Method and system for identifying abnormality of expander based on substation inspection image

Abstract

The invention provides an expander abnormity identification method and system based on a transformer substation patrol image, electronic equipment and a computer readable storage medium. The method comprises the following steps: acquiring the transformer substation patrol image; using a YOLOv11-Marge model used for expander position detection to carry out expander area detection on the transformer substation patrol image, and identifying and extracting an expander area image; performing expander defect detection on the expander area image by using a YOLOv11-Marge model used for expander defect detection; wherein the YOLOv11-Marge model used for expander defect detection is obtained after training of an expander defect expansion sample, and the expander defect expansion sample is obtained after expansion of the expander defect sample is carried out on the expander area image by using a diffusion model; and identifying and outputting the abnormal category of the expander. According to the method, the technical effects of effectively improving the defect recognition rate of the expander, reducing the missed recognition rate and enhancing the generalization performance of the model are achieved.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence and substation technology, and specifically relates to an expander anomaly recognition method, system, electronic equipment and computer-readable storage medium based on substation patrol images. Background Art

[0002] Expanders play a vital role in the stable operation of substations. Due to the magnetic field effect of current and low-energy discharge, expanders may have defects such as top punching, which will cause distortion of voltage and current measurements, and in severe cases may even cause safety problems such as falling off and bursting. Therefore, the performance stability of expanders is directly related to production safety and efficiency.

[0003] In actual operation, manual inspection is not only labor-intensive, but also involves potential safety hazards and unstable factors. Bad weather conditions can also affect inspection efficiency. Therefore, it is particularly important to timely, automatically and accurately identify expander defects through inspection images, discover potential problems in the early stages of defects, and avoid major safety accidents caused by small defects.

[0004] With the rapid development of deep learning technology, there are now more and more machine learning models to choose from, which can combine the characteristics of expander defects to realize fully automatic expander inspection tasks. However, in the current expander abnormality recognition technology, most methods are based on the recognition of expanders and perform defect recognition through simple image processing operations such as binarization and color gamut transformation. When classifying and judging expanders, the above methods often only perform simple classification on the whole. The disadvantages are as follows: Due to the limitations of the recognition data domain and the diversity of expander defect types, usually only the expander defects that have been seen can be recognized more accurately, resulting in overfitting problems. In addition, for expander defects that have not been seen, the above method may have the problem of missed recognition.

[0005] Therefore, there is an urgent need for an expander anomaly recognition method based on substation patrol images with high recognition rate and good generalization. Summary of the invention

[0006] The present invention provides an expander anomaly recognition method, system, electronic device and computer-readable storage medium based on substation patrol images, so as to overcome at least one technical problem existing in the prior art.

[0007] To achieve the above object, the present invention provides a method for identifying expander abnormality based on substation patrol images, comprising:

[0008] Acquire substation inspection images;

[0009] Using the YOLOv11-Large model for expander position detection, the substation inspection image is subjected to expander area detection, and the expander area image is identified and extracted;

[0010] Performing expander defect detection on the expander region image using a YOLOv11-Large model for expander defect detection; wherein the YOLOv11-Large model for expander defect detection is obtained by training with expander defect expansion samples, and the expander defect expansion samples are obtained by expanding the expander region image with expander defect samples using a diffusion model;

[0011] Identify and output expander anomaly categories.

[0012] Further, a preferred method is to use a diffusion model to expand the expander region image to expand the expander defect sample, and the method for obtaining the expander defect expansion sample includes:

[0013] labeling the expander region image with defect category labels;

[0014] Classifying the expanded region images into defect categories after annotation, and generating defect category text for the expanded region images of each defect category;

[0015] The labeled expander region image and the defect category text corresponding to the expander region image are expanded using a diffusion model to obtain expander defect expansion samples.

[0016] Furthermore, a preferred method is, before using a diffusion model to expand the annotated expander area image and the defect category text corresponding to the expander area image to obtain the expander defect expanded sample, it also includes training two low-rank matrices, the product of which approximates the model weight increment, to achieve fine-tuning of the diffusion model.

[0017] Further, a preferred method is that the defect category labels include rust defects, top cover lifting defects and gap defects.

[0018] Furthermore, a preferred method is that the YOLOv11-Large model for expander position detection is obtained by training a training data set; it also includes performing data augmentation on the training data set during the training process, and the data augmentation method includes perspective transformation, translation transformation, angle rotation, mosaic enhancement, Mixup image mixing and RandomAugment.

[0019] Furthermore, a preferred method is that the YOLOv11-Large model for expander defect detection is a YOLOv11-Large model with a decoupled head structure without an anchor frame.

[0020] Further, the preferred method is that the YOLOv11-Large model for expander defect detection is obtained after training based on a loss function;

[0021] The loss function is obtained by the following formula:

[0022] L=λ1Loss CIoU +λ1Loss DFL +λ2Loss BCE

[0023] Among them, λ1 and λ2 are hyperparameters; Loss BCE is the loss function of the classification branch; Loss DFL and Loss CIoU is the loss function of the regression branch.

[0024] In order to solve the above problems, the present invention also provides an expander abnormality identification system based on substation patrol images, which uses the above expander abnormality identification method based on substation patrol images to identify expander abnormalities. The system includes:

[0025] An image acquisition unit, used for acquiring substation inspection images;

[0026] A regional image acquisition unit, used to perform expander region detection on the substation patrol image using a YOLOv11-Large model for expander position detection, and identify and extract an expander region image;

[0027] A defect recognition unit is used to perform expander defect detection on the expander area image using the YOLOv11-Large model for expander defect detection; wherein the YOLOv11-Large model for expander defect detection is obtained after training with expander defect expansion samples, and the expander defect expansion samples are obtained after expanding the expander area image with expander defect samples using a diffusion model; and identify and output expander abnormality categories.

[0028] In order to solve the above problems, the present invention also provides an electronic device, which includes a memory, a processor, and an expander abnormality identification program based on substation patrol images stored in the memory and executable on the processor, wherein the expander abnormality identification program based on substation patrol images implements the steps in the above-mentioned expander abnormality identification method based on substation patrol images when executed by the processor.

[0029] In order to solve the above problems, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned expander abnormality identification method based on substation patrol images.

[0030] The present invention provides an expander anomaly recognition method, system, electronic device, and computer-readable storage medium based on substation inspection images. By adopting the YOLOv11-Large model and combining the diffusion model to expand defect samples, a two-stage recognition model framework is proposed; wherein the first stage is expander area recognition; and the second stage is specific expander defect recognition. The present invention achieves the technical effect of effectively improving the recognition rate of expander defects, reducing the missed recognition rate, and enhancing the generalization performance of the model. While reducing the cost of inspection personnel and ensuring the safety of inspection personnel, it is of great significance to maintain the stable operation of substations. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1 A schematic flow chart of a method for identifying expander anomalies based on substation patrol images provided in accordance with an embodiment of the present invention.

[0033] Figure 2 A schematic diagram of the model training principle of an expander anomaly recognition method based on substation patrol images provided in one embodiment of the present invention.

[0034] Figure 3 A schematic diagram of expander abnormality identification output results provided by an embodiment of the present invention.

[0035] Figure 4 A module diagram of an expander anomaly recognition system based on substation patrol images provided by an embodiment of the present invention.

[0036] Figure 5 A schematic diagram of the internal structure of an electronic device for implementing an expander anomaly identification method based on substation patrol images provided in one embodiment of the present invention.

[0037] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0038] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0039] The technical solutions in the embodiments of the present application will be described clearly and in detail below in conjunction with the accompanying drawings. In the description of the embodiments of the present application, unless otherwise specified, "and / or" in the text is only a description of the association relationship of the associated objects, indicating that three relationships may exist, for example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone.

[0040] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In addition, in the description of the embodiments of the present application, "plurality" means two or more than two.

[0041] References to "one embodiment" or "some embodiments" etc. described in this specification mean that a particular feature, structure or characteristic described in conjunction with the embodiment is included in one or more embodiments of the present application. Thus, the phrases "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear at different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0042] Computer Vision (CV) is a science that studies how to make machines "see". To put it more concretely, it refers to machine vision such as using cameras and computers to replace human eyes to identify and measure targets, and further do graphic processing to make computer processing more suitable for human eye observation or transmission to instrument detection. As a scientific discipline, computer vision studies related theories and technologies, and attempts to establish an artificial intelligence system that can obtain information from images or multidimensional data. Computer vision technology generally includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, synchronous positioning and map construction, and also includes common biometric recognition technologies such as face recognition and fingerprint recognition. The recognition model in the embodiment of the present application utilizes computer vision technology.

[0043] Machine Learning (ML) is a multi-disciplinary cross-disciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithmic complexity theory and other disciplines. It specializes in how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all fields of artificial intelligence. Machine learning and deep learning generally include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and learning by teaching. The recognition model in the embodiment of the present application is trained using machine learning or deep learning techniques.

[0044] Stable Diffusion is a latent diffusion model that can generate detailed images from text descriptions. It can also be used for tasks such as image restoration, image drawing, text to image, and image to image. Stable Diffusion converts the "image generation" process into a "diffusion" process that gradually removes noise. The whole process starts with random Gaussian noise, and gradually removes noise after training until there is no noise, and finally outputs an image that is closer to the text description. The disadvantage of this process is that the denoising process is very time-consuming and memory-intensive, especially when generating high-resolution images. Stable Diffusion introduces latent diffusion to solve this problem. Latent diffusion reduces memory and computational costs by applying a diffusion process on a lower-dimensional latent space instead of using the actual pixel space.

[0045] The software environment of this device is window10 and above, python3.8 and above, and the version of ultralytics package is 8.3.15 and later.

[0046] The hardware environment of this device: CPU is Intel Core i5 12400F or above, memory is 8G or above, independent graphics card is 3060 8G or above.

[0047] Example 1

[0048] Figure 1 The expander anomaly recognition method based on substation inspection images is described in general. Figure 1 A flow chart of a method for identifying expander anomalies based on substation patrol images provided in accordance with an embodiment of the present invention; the method may be executed by a system, which may be implemented by software and / or hardware.

[0049] like Figure 1As shown, in this embodiment, the expander abnormality identification method based on the substation inspection image includes steps S110 to S140.

[0050] S110, obtaining substation inspection images.

[0051] Specifically, according to the State Grid Corporation's 2022 remote intelligent patrol system detection plan for more than 500 substations, there are clear technical specifications for the collection of substation patrol images. Visible light photos should be in jpg format with a resolution of not less than 1920×1080; infrared pictures should also be in jpg format with a resolution of not less than 640×480. In the specific implementation process, the substation patrol image not only includes the expander image, but also includes records of other specific equipment and states, such as SF6 pressure gauges, switch action counters, lightning arrester leakage current meters and other meter readings, as well as the appearance of the equipment. In this embodiment, the substation patrol image of the expander area is to be obtained.

[0052] S120. Perform expander area detection on the substation inspection image using the YOLOv11-Large model for expander position detection, and identify and extract the expander area image.

[0053] Due to camera cost and efficiency considerations, the patrol images of substations are shot at a certain distance from the target. The size of the patrol images of substations is generally 1080p or 720p, and the expander is generally located at a higher position. In order to improve the recognition rate of expander defects in the expander abnormality recognition method based on substation patrol images, reduce the missed recognition rate, and achieve better generalization performance, the YOLOv11-Large model for expander position detection is first used to identify the expander area of ​​the substation patrol image.

[0054] Figure 2 A schematic diagram of the model training principle of an expander anomaly recognition method based on substation patrol images provided in one embodiment of the present invention.

[0055] like Figure 2As shown in the figure, image acquisition and annotation: collect the first image (substation inspection image); use labelimg software to annotate the expanders in the first image and obtain the annotation information of the expander area. Data preprocessing and enhancement: perform data equalization on the annotated first image to ensure that the number of each type of expander is basically the same. Use data enhancement techniques (such as translation, flipping, perspective transformation, adding random noise, etc.) to expand the image to improve the generalization ability of the model. In other words, in order to ensure that the number of samples of each type of expander in the data set is balanced, the first image and its annotations are first analyzed to identify and count the frequency of occurrence of different types of expanders. Subsequently, the open source data enhancement library is used to enhance the expander categories with insufficient sample numbers. The enhancement methods include image translation, flipping, applying perspective transformation, and injecting random noise to artificially increase the sample size of these categories. During the enhancement process, new images and their corresponding annotations are automatically generated to ensure that the number of samples of each type of expander is relatively balanced in the end.

[0056] Next, train the first model (i.e., the YOLOv11-Large model for expander position detection): Use the processed first image to train the first model, i.e., the expander position detection model, using the YOLOv11-Large model. During the training process, enable online data augmentation, including perspective transformation, translation transformation, rotation, mosaic enhancement, and mixup.

[0057] During the data augmentation process, the amplitude of the perspective transformation is carefully set to 0.0001 to achieve subtle changes in perspective. The parameter of the translation transformation is set to 0.1, allowing the image to be moderately shifted in the horizontal or vertical direction. The rotation transformation is set to a random angle, not exceeding 10°, to simulate the slight rotation that may occur during actual shooting. In addition, mosaic enhancement is activated with a probability of 90% to introduce a pixel-level mixing effect to the image; mixup enhancement is adopted with a probability of 80% to increase sample diversity by fusing two images. At the same time, randomaugment is also enabled to randomly select and apply a series of predefined enhancement operations. Except for these specifically set parameters, the remaining parameters remain at the default settings.

[0058] Expander area extraction and expansion: Cut out the marked expander area from the first image and expand it by 10% to generate the second image (i.e., expander area image). Defect annotation and sample expansion: Use labelimg software to annotate the expander defects in the second image. The annotation method is a rectangular frame, such as severe rust and fall-off, raised top cover, and lower edge raised gaps. The second image is annotated with a rectangular frame to obtain the annotation.

[0059] The Stable Diffusion model and LoRa fine-tuning technology are used to generate defect images to increase the diversity of defect samples. Incremental data generation and integration: The Stable Diffusion model is trained based on the generated defect images and corresponding text prompts. Manual screening is used to generate images with better quality as incremental data, and the proportion of each category is balanced. For example, severe rust and shedding generate rust or shedding in different directions and colors; the raised top cover generates different directions, opening and closing sizes, and heights; the raised gap on the lower edge generates gaps of different sizes.

[0060] Integrate the incremental data with the second image to form a third image (i.e., the training set of the YOLOv11-Large model for expander defect detection, the expander defect expansion sample). Second model training: Use the integrated third image to train the second model (i.e., the YOLOv11-Large model for expander defect detection), that is, the expander defect detection model, and also use the YOLOv11-Large model. Continue to enable online data augmentation during training, using the same settings as the first model. Defect detection and result output: Use the trained second model to perform defect detection on the third image, identify and output the expander anomaly category. Perform logical analysis based on the detection results to obtain the final defect detection report.

[0061] The YOLOv11-Large model for expander position detection is obtained after training with a training data set; the training data set is subjected to data augmentation, and the data augmentation method includes perspective transformation, translation transformation, angle rotation, mosaic enhancement, Mixup image mixing and RandomAugment.

[0062] S130. Perform expander defect detection on the expander area image using the YOLOv11-Large model for expander defect detection; wherein the YOLOv11-Large model for expander defect detection is obtained by training with expander defect expansion samples, and the expander defect expansion samples are obtained by expanding the expander area image with expander defect samples using a diffusion model.

[0063] Specifically, the method of using a diffusion model to expand the expander area image with expander defect samples to obtain the expanded expander defect samples includes: S1301, labeling the expander area image with defect category labels. The defect category labels include rust defects, top cover lifting defects and gap defects. S1302, classifying the labeled expander area image into defect categories, and generating defect category texts for the expander area image of each defect category; S1303, using a diffusion model to expand the labeled expander area image and the defect category texts corresponding to the expander area image to obtain expanded expander defect samples. Through the samples expanded by the diffusion model, the model can not only recognize the defect types seen in the training data, but also effectively recognize the defect types that have not been seen, which enhances the generalization ability of the model.

[0064] Combination Figure 2 As shown in the figure, in the specific implementation process, the category data of different defects in the second image are collected for label classification, such as severe rust and peeling label is xiushi, the top cover label is pzqcd, and the bottom edge top gap label is fengxi. In the second image, different categories of data are placed in different folders, and corresponding texts are automatically generated according to the category images of different defects, such as A photo of a pzqcd. A photo of a xiushi. A photo of afengxi. and other prompt sentences as the prompt corresponding text of the image. After the above processing of the second image, combined with the generated corresponding text, the stable diffusion model is trained.

[0065] An exemplary description is as follows: generate an expander defect image, collect expander types with different defects in the second image, such as xiushi, pzqcd, and fengxi, and divide them into 3 folders, each folder corresponds to each image, and generate text in json format at the same time. The text format is a photo of a*, where * is the corresponding category label. Finally, the three folders are summarized and packaged as training data, and trained using SD (stable diffusion). The trained SD model uses the text prompt a photo of a* to generate the desired expander defect image type.

[0066] In order to save video memory, fine-tuning training is performed on the basis of the public model based on the LoRA method. Specifically, the use of LoRA fine-tuning technology can effectively reduce video memory consumption, so that fine-tuning training of the model can also be performed on a GPU with limited resources. That is to say, before using the diffusion model to expand the labeled expander area image and the defect category text corresponding to the expander area image to obtain the expander defect expansion sample, it also includes training two low-rank matrices, and the product of the two low-rank matrices approximates the model weight increment to achieve fine-tuning of the diffusion model. In order to achieve effective adjustment and optimization of the model without significantly increasing the computational burden.

[0067] The YOLOv11-Large model optimizes the feature maps of the previous layer and enriches them with an attention mechanism to improve the performance of the model, enabling the model to achieve real-time fast reasoning on devices with less computing resources while ensuring high accuracy. Specifically, the YOLOv11-Large model has been significantly improved in many aspects compared to earlier versions of the YOLO series. First, it introduces the C3K2 module, which not only enhances the feature extraction capability, but also effectively reduces the model parameters through its channel separation strategy. Secondly, the addition of the SPFF module allows the model to more flexibly identify objects of different sizes, especially improving the detection capability of small targets.

[0068] In addition, the YOLOv11-Large model also introduces the C2PSA module, which integrates the attention mechanism to further improve the accuracy of the model. This attention mechanism allows the model to pay more attention to key areas in the image, especially when detecting small targets, which can improve the accuracy of detection. This module works in conjunction with the multi-head attention mechanism to further improve the model's ability to accurately perceive objects. By optimizing the feature maps of the previous layer and using the attention mechanism to enhance them, the YOLOv11-Large model reduces training parameters and optimizes the network feature extraction framework while maintaining high accuracy. As a result, the model can achieve fast and real-time inference on resource-constrained devices while maintaining high-precision detection performance. In short, the YOLOv11-Large model achieves efficient and accurate target detection in resource-constrained environments through a series of improvements.

[0069] The YOLOv11-Large model for expander defect detection is a YOLOv11-Large model with a decoupled head structure without an anchor box. Specifically, the detection adopts an anchor-free design method, which does not require the artificial design of the box size, simplifying the design, calculation and memory consumption. The detection head part uses a separable decoupled head, which separates the classification and regression tasks, reduces the mutual interference between tasks, and improves the detection accuracy of the model; by separating the pixel-level prediction from the feature extraction, the Decoupled Head can better utilize the semantic information between the bottom and high-level features. This separation enables the model to make more full use of features at different levels and improve the expressiveness of the model, thereby better handling targets of different scales and fineness, enhancing the expressiveness of the model; at the same time, it improves the flexibility of the model; reduces the number of parameters and computational complexity.

[0070] The YOLOv11-Large model for expander defect detection is obtained after training based on the loss function;

[0071] The loss function is obtained by the following formula:

[0072] L=λ1Loss CIoU +λ1Loss DFL +λ2Loss BCE

[0073] Among them, λ1 and λ2 are hyperparameters; Loss BCE is the loss function of the classification branch; Loss DFL and Loss CIoU is the loss function of the regression branch.

[0074] It can be understood that hyperparameters are adjustable hyperparameters used to control the plasticity of the model.

[0075] For the loss function of the classification branch, it is obtained by the following formula:

[0076]

[0077] Among them, y i is the label, p(y i ) is the probability of the label and N is the number of samples.

[0078] Loss function of regression branch DFL , obtained by the following formula:

[0079]

[0080] Where N is the sample size, C is the number of categories, and y ic is the true label of the i-th sample, p ic is the predicted probability that the ith sample belongs to category c, α is the balancing factor used to adjust the weights between positive and negative samples, and γ is the focusing parameter used to control the degree of attention paid to difficult samples.

[0081] Loss function of regression branch CIoU , obtained by the following formula:

[0082]

[0083] in, is the aspect ratio factor of the regression box, where w and h are the width and length of the regression box respectively. b is the predicted box coordinate, b gt is the coordinate of the true value box. b represents the length of the diagonal of the minimum outer rectangle of the true value box and the predicted box, ρ(b,b gt ) represents the distance between the center points of the true value box and the predicted box.

[0084]

[0085] Among them, Intersection represents the area of ​​the intersection of the true value box and the prediction box, and Union represents the area of ​​the true value box and the prediction box.

[0086] S140. Identify and output the expander abnormality category.

[0087] Based on the classification network, the features of the image to be detected are subjected to defect classification to obtain the predicted defect information of the image to be detected, wherein the predicted defect information is used to characterize the classification result of whether the image to be detected contains defects.

[0088] Specifically, in the process of detecting expander defects, a cascade method of two models is adopted. First, the image to be tested will be input into the YOLOv11-Large model for expander position detection, which is responsible for identifying and extracting the key area in the image, that is, the first area image. Then, this first area image will be sent to the YOLOv11-Large model for expander defect detection for further analysis and detection. The task of the YOLOv11-Large model for expander defect detection is to identify whether there are specific defect types in the first area image, including xiushi, pzqcd, and fengxi. If the detection results of the YOLOv11-Large model for expander defect detection show one or more of these defect types, the system will determine that the expander is defective and return the corresponding defect results. On the contrary, if these defect types are not found in the detection results of the YOLOv11-Large model for expander defect detection, that is, no defect is detected, then the system will conclude that the expander is in a normal state and return the corresponding results. Figure 3 In short, this two-stage inspection process first locates the critical area through the first model, then uses the second model to identify defects, and finally determines the status of the expander based on the output of the second model to ensure that it can accurately identify whether the expander has defects.

[0089] In summary, the YOLOv11-Large model, as the core of expander position detection and defect detection, improves the feature extraction capability through its advanced network structure, such as the C3K2 module and the SPPF module, so that the model can more accurately identify the expander area and the defects therein. In addition, the diffusion model is used to expand the expander defect samples, which helps the model learn more diverse defect features, thereby improving the recognition rate and reducing the missed recognition rate. Through the samples expanded by the diffusion model, the model can not only identify the defect types seen in the training data, but also effectively identify the defect types that have not been seen, which enhances the generalization ability of the model.

[0090] like Figure 4 As shown, the present invention provides an expander anomaly identification system 400 based on substation patrol images, and the present invention can be installed in an electronic device. According to the functions implemented, the expander anomaly identification system 400 based on substation patrol images can include an image acquisition unit 410, a regional image acquisition unit 420 and a defect identification unit 430. The unit of the present invention can also be called a module, which refers to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, which are stored in the memory of the electronic device.

[0091] In this embodiment, the functions of each module / unit are as follows:

[0092] An image acquisition unit 410 is used to acquire substation inspection images;

[0093] A regional image acquisition unit 420, configured to perform expander region detection on the substation patrol image using a YOLOv11-Large model for expander position detection, and identify and extract an expander region image;

[0094] The defect recognition unit 430 is used to perform expander defect detection on the expander area image using the YOLOv11-Large model for expander defect detection; wherein the YOLOv11-Large model for expander defect detection is obtained after training with expander defect expansion samples, and the expander defect expansion samples are obtained after expanding the expander area image with expander defect samples using a diffusion model; and identify and output the expander abnormality category.

[0095] The expander anomaly recognition system 400 based on substation patrol images of the present invention adopts the YOLOv11-Large model and combines the diffusion model to expand the defect samples, and proposes a two-stage recognition model framework; wherein the first stage is the expansion de-region recognition; the second stage is the specific defect recognition of the expander. The present invention achieves the technical effect of effectively improving the recognition rate of expander defects, reducing the missed recognition rate, and enhancing the generalization performance of the model. While reducing the cost of patrol personnel and ensuring the safety of patrol personnel, it is of great significance to maintain the stable operation of the substation.

[0096] like Figure 5 As shown, the present invention also provides an electronic device 1 for an expander abnormality recognition method based on substation patrol images.

[0097] The electronic device 1 may include a processor 10, a memory 11 and a bus, and may also include a computer program stored in the memory 11 and run on the processor 10, such as an expander anomaly recognition program 12 based on a substation patrol image. The memory 11 may also include both an internal storage unit of the expander anomaly recognition system based on a substation patrol image and an external storage device. The memory 11 may be used not only to store the code installed in the application software and various data, such as the expander anomaly recognition program based on a substation patrol image, but also to temporarily store the data that has been output or is to be output.

[0098] The electronic device 1 may include a processor 10, a memory 11 and a bus, and may also include a computer program stored in the memory 11 and run on the processor 10, such as an expander anomaly recognition program 12 based on a substation patrol image. The memory 11 may also include both an internal storage unit of the expander anomaly recognition system based on a substation patrol image and an external storage device. The memory 11 may be used not only to store the code installed in the application software and various data, such as the expander anomaly recognition program based on a substation patrol image, but also to temporarily store the data that has been output or is to be output.

[0099] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. The memory 11 may be an internal storage unit of the electronic device 1 in some embodiments, such as a mobile hard disk of the electronic device 1. The memory 11 may also be an external storage device of the electronic device 1 in other embodiments, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 11 may also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 11 may not only be used to store application software and various types of data installed in the electronic device 1, such as the code of the expander abnormality recognition program based on the substation patrol image, but also be used to temporarily store data that has been output or is to be output.

[0100] The processor 10 may be composed of an integrated circuit in some embodiments, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips, etc. The processor 10 is the control core (Control Unit) of the electronic device, and uses various interfaces and lines to connect various components of the entire electronic device, and executes or executes programs or modules stored in the memory 11 (such as an expander abnormality recognition program based on a substation patrol image, etc.), and calls data stored in the memory 11 to execute various functions of the electronic device 1 and process data.

[0101] The bus may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to realize connection and communication between the memory 11 and at least one processor 10, etc.

[0102] Figure 5 Only an electronic device with components is shown, and those skilled in the art will understand that Figure 5 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0103] For example, although not shown, the electronic device 1 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to the at least one processor 10 through a power management system, so that the power management system can realize functions such as charging management, discharging management, and power consumption management. The power source may also include any components such as one or more DC or AC power sources, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators. The electronic device 1 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0104] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0105] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode) touch device. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0106] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0107] The expander abnormality identification program 12 based on the substation patrol image stored in the memory 11 in the electronic device 1 is a combination of multiple instructions. When running in the processor 10, it can achieve: obtaining a substation patrol image; using the YOLOv11-Large model for expander position detection to perform expander area detection on the substation patrol image, and identify and extract the expander area image; using the YOLOv11-Large model for expander defect detection to perform expander defect detection on the expander area image; wherein the YOLOv11-Large model for expander defect detection is obtained after training with expander defect expansion samples, and the expander defect expansion samples are obtained by expanding the expander area image with expander defect samples using a diffusion model; and identifying and outputting the expander abnormality category.

[0108] Specifically, the specific implementation method of the processor 10 for the above instructions can refer to Figure 1 The description of the relevant steps in the corresponding embodiment is not repeated here. Furthermore, if the module / unit integrated in the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable medium may include: any entity or system capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, and a read-only memory (ROM).

[0109] An embodiment of the present invention also provides a computer-readable storage medium, which may be non-volatile or volatile, and stores a computer program, which, when executed by a processor, implements: obtaining a substation inspection image; performing expander area detection on the substation inspection image using a YOLOv11-Large model for expander position detection, and identifying and extracting an expander area image; performing expander defect detection on the expander area image using a YOLOv11-Large model for expander defect detection; wherein the YOLOv11-Large model for expander defect detection is obtained after training with expander defect expansion samples, and the expander defect expansion samples are obtained after expanding the expander area image with expander defect samples using a diffusion model; and identifying and outputting expander abnormality categories.

[0110] Specifically, the specific implementation method when the computer program is executed by the processor can refer to the description of the relevant steps in the expander abnormality identification method based on the substation patrol image in the embodiment, which will not be repeated here.

[0111] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0112] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0113] In addition, each functional module in each embodiment of the present invention may be integrated into one processing unit, each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0114] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0115] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0116] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or systems stated in a system claim can also be implemented by one unit or system through software or hardware. Secondary words are used to indicate names, but not to indicate any particular order.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for identifying expander anomalies based on substation patrol images A method for identifying expander anomalies based on substation patrol images, characterized in that: include, Acquire substation inspection images; Using the YOLOv11-Large model for expander position detection, the substation inspection image is subjected to expander area detection, and the expander area image is identified and extracted; Performing expander defect detection on the expander region image using a YOLOv11-Large model for expander defect detection; wherein the YOLOv11-Large model for expander defect detection is obtained by training with expander defect expansion samples, and the expander defect expansion samples are obtained by expanding the expander region image with expander defect samples using a diffusion model; Identify and output expander anomaly categories.

2. The expander abnormality identification method based on substation patrol images according to claim 1 is characterized in that: The method of expanding the expander defect sample by using a diffusion model to expand the expander region image to obtain the expander defect expansion sample comprises: labeling the expander region image with defect category labels; Classifying the expanded region images into defect categories after annotation, and generating defect category text for the expanded region images of each defect category; The labeled expander region image and the defect category text corresponding to the expander region image are expanded using a diffusion model to obtain expander defect expansion samples.

3. The expander abnormality identification method based on substation inspection image according to claim 2 is characterized in that: Before using the diffusion model to expand the annotated expander area image and the defect category text corresponding to the expander area image to obtain the expander defect expansion sample, it also includes training two low-rank matrices, the product of the two low-rank matrices approximating the model weight increment, so as to achieve fine-tuning of the diffusion model.

4. The expander abnormality identification method based on substation inspection image according to claim 2 is characterized in that: The defect category labels include rust defects, top cover lifting defects and gap defects.

5. The expander abnormality identification method based on substation inspection images according to claim 1 is characterized in that: The YOLOv11-Large model for expander position detection is obtained by training the training data set; Including, during the training process, data augmentation is performed on the training data set, and the data augmentation methods include perspective transformation, translation transformation, angle rotation, mosaic enhancement, Mixup image mixing and RandomAugment.

6. The expander abnormality identification method based on substation inspection images according to claim 1 is characterized in that: The YOLOv11-Large model for expander defect detection is a YOLOv11-Large model with a decoupled head structure without an anchor frame.

7. The expander abnormality identification method based on substation inspection images according to claim 1 is characterized in that: The YOLOv11-Large model for expander defect detection is obtained after training based on the loss function; The loss function is obtained by the following formula: L=λ1Loss CIoU +λ1Loss DFL +λ2Loss BCE Among them, λ1 and λ2 are hyperparameters; Loss BCE is the loss function of the classification branch; Loss DFL and Loss CIoU is the loss function of the regression branch.

8. An expander anomaly recognition system based on substation patrol images, characterized in that: The expander abnormality identification method based on the substation patrol image according to any one of claims 1 to 7 is used to identify the expander abnormality, and the system includes: An image acquisition unit, used for acquiring substation inspection images; A regional image acquisition unit, used to perform expander region detection on the substation patrol image using a YOLOv11-Large model for expander position detection, and identify and extract an expander region image; A defect recognition unit is used to perform expander defect detection on the expander area image using the YOLOv11-Large model for expander defect detection; wherein the YOLOv11-Large model for expander defect detection is obtained after training with expander defect expansion samples, and the expander defect expansion samples are obtained after expanding the expander area image with expander defect samples using a diffusion model; and identify and output expander abnormality categories.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, and an expander abnormality identification program based on substation patrol images stored in the memory and executable on the processor. When the expander abnormality identification program based on substation patrol images is executed by the processor, the steps of the expander abnormality identification method based on substation patrol images as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the expander abnormality identification method based on substation patrol images as described in any one of claims 1 to 7 is implemented.

Images