherein the fault features include fault coefficients; perform homogeneous clustering on the fault features to determine M fault clusters, wherein faults within a cluster are homogeneous; traverse the M fault clusters to generate M sets of fault warning information, wherein the warning type of each set of fault warning information is consistent, and determine the power operation and maintenance priority based on the fault coefficients; and display and warn of power faults on the terminal interface based on the M sets of fault warning information.

[0013] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0014] This application provides a power fault diagnosis and early warning system based on infrared imagery, comprising: a scan image acquisition module for infrared scanning monitoring of power equipment areas to determine infrared scan images; an infrared image acquisition module for filtering, denoising, and grayscale processing the infrared scan images based on a power data interface to determine power infrared images; a fault analysis module for determining local image localization using corner windows determined by pixel structure tensors, and supervising the training of the power fault diagnosis module through transfer learning; a power fault diagnosis module for transmitting the power infrared images back to the power fault diagnosis module, performing structure tensor calculations pixel by pixel, locating corner window images, and determining fault features to determine fault diagnosis orders; and a display and early warning module for traversing the fault diagnosis orders, marking faults in the power infrared images, and displaying power fault warnings on a terminal interface. This system solves the technical problem in existing power fault detection systems that cannot identify and prioritize faults in real time and accurately, leading to untimely maintenance responses and increased system operational risks. It achieves the technical effect of real-time and accurate detection and classification of power faults, and through fault feature clustering and priority ranking, enables efficient maintenance responses and improves the operational reliability and safety of power systems. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0016] Figure 1 This is a schematic diagram of the structure of a power fault diagnosis and early warning system based on infrared spectra provided in an embodiment of this application.

[0017] Figure 2 This is a schematic diagram illustrating the process of integrating and outputting the fault diagnosis form in the power fault diagnosis and early warning system based on infrared spectra in this application.

[0018] Explanation of reference numerals in the attached diagram: Scan image acquisition module 11, Infrared image acquisition module 12, Fault analysis module 13, Power fault diagnosis module 14, Display and early warning module 15. Detailed Implementation

[0019] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, systems, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0022] This application provides a power fault diagnosis and early warning system based on infrared imagery, such as... Figure 1 As shown, the system includes:

[0023] The image acquisition module 11 is used to perform infrared scanning monitoring on the power equipment area and determine the infrared scanning image.

[0024] The infrared image acquisition module 12 is used to perform filtering, noise reduction, and grayscale processing on the infrared scanned image based on the power data interface to determine the power infrared image.

[0025] The fault analysis module 13 is used to determine the local image location using a corner window determined by the pixel structure tensor, and to supervise the training of the power fault diagnosis module through transfer learning.

[0026] Specifically, the image acquisition module 11 is used to perform thermal imaging scanning and monitoring of the operating area of ​​the power equipment using an infrared camera to determine the infrared scan image. The infrared camera can detect the temperature distribution on the surface of the equipment. By capturing thermal images, abnormally heated areas can be identified, which may indicate potential power faults. Subsequently, the infrared image acquisition module 12 receives the infrared scan image through the power data interface and performs filtering, noise reduction, and grayscale processing on the infrared scan image, converting the color infrared image into a grayscale image to simplify the subsequent image analysis process. The grayscale image only contains brightness information, which helps to highlight the differences in temperature distribution, thereby determining the power infrared image. Further, the fault analysis module 13 uses a corner window determined based on the pixel structure tensor to determine the local image location, and supervises the training of the power fault diagnosis module through transfer learning. The power fault diagnosis module, for abnormal temperature areas, combines the mapping relationship between the temperature area and the power equipment area to obtain the fault diagnosis of the corresponding power structure under the abnormal temperature mapping relationship.

[0027] Furthermore, the fault analysis module 13 is also used to: determine the critical temperature value under standard power operation for the power equipment area and generate an infrared reference spectrum; establish a mapping between the infrared reference spectrum and the power equipment area, and determine the power fault characteristics of the mapped power structure based on the mapping relationship; and perform sample-driven training based on the infrared reference spectrum and the power fault characteristics to generate the power fault diagnosis module.

[0028] The supervised training power fault diagnosis module includes: determining the critical temperature value of the heating area of ​​the power equipment under standard power operation conditions, where the critical temperature value is the highest allowable temperature of the heating area of ​​the power equipment under normal operation; generating infrared reference maps for each area based on the power equipment and the critical temperature value of the heating area, with the critical temperature of different areas reflecting different fault characteristics of the equipment. Taking a transformer as an example, the critical temperature values ​​of key components such as transformer windings, tanks, and terminals are obtained, and corresponding infrared reference maps are constructed. When the temperature at a certain point of the transformer winding exceeds a preset critical value, there may be a fault of winding insulation aging or short circuit at the corresponding location. Further, a mapping relationship is established between the infrared reference map and the power equipment area, and the power fault characteristics of the mapped power structure are determined based on the mapping relationship. The power fault characteristics of the power structure refer to the critical temperature characteristics exhibited in the infrared reference map of the heating area of ​​the power equipment. Finally, based on the infrared reference map and the power fault characteristics, sample-driven training is performed to generate the power fault diagnosis module.

[0029] Furthermore, the fault analysis module 13 is also used to: determine a large diagnostic model by performing sample-supervised training based on power fault records; introduce a functional relationship between gray values ​​and temperature values, perform transfer calling on the large diagnostic model, add a corner reconstruction layer and perform supervised training until convergence, and determine the power fault diagnosis module.

[0030] The generation of the power fault diagnosis module includes: supervised training based on power fault records, using labeled fault record samples, and training the model through machine learning algorithms to enable it to identify and classify different types of power faults, thus obtaining a large-scale diagnostic model. The power fault record samples include infrared spectra of various fault samples, as well as corresponding fault category identifiers and abnormal temperature region identifiers. Subsequently, a functional relationship between grayscale values ​​and temperature values ​​is introduced. This functional relationship represents the mathematical relationship between grayscale values ​​and actual temperatures, allowing for accurate conversion of grayscale information in the image into temperature information, enhancing the model's ability to identify temperature anomalies. Based on this functional relationship, the large-scale diagnostic model is transferred and applied. Using the pre-trained large-scale diagnostic model as a foundation, a transfer learning method is applied, adding a corner reconstruction layer, and supervised training is performed until convergence to determine the power fault diagnosis module. The corner reconstruction layer first performs corner window filtering analysis to obtain the heating areas of the power equipment for image segmentation and reconstruction. The reconstructed images of the heating areas are then converted to grayscale. Furthermore, based on the functional relationship between grayscale values ​​and temperature values ​​in the corner window images, a limit-breaking judgment based on critical temperatures is performed to obtain regions exceeding the critical temperature. The infrared abnormal temperature region is obtained based on the area exceeding the critical temperature. By combining the mapping relationship between the infrared abnormal temperature region and the power equipment region, the corresponding fault diagnosis result is determined. The power fault diagnosis module is used to obtain the infrared abnormal temperature region from the power infrared image and, based on the mapping relationship between the infrared abnormal temperature region and the power equipment region, perform fault diagnosis on the corresponding power structure under the abnormal temperature mapping relationship.

[0031] The power fault diagnosis module 14 is used to transmit the power infrared image back to the power fault diagnosis module, perform structural tensor calculation pixel by pixel, locate corner window images and determine fault features to determine the fault diagnosis form; the display warning module 15 is used to traverse the fault diagnosis form, mark the fault in the power infrared image, and display the power fault warning on the terminal interface.

[0032] The power fault diagnosis module 14 is used to transmit the power infrared image back to the power fault diagnosis module, perform structural tensor calculations pixel by pixel on the power infrared image, determine corner windows based on the calculation results, and determine fault features to identify fault diagnosis orders. Finally, the display and warning module 15 iterates through the fault diagnosis orders, marks faults in the power infrared image, and displays power fault warnings on the terminal interface. This solves the technical problem in the prior art that power fault detection systems cannot identify and prioritize faults in real time and accurately, leading to untimely maintenance responses and increased system operation risks. It achieves the technical effect of being able to detect and classify power faults in real time and accurately, and through fault feature clustering and prioritization, achieve efficient maintenance responses and improve the operational reliability and safety of the power system.

[0033] Furthermore, such as Figure 2 As shown, the power fault diagnosis module 14 is further configured to: transmit the power infrared image to the power fault diagnosis module, perform corner window determination based on pixel structure tensor, and reconstruct and determine the corner window image; identify the corner window image, and, based on the functional relationship between grayscale value and temperature value, perform temperature value exceeding the limit determination based on infrared reference spectrum to locate abnormal temperature areas; for the abnormal temperature areas, perform power fault decision based on mapping relationship, and integrate and output the fault diagnosis form.

[0034] The process of determining the fault diagnosis form includes: transmitting the power infrared image to the power fault diagnosis module, performing corner window determination based on pixel structure tensors, obtaining the gradient change features of each pixel in its neighborhood, obtaining corner windows, and reconstructing and determining the corner window image. Further, identifying the corner window image, and using the functional relationship between grayscale values ​​and temperature values ​​as a benchmark, determining whether the temperature values ​​in the corner window image exceed the limit based on the infrared reference spectrum, and obtaining abnormal temperature regions where the temperature values ​​exceed the limit. Finally, for the abnormal temperature regions, obtaining the fault type corresponding to the mapping relationship based on the mapping relationship, completing the power fault decision, and integrating and outputting the fault diagnosis form. The fault diagnosis form includes specific fault categories and abnormal temperature regions.

[0035] Furthermore, the power fault diagnosis module 14 is also used to: traverse the power infrared image and identify a first pixel, wherein the first pixel is any pixel in the power infrared image; for the first pixel, combine neighboring pixels to identify and determine a first pixel value matrix; perform corner window determination on the first pixel value matrix, and if it is a corner window, retain the first pixel value matrix; if it is not a corner window, filter out the first pixel value matrix.

[0036] The corner window determination based on pixel structure tensor includes: traversing the power infrared image and identifying a first pixel, where the first pixel is any pixel within the power infrared image. Then, for the first pixel, the neighboring pixels of the first pixel are obtained to obtain a first pixel value matrix. Corner window determination is performed on the first pixel value matrix. If it is a corner window, it indicates that the corresponding first pixel value may be an edge pixel value, and the first pixel value matrix is ​​retained. If it is not a corner window, the corresponding first pixel value is not an edge pixel value, and the first pixel value matrix is ​​filtered out.

[0037] Furthermore, the power fault diagnosis module 14 is also used to: determine a preset gradient value, wherein the preset gradient value is a gradient threshold between pixels; traverse the first pixel value matrix and calculate the gradient features of neighboring pixels as the structure tensor of the first pixel, wherein the structure tensor contains 4 terms used to quantify the gradient change features of the pixel, and the gradient features of neighboring pixels contain gradient coefficients and gradient directions; based on the preset gradient value, perform corner window determination by judging the structure tensor; wherein if at least one term of the structure tensor is greater than the preset gradient value, the first pixel value matrix is ​​used as the corner window.

[0038] Corner window determination is performed on the first pixel value matrix, including: determining a preset gradient value, where the preset gradient value is a gradient threshold between pixels. When the gradient coefficient is greater than the preset gradient value, the difference between adjacent pixels is large, and the corresponding pixel may be an edge pixel. The first pixel value matrix is ​​traversed, and the gradient features of the first pixel value and its neighboring pixels are calculated as the structure tensor of the first pixel. The structure tensor contains four terms used to quantify the gradient change features of the pixel, and the gradient features of the neighboring pixels include gradient coefficients and gradient directions. Based on the preset gradient value, the first pixel with a gradient greater than the preset gradient value is obtained by determining whether the structure tensor is greater than the preset gradient value, thus completing the corner window determination. Specifically, if at least one term of the structure tensor is greater than the preset gradient value, the first pixel value matrix is ​​used as the corner window.

[0039] Furthermore, the power fault diagnosis module 14 is also used to: determine corner points pixel by pixel in the power infrared image to identify N corner point windows; and to discretely stitch the N corner point windows based on the relative distribution of the image to determine the corner point window image.

[0040] The reconstruction and determination of the corner window image includes: based on the acquired power infrared image, determining corner windows for pixels in the power infrared image; identifying corner windows where at least one of the structure tensors is greater than the preset gradient value; obtaining all corner windows that meet the requirements; and determining N corner windows. After completing the corner detection of the power infrared image, these scattered corner windows need to be further processed to form a "corner window image" for subsequent fault diagnosis analysis or model input. Based on the relative positions of the corner windows in the image, i.e., the spatial positions of each corner window in the infrared image coordinate system and their distance relationships, each of the N corner windows is considered as a discrete small image block. The corner windows are sequentially stitched together according to the order from left to right and from top to bottom, and when the distance between corner windows is less than a preset distance threshold. This determines the stitched corner window image. These corner window images only retain information around the most identifiable key points in the image, reducing interference from a large number of irrelevant areas.

[0041] Furthermore, the display and warning module 15 is also used to: traverse the fault diagnosis sheet, determine fault features based on fault level and fault propagation characteristics, wherein the fault features include fault coefficients; perform homogeneous clustering processing on the fault features to determine M fault clusters, wherein faults within a cluster are homogeneous; traverse the M fault clusters to generate M sets of fault warning information, wherein the warning type of each set of fault warning information is consistent, and determine the power operation and maintenance priority based on the fault coefficients; and display and warn of power faults on the terminal interface based on the M sets of fault warning information.

[0042] The fault diagnosis forms are iterated through, and each fault record in the forms is checked one by one. Assume the fault diagnosis forms contain three faults: transformer terminal overheating, poor contact in switchgear, and abnormal temperature of the distribution panel. Fault characteristics are determined based on the fault level and fault propagation characteristics. Each fault type corresponds to a preset specific fault level and fault propagation characteristic parameter, i.e., the impact level on other equipment. These two parameters are summed to obtain the fault characteristics. For example: transformer terminal overheating may cause equipment damage, which is level 3. Poor contact in switchgear may cause a short circuit, which is level 2. Further, a clustering algorithm is used to group faults from the same source together. Faults in each cluster have high similarity in characteristics, resulting in M ​​fault clusters, where M is a positive integer. Faults within a cluster are considered to be from the same source, meaning they have the same fault characteristics, including the same fault location and fault coefficient. Further, the M fault clusters are iterated through, generating M sets of fault warning information. The warning types of each set of fault warning information are consistent, and the power operation and maintenance priority is determined based on the fault coefficient. Faults originating from the same source use the same warning method, such as icon flashing at different frequencies, which facilitates identification and allows for direct fault determination. Based on the aforementioned M sets of fault warning information, power fault warnings are displayed on the terminal interface.

[0043] This application embodiment employs a scanning image acquisition module for infrared scanning monitoring of power equipment areas to determine infrared scanning images; an infrared image acquisition module for filtering, noise reduction, and grayscale processing of the infrared scanning images based on a power data interface to determine power infrared images; a fault analysis module for determining local image location using corner windows determined by pixel structure tensors, and supervising training of a power fault diagnosis module through transfer learning; a power fault diagnosis module for transmitting the power infrared images back to the power fault diagnosis module, performing structure tensor calculations pixel by pixel, locating corner window images, and determining fault features to determine fault diagnosis orders; and a display and warning module for traversing the fault diagnosis orders, marking faults in the power infrared images, and displaying power fault warnings on the terminal interface. This solves the technical problem in existing power fault detection systems that cannot identify and prioritize faults in real time and accurately, leading to untimely maintenance responses and increased system operational risks. It achieves the technical effect of real-time and accurate detection and classification of power faults, and through fault feature clustering and priority ranking, realizes efficient maintenance responses, and improves the operational reliability and safety of power systems.

[0044] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A power fault diagnosis and early warning system based on infrared imagery, characterized in that, The system includes: The image acquisition module is used to perform infrared scanning monitoring on the power equipment area and determine the infrared scan image; The infrared image acquisition module is used to perform filtering, noise reduction, and grayscale processing on the infrared scanned image based on the power data interface to determine the power infrared image; The fault analysis module is used to determine the local image location using a corner window determined by the pixel structure tensor. Through transfer learning, the power fault diagnosis module is trained in a supervised manner. The power fault diagnosis module is used to transmit the power infrared image back to the power fault diagnosis module, perform structural tensor calculation pixel by pixel, locate corner window images and determine fault features, and determine the fault diagnosis order. The display and warning module is used to traverse the fault diagnosis form, mark the fault in the power infrared image, and display and warn of power faults on the terminal interface. The power fault diagnosis module is also used for: The power infrared image is transmitted to the power fault diagnosis module, where corner window determination is performed based on pixel structure tensor, and the corner window image is reconstructed and determined. The corner window image is identified, and based on the functional relationship between grayscale value and temperature value, a temperature value exceeding the limit is determined based on the infrared reference spectrum to locate abnormal temperature areas. For the abnormal temperature region, power fault decisions are made based on the mapping relationship, and the fault diagnosis report is integrated and output. The power fault diagnosis module is also used for: Traverse the power infrared image and identify the first pixel, wherein the first pixel is any pixel within the power infrared image; For the first pixel, combine the neighboring pixels to identify and determine the first pixel value matrix; The first pixel value matrix is ​​subjected to corner window determination. If it is a corner window, the first pixel value matrix is ​​retained; if it is not a corner window, the first pixel value matrix is ​​filtered out. The power fault diagnosis module is also used for: Determine a preset gradient value, wherein the preset gradient value is a gradient threshold between pixels; Traverse the first pixel value matrix and calculate the gradient features of neighboring pixels as the structure tensor of the first pixel. The structure tensor contains four terms to quantify the gradient change features of the pixel. The gradient features of neighboring pixels include gradient coefficients and gradient directions. Based on the preset gradient value, corner window determination is performed by judging the structure tensor; If at least one of the structural tensors is greater than the preset gradient value, the first pixel value matrix is ​​used as the corner window.

2. The power fault diagnosis and early warning system based on infrared spectrum as described in claim 1, characterized in that, The fault analysis module is also used for: For the power equipment area, the critical temperature value under standard power operation is determined, and an infrared reference spectrum is generated; Establish a mapping between the infrared reference spectrum and the power equipment area, and determine the power fault characteristics of the mapped power structure based on the mapping relationship; Based on the infrared reference spectrum and the power fault characteristics, sample-driven training is performed to generate the power fault diagnosis module.

3. The power fault diagnosis and early warning system based on infrared spectrum as described in claim 2, characterized in that, The fault analysis module is also used for: A large diagnostic model was determined by conducting supervised training based on power fault records. By introducing a functional relationship between grayscale values ​​and temperature values, the diagnostic model is transferred and called, a corner reconstruction layer is added, and supervised training is performed until convergence to determine the power fault diagnosis module.

4. The power fault diagnosis and early warning system based on infrared spectrum as described in claim 1, characterized in that, The power fault diagnosis module is also used for: For the aforementioned power infrared image, corner points are determined pixel by pixel to identify N corner point windows; Based on the relative distribution of the image, the N corner point windows are discretely stitched together to determine the corner point window image.

5. The power fault diagnosis and early warning system based on infrared spectrum as described in claim 1, characterized in that, The power fault diagnosis module is also used for: The fault diagnosis forms are traversed, and fault characteristics are determined based on fault level and fault propagation characteristics, wherein the fault characteristics include fault coefficients; The fault characteristics are subjected to homogeneous clustering to determine M fault clusters, where faults within a cluster are homogeneous. Traverse the M fault clusters to generate M sets of fault warning information, wherein the warning type of each set of fault warning information is consistent, and the power operation and maintenance priority is determined based on the fault coefficient. Based on the M sets of fault warning information, power fault warnings are displayed on the terminal interface.

Images