Electrical equipment infrared image abnormity identification method and system based on parallel operation characteristics

Abstract

The invention provides an electrical equipment infrared image anomaly recognition method and system based on parallel operation characteristics, and belongs to the technical field of image anomaly recognition. And processing the obtained infrared image by using a pre-trained anomaly identification model to obtain an identification result of whether the electrical equipment is abnormal or not. According to the invention, a YOLO target detection algorithm is utilized to accurately segment key parts of equipment, and then an improved region growing algorithm is combined to position an abnormal temperature region in an infrared image. By fully utilizing the high consistency of the two parallel devices in electrical response and thermodynamic behavior, the local temperature abnormal region can be accurately identified even under the condition of no fault sample, and a new technical support is provided for online state monitoring and intelligent operation and maintenance of the electrical device.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image anomaly recognition, and in particular to an electrical equipment infrared image anomaly recognition method and system based on parallel operation characteristics. BACKGROUND

[0002] The traction power supply system is the core energy hub of urban rail transit, and the operation state of the key power equipment directly affects the safe and reliable operation of the whole system. In the long-term service process, due to factors such as electrical insulation deterioration and mechanical stress accumulation, abnormal temperature rise may occur at the key parts of the equipment, which may cause a chain of failures. The infrared thermal imaging technology has become an important means of equipment state monitoring due to its non-contact, real-time and high sensitivity. However, the traditional infrared thermal fault detection method mainly relies on manual inspection, and therefore has the problems of low efficiency and high false detection rate.

[0003] Deep learning technology has strong signal processing capability and is widely used in infrared image fault diagnosis of electrical equipment. However, the recognition algorithm based on deep learning needs a large number of fault samples to train the model, but the fault samples are extremely scarce in actual operation, which to some extent limits the actual application effect of the deep learning method.

[0004] In the traction power supply system of urban rail transit, two same type devices (such as transformers, rectifiers, switch cabinets, etc.) usually adopt parallel operation mode, and their electrical parameters, load conditions and operating environments are highly consistent, and they are coupled with each other, so that the two parallel devices show strong comparability in electrical response and thermodynamic behavior. Theoretical analysis and engineering practice show that under normal operating conditions, the temperature distribution of the two parallel devices is basically consistent, and once one of them fails, the local temperature field will deviate from the normal state, which provides a solid theoretical basis for anomaly detection based on difference analysis. By comparing the temperature symmetry difference between the parallel devices, the dependence of the model on the fault sample data is significantly reduced. SUMMARY

[0005] The present application aims to provide an electrical equipment infrared image anomaly recognition method and system based on parallel operation characteristics to solve at least one of the technical problems in the background.

[0006] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0007] In a first aspect, the present application provides an electrical equipment infrared image anomaly recognition method based on parallel operation characteristics, comprising:

[0008] obtaining an infrared image of an electrical equipment in parallel operation;

[0009] The acquired infrared image is processed using a pre-trained anomaly recognition model to obtain an identification result of whether the electrical equipment is abnormal; wherein, the anomaly recognition model includes a segmentation network and a recognition network; the segmentation network is used to segment the key components of the electrical equipment using the YOLOv8 target detection algorithm; the recognition network is used to detect abnormal temperatures of the electrical equipment using an improved region growing algorithm, and the improved region growing algorithm includes an initialization stage, a region growing stage and a post-processing stage. In the initialization stage, adaptive selection of seed points is performed; in the region growing stage, the temperature mean and standard deviation are calculated in a local window with the current seed point as the center, and the growth threshold is dynamically adjusted in combination with the temperature gradient; in the post-processing stage, the preliminary segmentation results are optimized, and overly small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference.

[0010] As a further limitation of the first aspect of the present invention, a YOLOv8 target detection algorithm is used to segment key components of electrical equipment, including: using the COCO dataset to pre-train the YOLOv8 model, setting the number of training rounds to 100 rounds and the batch size to 16, and saving the best weights of all optimal rounds after the training is completed; using LabelImg External Tools to label the collected electrical equipment infrared image dataset, constructing a labeled dataset in YOLO format, and inputting the dataset into the model for fine-tuning training, setting the number of fine-tuning rounds to 50 rounds and the batch size to 32, thereby improving the recognition accuracy of the model for the specific task of electrical equipment infrared image recognition and achieving accurate segmentation of electrical equipment components.

[0011] As a further limitation of the first aspect of the present invention, in the initialization stage, adaptive selection of seed points is performed, including: converting the infrared temperature matrix T(x, y) into the frequency domain through a two-dimensional Fourier transform to generate a complex spectrum containing spatial frequency distribution information, taking a logarithmic operation on the spectrum amplitude and fitting its smooth component with a Gaussian smoothing kernel; by calculating the residual between the logarithmic spectrum and the smooth component, the high-frequency component characterizing the local abnormal temperature change is separated.

[0012] As a further limitation of the first aspect of the present invention, after the residual is inverse Fourier transformed, a heat map reflecting the significance of the temperature distribution is generated, and the pixel areas with the highest significance values ​​are extracted as primary candidate areas. The pixel temperature values ​​p(x, y) in the candidate areas are subjected to multimodal analysis using the K-means clustering algorithm; the cluster centers are iteratively optimized until the convergence conditions are met; after locking the category C3 with the highest temperature as the target cluster, the high-temperature pixel points at the top of the temperature distribution are extracted as candidate seeds, and the connectivity of their spatial distribution is verified, the local temperature consistency of the candidate seeds is tested, isolated points are eliminated, and finally the points that pass the verification are used as initial seeds of the growth algorithm.

[0013] As a further limitation of the first aspect of the present invention, in the region growing stage, with the current seed point as the center, a 7×7 local window is taken to calculate the temperature mean and standard deviation, and the growth threshold is dynamically adjusted in combination with the temperature gradient; for the pixel p(x, y) to be grown, a temperature and gradient dual feature fusion criterion is introduced to perform similarity judgment. When S(p) < 1, the pixel is determined to belong to the abnormal area and is included in the growth range; during the growth process, if the current pixel gradient amplitude || G p (x,y)||Exceeds the dynamic threshold G th , and there are at least 3 strong edge points in its 8-neighborhood, then the growth in this direction is terminated to prevent cross-region mismerging.

[0014] As a further limitation of the first aspect of the present invention, in the post-processing stage, the preliminary segmentation results are optimized, and excessively small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference: if the number of pixels contained in a region is less than a set threshold, the region is removed.

[0015] In a second aspect, the present invention provides an electrical equipment infrared image anomaly recognition system based on parallel operation characteristics, comprising:

[0016] An acquisition module, used for acquiring infrared images of electrical equipment operating in parallel;

[0017] The processing module is used to process the acquired infrared image using a pre-trained anomaly recognition model to obtain an identification result of whether the electrical equipment is abnormal; wherein, the anomaly recognition model includes a segmentation network and a recognition network; the segmentation network is used to segment the key components of the electrical equipment using the YOLOv8 target detection algorithm; the recognition network is used to detect abnormal temperatures of the electrical equipment using an improved region growing algorithm, and the improved region growing algorithm includes an initialization stage, a region growing stage and a post-processing stage. In the initialization stage, adaptive selection of seed points is performed; in the region growing stage, the temperature mean and standard deviation are calculated in a local window with the current seed point as the center, and the growth threshold is dynamically adjusted in combination with the temperature gradient; in the post-processing stage, the preliminary segmentation results are optimized, and overly small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference.

[0018] In a third aspect, the present invention provides a non-transitory computer-readable storage medium, which is used to store computer instructions. When the computer instructions are executed by a processor, the method for identifying anomalies in infrared images of electrical equipment based on parallel operation characteristics as described in the first aspect is implemented.

[0019] In a fourth aspect, the present application provides a computer device, comprising a memory and a processor, the processor and the memory being in communication with each other, the memory storing program instructions executable by the processor, and the processor invoking the program instructions to execute the method for identifying infrared image abnormalities of electrical equipment based on parallel operation characteristics according to the first aspect.

[0020] In a fifth aspect, the present application provides an electronic device, comprising a processor, a memory and a computer program, wherein the processor is connected with the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to make the electronic device execute instructions for realizing the method for identifying infrared image abnormalities of electrical equipment based on parallel operation characteristics according to the first aspect.

[0021] The present application has the following advantages: the YOLO target detection algorithm is used to accurately segment the key components of the equipment, and then the improved region growing algorithm is combined to locate the abnormal temperature region in the infrared image. By fully utilizing the high consistency of the two parallel devices in electrical response and thermodynamic behavior, even under the condition of no fault sample, the local temperature abnormal region can be accurately identified, which provides new technical support for the online state monitoring and intelligent operation and maintenance of electrical equipment.

[0022] The advantages of the additional aspects of the present application will be more apparent from the following description section or will be understood through the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0024] Figure 1 The flowchart of the method for identifying infrared image abnormalities of electrical equipment based on parallel operation characteristics according to the embodiments of the present application.

[0025] Figure 2 The strategy diagram of the infrared image abnormality detection of the "comparison-identification" according to the embodiments of the present application.

[0026] Figure 3 The flowchart of the improved region growing algorithm according to the embodiments of the present application.

[0027] Figure 4 The flowchart of the adaptive seed point selection according to the embodiments of the present application. DETAILED DESCRIPTION

[0028] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and are not to be construed as limiting the present invention.

[0029] Those skilled in the art will understand that unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs.

[0030] It should also be understood that terms, such as those defined in commonly used dictionaries, should be understood to have a meaning consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless as defined herein.

[0031] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a," "an," "said," and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the stated features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, and / or groups thereof.

[0032] In the description of this specification, reference to the terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Those skilled in the art may combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless otherwise contradictory.

[0033] To facilitate understanding of the present invention, the present invention is further explained below with reference to specific embodiments in conjunction with the accompanying drawings. However, the specific embodiments do not constitute a limitation on the embodiments of the present invention.

[0034] Those skilled in the art should understand that the drawings are merely schematic diagrams of embodiments, and the components in the drawings are not necessarily necessary for implementing the present invention.

[0035] Example 1

[0036] In this embodiment 1, an electrical equipment infrared image anomaly recognition system based on parallel operation characteristics is first provided, including: an acquisition module for acquiring infrared images of electrical equipment operating in parallel; a processing module for processing the acquired infrared images using a pre-trained anomaly recognition model to obtain an identification result of whether the electrical equipment is abnormal; wherein, the anomaly recognition model includes a segmentation network and a recognition network; the segmentation network is used to segment the key components of the electrical equipment using the YOLOv8 target detection algorithm; the recognition network is used to detect abnormal temperatures of the electrical equipment using an improved region growing algorithm, and the improved region growing algorithm includes an initialization stage, a region growing stage and a post-processing stage. In the initialization stage, adaptive selection of seed points is performed; in the region growing stage, the temperature mean and standard deviation are calculated in a local window with the current seed point as the center, and the growth threshold is dynamically adjusted in combination with the temperature gradient; in the post-processing stage, the preliminary segmentation results are optimized, and excessively small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference.

[0037] In this embodiment, the above-mentioned system is used to implement an abnormality recognition method for infrared images of electrical equipment based on parallel operation characteristics, including: obtaining infrared images of electrical equipment running in parallel; processing the obtained infrared images using a pre-trained abnormality recognition model to obtain an identification result of whether the electrical equipment is abnormal; wherein, the abnormality recognition model includes a segmentation network and a recognition network; the segmentation network is used to segment the key components of the electrical equipment using the YOLOv8 target detection algorithm; the recognition network is used to detect abnormal temperatures of the electrical equipment using an improved region growing algorithm, and the improved region growing algorithm includes an initialization stage, a region growing stage and a post-processing stage. In the initialization stage, adaptive selection of seed points is performed; in the region growing stage, the temperature mean and standard deviation are calculated in a local window with the current seed point as the center, and the growth threshold is dynamically adjusted in combination with the temperature gradient; in the post-processing stage, the preliminary segmentation results are optimized, and excessively small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference.

[0038] The YOLOv8 target detection algorithm was used to segment key components of electrical equipment. This included pre-training the YOLOv8 model using the COCO dataset, setting the number of training rounds to 100 and the batch size to 16. After training, the optimal weights from all rounds were saved. LabelImg External Tools was used to annotate the collected infrared image dataset of electrical equipment, constructing a labeled dataset in YOLO format, and inputting it into the model for fine-tuning training. The fine-tuning number of rounds was set to 50 and the batch size was set to 32. This improved the model's recognition accuracy for the specific task of electrical equipment infrared image recognition and achieved accurate segmentation of electrical equipment components.

[0039] During the initialization phase, adaptive seed point selection is performed, including: converting the infrared temperature matrix T(x,y) into the frequency domain through a two-dimensional Fourier transform to generate a complex spectrum containing spatial frequency distribution information, taking the logarithm of the spectrum amplitude and fitting its smooth component with a Gaussian smoothing kernel; and separating the high-frequency components representing local abnormal temperature changes by calculating the residual between the logarithmic spectrum and the smooth component.

[0040] After performing an inverse Fourier transform on the residual, a heat map reflecting the significance of the temperature distribution is generated. The pixel areas with the highest significance values ​​are extracted as primary candidate areas. The pixel temperature values ​​p(x, y) in the candidate areas are subjected to multimodal analysis using the K-means clustering algorithm. The cluster centers are iteratively optimized until the convergence conditions are met. After locking the category C3 with the highest temperature as the target cluster, the high-temperature pixel points at the top of the temperature distribution are extracted as candidate seeds, and their spatial distribution is verified for connectivity. The local temperature consistency of the candidate seeds is tested, and isolated points are eliminated. Finally, the points that pass the verification are used as the initial seeds of the growth algorithm.

[0041] In the region growing stage, with the current seed point as the center, a 7×7 local window is taken to calculate the temperature mean and standard deviation, and the growth threshold is dynamically adjusted in combination with the temperature gradient. For the pixel p(x, y) to be grown, the temperature and gradient dual feature fusion criterion is introduced to perform similarity judgment. When S(p)<1, the pixel is judged to belong to the abnormal area and included in the growth range. During the growth process, if the current pixel gradient amplitude || G p (x,y)||Exceeds the dynamic threshold G th , and there are at least 3 strong edge points in its 8-neighborhood, then the growth in this direction is terminated to prevent cross-region mismerging.

[0042] In the post-processing stage, the preliminary segmentation results are optimized, and too small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference: if the number of pixels contained in a region is less than the set threshold, the region is removed.

[0043] Example 2

[0044] This embodiment provides a method for identifying anomalies in infrared images of equipment based on parallel operation characteristics, which enables precise positioning of thermal fault areas in infrared images of electrical equipment. The method comprises the following steps: S1, through a fault simulation experiment, collects infrared images of electrical equipment under different early-stage fault conditions and constructs a corresponding infrared image dataset. S2, preprocesses the original infrared image dataset and appropriately amplifies the dataset. S3, constructs a model for identifying anomalies in infrared images of equipment based on parallel operation characteristics. S4, inputs the infrared image dataset into the constructed model to verify the model's performance in detecting abnormal areas in infrared images.

[0045] Furthermore, the infrared image dataset in S1 includes infrared image data of common electrical and mechanical faults, as well as normal operating conditions. In S2, the raw data is preprocessed using temperature-driven cropping technology, and the image resolution is uniformly adjusted to 512×384. The data augmentation method in S2 specifically involves image augmentation using mirroring and rotation techniques. In S3, based on the consistency of parallel operation of electrical equipment, a "contrast-recognition" anomaly detection strategy is proposed. Key components are segmented using the YOLOv8 object detection algorithm, and abnormal regions are identified using an improved region growing algorithm. This results in a device infrared image anomaly recognition model based on parallel operation characteristics. In S4, this model is used to identify abnormal regions in infrared images of electrical equipment. Based on the consistent thermodynamic behavior of parallel electrical equipment, if two parallel devices are operating normally, no abnormal region will be identified. However, if one of the devices experiences a thermal failure, an overheating region will be identified.

[0046] like Figure 1 As shown, the device infrared image anomaly recognition method based on the parallel operation characteristics proposed in this embodiment specifically includes the following steps:

[0047] Step 1: Use an infrared thermal imager to capture infrared images of electrical equipment under typical electrical faults, mechanical faults, and normal operation, and construct a corresponding infrared image dataset. The sampling interval is set to 2 minutes to ensure coverage of the early temperature rise stage of fault development. The collected original infrared image contains temperature matrix data, and the temperature value T of each pixel point p(x,y) is p Calculated by the following formula:

[0048] I p =0.299M r +0.587M g +0.114M b

[0049] b=T max -kY max

[0050] T p =k·I p +b

[0051] Among them, I p is the pixel gray value, M i is the matrix of the extracted color channels, r, g, b represent the red, green and blue color channels respectively. k and b are the linear coefficients corresponding to the actual temperature value and the image grayscale value, T max and T min are the maximum and minimum temperature values ​​in the thermal imager software, Y max and Ymin The maximum and minimum values in the grayscale image.

[0052] Step two: Firstly, in order to solve the problem of background interference in the original infrared image, the temperature distribution histogram of the whole infrared image is calculated and analyzed, and the temperature T p The top 10% high-temperature pixel area;

[0053] Then, morphological closing operation is performed on the high-temperature pixels to connect the discrete areas, and the largest connected region is extracted as the main position of the electrical equipment. The formula of morphological closing operation is:

[0054]

[0055] wherein, represents the expansion operation, ° represents the corrosion operation, and B is a 5x5 circular structural element.

[0056] Then, the minimum circumscribed rectangle is generated, and a 10% boundary is expanded around the center of the rectangle to complete the initial cropping, and the main body of the electrical equipment and the key heat components are retained.

[0057] On this basis, geometric expansion operation is performed on the cropped non-standard size image: through horizontal mirror reflection simulation of different observation angles, random rotation within ±60° is applied, and bicubic interpolation is used to maintain temperature continuity to avoid introducing false temperature values. The interpolation formula is:

[0058]

[0059] wherein, w ij (x, y) is the weight coefficient of bicubic interpolation.

[0060] Finally, all the expanded images are uniformly scaled to 512x384 resolution, which provides higher pixel density and retains more target feature details without changing the proportion of the infrared image.

[0061] Step three: Since the two parallel electrical equipment have high similarity in design parameters and load distribution. This characteristic makes the two parallel equipment in the same substation be regarded as "twins" in design, operation and performance. Therefore, an "contrast-identification" abnormality detection strategy is proposed, and an electrical equipment infrared image abnormality recognition model is built based on this strategy, as shown in Figure 2 .

[0062] The YOLOv8 target detection algorithm is used to segment the key components of the electrical equipment, which lays the foundation for the comparison of the same type and position infrared images of the two parallel electrical equipment. The specific steps of segmentation are as follows:

[0063] First, the COCO dataset is used to pre-train the YOLOv8 model. The number of training rounds is set to 100 and the batch size is set to 16. After the training is completed, the best weights of all rounds are saved.

[0064] Next, we used LabelImg External Tools to annotate the collected infrared image dataset of electrical equipment, constructing a labeled dataset in YOLO format. This dataset was then input into the model for fine-tuning training. We set the number of fine-tuning rounds to 50 and the batch size to 32. This improved the model's recognition accuracy for the specific task of electrical equipment infrared image recognition and enabled accurate segmentation of electrical equipment components.

[0065] After completing the above-mentioned component segmentation, the improved region growing algorithm is used to detect abnormal temperature of electrical equipment. The region growing algorithm mainly includes three stages: initialization, region growing, and post-processing. The specific process is as follows: Figure 3 shown.

[0066] In the initialization stage, the seed point is adaptively selected. The specific process is as follows: Figure 4 First, the infrared temperature matrix T(x,y) after the above processing is converted to the frequency domain through two-dimensional Fourier transform to generate a complex spectrum containing spatial frequency distribution information.

[0067]

[0068] Where (u,v) is the frequency domain coordinate and F(u,v) is the complex spectrum.

[0069] On this basis, the spectrum amplitude is logarithmically calculated and the Gaussian smoothing kernel G is used. σ (σ=3) fitting its smooth component L smooth .

[0070] L smooth (u,v)=log|F(u,v)|×G σ

[0071] By calculating the residual R between the logarithmic spectrum and the smoothed component, the high-frequency components representing the local abnormal temperature change can be effectively separated.

[0072] R(u,v)=log|F(u,v)|-L smooth (u,v)

[0073] After performing inverse Fourier transform on the residual, a heat map reflecting the significance of temperature distribution is generated, and the pixel areas with the top 20% significance value S are extracted as primary candidate areas.

[0074] S(x,y)=Re[F -1 {exp(R(u,v)+i·φ(u,v)}]

[0075] Where Φ(u,v) is the original phase spectrum.

[0076] After obtaining the salient candidate region, the pixel temperature value p(x,y) in the candidate region is subjected to multimodal analysis using the K-means clustering algorithm, and its objective function is:

[0077]

[0078] Among them, C k represents the kth cluster, μ k is the cluster center temperature, and the three cluster center background temperature zones, normal heating zones and potential fault zones are set to correspond to μ1=T 50％ 、μ2=T 75％ 、μ3=T 90％ .

[0079] The cluster centers are optimized iteratively until the convergence conditions are met:

[0080]

[0081] After locking the category C3 with the highest temperature as the target cluster, we extract the top 5% of the extreme high temperature pixels P in the temperature distribution. candidate As candidate seeds:

[0082] P candidate ={T p ∈C3|T p >μ3+1.645σ3}

[0083] The connectivity of its spatial distribution is verified, the consistency of the local temperature of the candidate seeds is tested, and isolated points are eliminated:

[0084] and

[0085] It is required that the temperature fluctuation in the 5×5 area N5 around the seed point does not exceed 3°C, and the average temperature reaches the global mean, which is the global average temperature μ global More than 1.2 times of the original value, thus eliminating the interference of environmental heat sources, and finally the verified points are used as the initial seeds of the growth algorithm.

[0086] Then enter the region growing stage. Taking the current seed point as the center, take a 7×7 local window to calculate the temperature mean μ local and standard deviation σ local , dynamically adjust the growth threshold in combination with the temperature gradient:

[0087]

[0088] in, Through G x and G y Convolved with the temperature image.

[0089]

[0090] For the pixel p(x,y) to be grown, the temperature and gradient dual feature fusion criteria are introduced to perform similarity judgment. When S(p)<1, the pixel is determined to belong to the abnormal area and included in the growth range.

[0091]

[0092]

[0093] Among them, T seed is the average temperature of the current area, cosθ is the cosine similarity between the pixel gradient direction and the regional average gradient direction, and |||| is the gradient amplitude.

[0094] During the growth process, if the current pixel gradient amplitude ||G p (x,y)||Exceeds the dynamic threshold G th , and there are at least 3 strong edge points in its 8-neighborhood, then the growth in this direction is terminated immediately to prevent cross-region mismerging.

[0095] G th =0.2×max(G)

[0096] Where G is the Sobel gradient magnitude of the entire image.

[0097] Finally, we enter the post-processing stage to optimize the initial segmentation results, remove the small growth areas, and avoid the inclusion of abnormal neighborhoods caused by noise interference. k The number of pixels contained in is A k , if the area is less than the set threshold A min , then remove the area.

[0098]

[0099]

[0100] Where W×H is the image resolution, N scale is the scaling factor and β is the safety factor.

[0101] After constructing the region growing algorithm, it was applied to the abnormal temperature detection of infrared images of parallel electrical equipment, thereby building an equipment infrared image abnormality recognition model based on parallel operation characteristics.

[0102] Step 4: First, read the original infrared images A and B of the two electrical devices running in parallel and perform preprocessing such as image cropping. Next, use the YOLO algorithm to segment key components. Then, normalize the segmented images and generate a differential image by subtracting them. After obtaining the differential image, input the differential image into the region growing algorithm for recognition. Based on the consistency of the thermodynamic behavior of electrical devices running in parallel, if both parallel devices are normal, the infrared images of A and B are the same, and the region growing algorithm should not divide any regions. However, if one of the devices has a thermal fault, the region growing algorithm will automatically divide the pixels into different regions, that is, the overtemperature region is divided and identified.

[0103] Example 3

[0104] This embodiment 3 provides a non-transitory computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics as described above is implemented. The method includes:

[0105] Acquire infrared images of electrical equipment operating in parallel;

[0106] The acquired infrared image is processed using a pre-trained anomaly recognition model to obtain an identification result of whether the electrical equipment is abnormal; wherein, the anomaly recognition model includes a segmentation network and a recognition network; the segmentation network is used to segment the key components of the electrical equipment using the YOLOv8 target detection algorithm; the recognition network is used to detect abnormal temperatures of the electrical equipment using an improved region growing algorithm, and the improved region growing algorithm includes an initialization stage, a region growing stage and a post-processing stage. In the initialization stage, adaptive selection of seed points is performed; in the region growing stage, the temperature mean and standard deviation are calculated in a local window with the current seed point as the center, and the growth threshold is dynamically adjusted in combination with the temperature gradient; in the post-processing stage, the preliminary segmentation results are optimized, and overly small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference.

[0107] Example 4

[0108] This embodiment 4 provides a computer device, including a memory and a processor, wherein the processor and the memory communicate with each other, the memory stores program instructions executable by the processor, and the processor invokes the program instructions to execute the above-described method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics, the method comprising:

[0109] Acquire infrared images of electrical equipment operating in parallel;

[0110] The acquired infrared image is processed using a pre-trained anomaly recognition model to obtain an identification result of whether the electrical equipment is abnormal; wherein, the anomaly recognition model includes a segmentation network and a recognition network; the segmentation network is used to segment the key components of the electrical equipment using the YOLOv8 target detection algorithm; the recognition network is used to detect abnormal temperatures of the electrical equipment using an improved region growing algorithm, and the improved region growing algorithm includes an initialization stage, a region growing stage and a post-processing stage. In the initialization stage, adaptive selection of seed points is performed; in the region growing stage, the temperature mean and standard deviation are calculated in a local window with the current seed point as the center, and the growth threshold is dynamically adjusted in combination with the temperature gradient; in the post-processing stage, the preliminary segmentation results are optimized, and overly small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference.

[0111] Example 5

[0112] This embodiment 5 provides an electronic device, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to cause the electronic device to execute instructions for implementing the above-mentioned method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics. The method includes:

[0113] Acquire infrared images of electrical equipment operating in parallel;

[0114] The acquired infrared image is processed using a pre-trained anomaly recognition model to obtain an identification result of whether the electrical equipment is abnormal; wherein, the anomaly recognition model includes a segmentation network and a recognition network; the segmentation network is used to segment the key components of the electrical equipment using the YOLOv8 target detection algorithm; the recognition network is used to detect abnormal temperatures of the electrical equipment using an improved region growing algorithm, and the improved region growing algorithm includes an initialization stage, a region growing stage and a post-processing stage. In the initialization stage, adaptive selection of seed points is performed; in the region growing stage, the temperature mean and standard deviation are calculated in a local window with the current seed point as the center, and the growth threshold is dynamically adjusted in combination with the temperature gradient; in the post-processing stage, the preliminary segmentation results are optimized, and overly small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference.

[0115] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0116] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0117] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0118] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide the functions for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0119] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solutions disclosed in the present invention without the need for creative work should be included in the scope of protection of the present invention.

Claims

1. A method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics, characterized in that: include: Acquire infrared images of electrical equipment operating in parallel; The acquired infrared image is processed using a pre-trained anomaly recognition model to obtain an identification result of whether the electrical equipment is abnormal; wherein, the anomaly recognition model includes a segmentation network and a recognition network; the segmentation network is used to segment the key components of the electrical equipment using the YOLOv8 target detection algorithm; the recognition network is used to detect abnormal temperatures of the electrical equipment using an improved region growing algorithm, and the improved region growing algorithm includes an initialization stage, a region growing stage and a post-processing stage. In the initialization stage, adaptive selection of seed points is performed; in the region growing stage, the temperature mean and standard deviation are calculated in a local window with the current seed point as the center, and the growth threshold is dynamically adjusted in combination with the temperature gradient; in the post-processing stage, the preliminary segmentation results are optimized, and overly small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference.

2. The method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics according to claim 1, characterized in that: The YOLOv8 object detection algorithm is used to segment key components of electrical equipment. This includes: pre-training the YOLOv8 model using the COCO dataset, setting the number of training rounds to 100 and the batch size to 16. After training, the best weights from all rounds are saved. The collected infrared image dataset of electrical equipment is annotated using LabelImg External Tools to construct a labeled dataset in YOLO format. This dataset is then input into the model for fine-tuning training, with a number of fine-tuning rounds of 50 and a batch size of 32. This improves the model's recognition accuracy for the specific task of electrical equipment infrared image recognition and achieves accurate segmentation of electrical equipment components.

3. The method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics according to claim 1, characterized in that: During the initialization phase, adaptive seed point selection is performed, including: converting the infrared temperature matrix T(x,y) into the frequency domain through a two-dimensional Fourier transform to generate a complex spectrum containing spatial frequency distribution information, taking the logarithm of the spectrum amplitude and fitting its smooth component with a Gaussian smoothing kernel; and separating the high-frequency components representing local abnormal temperature changes by calculating the residual between the logarithmic spectrum and the smooth component.

4. The method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics according to claim 3, characterized in that: After performing an inverse Fourier transform on the residual, a heat map reflecting the significance of the temperature distribution is generated. The pixel areas with the highest significance values ​​are extracted as primary candidate areas. The pixel temperature values ​​p(x, y) in the candidate areas are subjected to multimodal analysis using the K-means clustering algorithm. The cluster centers are iteratively optimized until the convergence conditions are met. After locking the category C3 with the highest temperature as the target cluster, the high-temperature pixel points at the top of the temperature distribution are extracted as candidate seeds, and their spatial distribution is verified for connectivity. The local temperature consistency of the candidate seeds is tested, and isolated points are eliminated. Finally, the points that pass the verification are used as the initial seeds of the growth algorithm.

5. The method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics according to claim 1, characterized in that: In the region growing stage, with the current seed point as the center, a 7×7 local window is taken to calculate the temperature mean and standard deviation, and the growth threshold is dynamically adjusted in combination with the temperature gradient. For the pixel p(x, y) to be grown, the temperature and gradient dual feature fusion criterion is introduced to perform similarity judgment. When S(p)<1, the pixel is judged to belong to the abnormal area and included in the growth range. During the growth process, if the current pixel gradient amplitude || G p (x,y)||Exceeds the dynamic threshold G th , and there are at least 3 strong edge points in its 8-neighborhood, then the growth in this direction is terminated to prevent cross-region mismerging.

6. The method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics according to claim 1, characterized in that: In the post-processing stage, the preliminary segmentation results are optimized, and too small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference: if the number of pixels contained in a region is less than the set threshold, the region is removed.

7. An electrical equipment infrared image anomaly recognition system based on parallel operation characteristics, characterized in that: include: An acquisition module, used for acquiring infrared images of electrical equipment operating in parallel; The processing module is used to process the acquired infrared image using a pre-trained anomaly recognition model to obtain an identification result of whether the electrical equipment is abnormal; wherein, the anomaly recognition model includes a segmentation network and a recognition network; the segmentation network is used to segment the key components of the electrical equipment using the YOLOv8 target detection algorithm; the recognition network is used to detect abnormal temperatures of the electrical equipment using an improved region growing algorithm, and the improved region growing algorithm includes an initialization stage, a region growing stage and a post-processing stage. In the initialization stage, adaptive selection of seed points is performed; in the region growing stage, the temperature mean and standard deviation are calculated in a local window with the current seed point as the center, and the growth threshold is dynamically adjusted in combination with the temperature gradient; in the post-processing stage, the preliminary segmentation results are optimized, and overly small growth areas are removed to avoid the inclusion of neighborhood anomalies caused by noise interference.

8. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, the method for identifying abnormalities in infrared images of electrical equipment based on parallel operation characteristics as described in any one of claims 1 to 6 is implemented.

9. A computer device, characterized in that: It includes a memory and a processor, the processor and the memory communicate with each other, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the infrared image anomaly recognition method of electrical equipment based on parallel operation characteristics as described in any one of claims 1 to 6.

10. An electronic device, characterized in that: include: A processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to execute instructions for implementing the method for identifying anomalies in infrared images of electrical equipment based on parallel operation characteristics as described in any one of claims 1 to 6.

Images