Power equipment defect intelligent identification method, system, equipment and medium

Abstract

The invention discloses a power equipment defect intelligent identification method, system and device and a medium, and the method comprises the steps: collecting an original image of power equipment, and screening the original image of the power equipment to obtain a channel image; carrying out enhancement processing on the channel image and then calculating a gray scale difference value to obtain a gray scale image; converting the gray level image into a frequency spectrum image by adopting fast Fourier transform, and constructing a Gaussian filtering function to carry out convolution and inverse transformation on the frequency spectrum image to obtain a spatial domain image; and performing adaptive threshold segmentation on the spatial domain image, dividing the image into a defect area and a non-defect area, obtaining a segmented image, and performing morphological processing on the segmented image to obtain an electrical equipment defect identification result. According to the method, the problem of aliasing in traditional spatial domain processing is solved, the defect area is accurately extracted, short-time interference and real defects can be effectively distinguished, and the segmentation accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of power equipment technology, and in particular to a method, system, device and medium for intelligent identification of defects in power equipment. Background Technology

[0002] With the advancement of smart grid construction, the safe and stable operation of power equipment (such as transmission lines and substation equipment) has become the core guarantee for reliable power supply, and timely defect identification is a key link in power equipment operation and maintenance. Traditional manual inspection methods are inefficient, costly, and greatly affected by environmental and human factors, making them difficult to meet the operation and maintenance needs of large-scale power grids. Therefore, intelligent defect identification technology for power equipment based on computer vision has become a research hotspot. It achieves automatic defect detection through image or video analysis, which can significantly improve inspection efficiency and accuracy. Especially in scenarios such as drone inspection and real-time monitoring by fixed cameras, video analysis technology can dynamically capture changes in equipment status, providing data support for preventive maintenance of power systems and serving as an important technical means to promote the intelligent transformation of power operation and maintenance.

[0003] Existing technologies for intelligent defect identification in power equipment have significant limitations. First, they rely on spatial domain image processing, which makes it difficult to distinguish between complex surface textures and defect features. This is because high-frequency noise and low-frequency defect signals overlap in the spatial domain, leading to a high rate of missed detection for low-contrast defects. Second, they employ a single-frame independent processing method, lacking modeling of the spatiotemporal relationship between video frames. This makes it impossible to identify defect evolution trends and easily misidentifies transient interferences such as changes in illumination and short-term occlusion as defects. Third, they have poor environmental adaptability. In scenarios with uneven illumination, shooting angle deviations, and differences in equipment models, their stability and robustness are insufficient, requiring frequent parameter tuning or model retraining. Fourth, high-resolution image processing is time-consuming and computationally inefficient, making it difficult to meet the real-time requirements of large-scale inspections and restricting its deployment and application on edge terminals (such as drones). Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an intelligent identification method for power equipment defects to solve the problem that traditional technologies struggle to separate equipment defects from noise.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides an intelligent identification method for defects in power equipment, comprising:

[0008] Acquire raw images of power equipment, and filter the raw images of power equipment to obtain channel images;

[0009] After enhancing the channel image, the grayscale difference is calculated to obtain a grayscale image;

[0010] The grayscale image is converted into a spectral image using a fast Fourier transform, and a Gaussian filter function is constructed to perform convolution and inverse transform on the spectral image to obtain a spatial domain image;

[0011] Adaptive threshold segmentation is performed on the spatial domain image to divide the image into defective and non-defective regions, obtaining segmented images. The segmented images are then subjected to morphological processing to obtain the power equipment defect identification results.

[0012] As a preferred embodiment of the intelligent defect identification method for power equipment described in this invention, acquiring a grayscale image includes:

[0013] The channel image is subjected to mean filtering, and the difference between the mean-filtered gray value and the original gray value of the channel image is calculated to obtain the gray value difference.

[0014] Based on the grayscale difference, a grayscale image is generated by combining the influence factor with the original grayscale value.

[0015] The beneficial effects of this preferred technical solution are as follows: By applying mean filtering to the channel image and calculating the gray-level difference, and then fusing it with the influencing factor and the original gray-level value to generate a gray-level image, the mean filtering can smooth image noise, while the gray-level difference can enhance edge details and defect features. This effectively solves the problem of detail blurring caused by single filtering and enhances the contrast between defect areas and the background.

[0016] As a preferred embodiment of the intelligent identification method for power equipment defects described in this invention, acquiring a spatial domain image includes:

[0017] The low-frequency defect information and high-frequency texture noise in the grayscale image are separated by the Fast Fourier Transform method, and the image is converted from the spatial domain to a spectral image.

[0018] Construct a Gaussian filter function, and perform a convolution operation between the Gaussian filter function and the spectrum image to attenuate high-frequency components and retain low-frequency components;

[0019] A fast inverse Fourier transform is performed on the convolutional spectral image to convert it back to the spatial domain, resulting in a smoothed spatial domain image that highlights the defect features.

[0020] The beneficial effects of this preferred technical solution are as follows: It successfully separates defect and texture signals that are difficult to distinguish in traditional spatial domain processing by using the fast Fourier transform method, specifically removes the interference of complex textures on the surface of the equipment, highlights the overall structural features of the defects, and solves the problem of recognition difficulties caused by the aliasing of low-frequency defect information and high-frequency texture noise.

[0021] As a preferred embodiment of the intelligent defect identification method for power equipment according to the present invention, the method includes: performing adaptive threshold segmentation on the spatial domain image to divide the image into defect regions and non-defect regions, and obtaining a segmented image, including:

[0022] Extract the grayscale values ​​of pixels within the target region of the spatial domain image, and perform statistical analysis and grouping of the pixels based on the grayscale values;

[0023] The optimal segmentation threshold is calculated based on the candidate thresholds and selected as the threshold that maximizes the inter-class variance of different grayscale values. Based on the optimal segmentation threshold, the spatial domain image is divided into defective and non-defective regions to obtain the segmented image.

[0024] The beneficial effects of this preferred technical solution are as follows: By extracting the grayscale values ​​of the target region and selecting the optimal segmentation threshold based on maximizing the inter-class variance, adaptive threshold segmentation is achieved, avoiding the sensitivity of traditional fixed thresholds to changes in illumination and differences in image contrast. It can dynamically adjust the segmentation criteria according to the image grayscale distribution characteristics, accurately dividing defective and non-defective regions, and reducing misjudgments or missed detections caused by fixed parameters.

[0025] As a preferred embodiment of the intelligent identification method for power equipment defects described in this invention, the method includes: statistically analyzing and grouping pixels, including:

[0026] Statistically analyze the frequency of each grayscale value and its proportion in the target area;

[0027] A candidate threshold is set, and the gray values ​​are divided into two groups. The proportion of pixels in the target area, the sum of gray values, and the average gray value of the two groups are calculated respectively to obtain the overall average gray value of the spatial domain image.

[0028] As a preferred embodiment of the intelligent defect identification method for power equipment according to the present invention, the method includes: filtering the original image of the power equipment to obtain channel images, including:

[0029] The original images of the acquired power equipment are preprocessed by using median filtering to remove noise and correct geometric distortion and color deviation in the original images of the power equipment.

[0030] The preprocessed original image of the power equipment is decomposed into three color channels: R, G, and B. The clarity and prominence of the defect area in each channel are compared and analyzed, and the R channel image is selected as the channel image.

[0031] As a preferred embodiment of the intelligent identification method for power equipment defects described in this invention, a grayscale image is generated based on the grayscale difference, combined with the fusion of the influencing factor and the original grayscale value. The calculation formula is expressed as follows:

[0032]

[0033] res=[f(x,y)-g(x,y)]*Factor+f(x,y)

[0034] Where f(x,y) represents the original gray value of each point on the image, G represents the local area covered by the filter kernel, M and N represent the size of the filter kernel, g(x,y) is the new gray value calculated by the mean filtering operation, Factor represents the influence factor, which can adjust the filtering intensity, and res represents the gray image obtained after a series of processing.

[0035] Secondly, the present invention provides an intelligent identification system for defects in power equipment, comprising: an image acquisition module for acquiring original images of power equipment;

[0036] The image processing module is used to filter the original image of the power equipment to obtain channel images; after enhancing the channel images, calculate the gray-level difference to obtain gray-level images; use fast Fourier transform to convert the gray-level images into spectrum images, and construct a Gaussian filter function to perform convolution and inverse transform on the spectrum images to obtain spatial domain images;

[0037] The image recognition module is used to perform adaptive threshold segmentation on the spatial domain image, divide the image into defective and non-defective regions, obtain segmented images, and obtain power equipment defect recognition results through morphological processing of the segmented images.

[0038] Thirdly, the present invention provides an electronic device, comprising:

[0039] Memory and processor;

[0040] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the intelligent identification method for power equipment defects are implemented.

[0041] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the intelligent identification method for power equipment defects.

[0042] Compared with existing technologies, the advantages of this invention are as follows: This invention effectively removes salt-and-pepper noise through median filtering, and then highlights defect features through image enhancement; it employs Fast Fourier Transform and Gaussian filtering to separate high-frequency texture and low-frequency defect information in the frequency domain, solving the problem of aliasing between the two in traditional spatial domain processing; through adaptive segmentation using the Otsu's method and morphological processing, it accurately extracts defect regions, improving segmentation accuracy. Simultaneously, this invention enhances the perception of defect evolution trends in dynamic scenes, effectively distinguishes between short-term interference and real defects, improves robustness and generalization ability in complex environments, reduces dependence on hardware performance, and provides an efficient and reliable solution for intelligent inspection of power equipment. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a schematic diagram of the overall process of the intelligent identification method for power equipment defects according to an embodiment of the present invention. Detailed Implementation

[0045] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0046] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for intelligent identification of defects in power equipment is provided, comprising:

[0047] S100: Acquire raw images of power equipment and filter the raw images of power equipment to obtain channel images;

[0048] S101: After enhancing the channel image, calculate the grayscale difference to obtain a grayscale image;

[0049] S102: The grayscale image is converted into a spectrum image using fast Fourier transform, and a Gaussian filter function is constructed to perform convolution and inverse transform on the spectrum image to obtain the spatial domain image;

[0050] S103: Perform adaptive threshold segmentation on the spatial domain image to divide the image into defective and non-defective regions, obtain segmented images, and then perform morphological processing on the segmented images to obtain the power equipment defect identification results.

[0051] It should be noted that this invention selects and obtains channel images from the original image, selectively retaining the most significant defect features and reducing irrelevant interference. By enhancing processing and calculating grayscale differences to generate grayscale images, the contrast between defects and the background is strengthened, laying a clear feature foundation for subsequent processing. Fast Fourier Transform and Gaussian filtering are used to separate high-frequency texture noise and low-frequency defect information in the frequency domain, solving the recognition problem caused by their aliasing in the traditional spatial domain. Adaptive threshold segmentation combined with morphological processing can dynamically adapt to the image's grayscale distribution, accurately dividing and optimizing defect areas. This invention effectively reduces false positives and false negatives caused by complex textures and changes in lighting, improving the efficiency and reliability of power equipment defect identification, and is suitable for intelligent inspection needs in various scenarios.

[0052] Example 2, refer to Figure 1 As an embodiment of the present invention, based on the above embodiment, a method for intelligent identification of defects in power equipment is provided.

[0053] In this embodiment of the invention, step S100, which involves acquiring original images of power equipment and filtering the original images to obtain channel images, further includes:

[0054] Specifically, to achieve rapid and high-quality image acquisition and provide support for subsequent accurate defect detection and analysis, this invention employs an industrial area scan camera as the image acquisition device. Industrial area scan cameras possess excellent imaging capabilities; when used in conjunction with a matching LED light source, they can capture subtle texture changes and potential defects on the surface of electrical equipment, ensuring clear, shadow-free images even under complex lighting conditions and avoiding interference from reflections on the recognition results.

[0055] The raw images of the acquired power equipment contain noise, which may interfere with subsequent processing. To improve image quality and enhance defect extraction accuracy, preprocessing of the raw images of the power equipment is necessary. Dust and white spots on the power equipment resemble salt-and-pepper noise; to avoid misjudgment in subsequent processing, median filtering is used to remove them.

[0056] For example, median filtering is represented as:

[0057] g(x,y)=med{f(xk,yi)}

[0058]

[0059] Where f(x,y) represents the original image data, g(x,y) represents the image data after being sorted in descending order, and m is the dimension of the filter kernel used in the image processing, which is used for feature extraction or local region smoothing.

[0060] The raw images of power equipment acquired by an industrial area scan camera are RGB images. To improve image quality and processing speed, and to avoid systematic errors, the acquired images are decomposed into three color channels: R, G, and B. During image processing, each color channel can be adjusted and edited individually for more precise color control. After image decomposition, it was found that surface defect areas were most prominent in the R channel; therefore, all subsequent operations are closely focused on the R channel to ensure accurate identification and processing of defect areas.

[0061] In one optional implementation, a high-definition line scan camera can be used to acquire the original image of the power equipment. After the image is acquired, a ring light source is used to eliminate the reflective interference on the surface of the equipment, which is especially suitable for continuous scanning imaging of linear equipment such as long-distance transmission lines. The acquired image is first filtered by bilateral filtering to remove Gaussian noise, and then the local contrast of the image is enhanced by multi-scale Retinex algorithm. Finally, the processed image is decomposed into R, G, and B channels, and the G channel with the clearest defect edge is selected as the channel image for subsequent processing.

[0062] In another alternative implementation, an infrared thermal imaging camera can be used to acquire the original image of the power equipment. After the image is acquired, the area where there may be overheating defects is initially located by temperature threshold segmentation. The infrared image and the visible light image are registered and fused to retain the temperature characteristics of the infrared image and the texture details of the visible light image. The fused image is decomposed into R, G and B channels, and the B channel, which can simultaneously highlight the temperature abnormal area and the structural defect, is selected as the processing object.

[0063] In this embodiment of the invention, step S101, which involves enhancing the channel image and calculating the grayscale difference to obtain a grayscale image, further includes sub-steps A1-A2:

[0064] A1: Perform mean filtering on the channel image, and calculate the difference between the mean-filtered gray value and the original gray value of the channel image to obtain the gray value difference.

[0065] A2: Based on the grayscale difference, the grayscale image is calculated and generated by combining the influence factor with the original grayscale value.

[0066] It should be noted that the smoothing operation in the original image of the power equipment after denoising results in blurred edge details, which is detrimental to subsequent feature filtering. Therefore, this invention employs an emphasis enhancement operation to enhance the original image of the power equipment.

[0067] In this embodiment of the invention, the original image of the power equipment is first subjected to mean filtering. The difference between the mean-filtered gray value and the original gray value is multiplied by an influence factor and then added to the original gray value to make the original image of the power equipment clearer, so as to obtain a grayscale image.

[0068] For example, the formula for calculating a grayscale image is:

[0069]

[0070] res=[f(x,y)-g(x,y)]*Factor+f(x,y)

[0071] Where f(x,y) represents the original grayscale value of each point on the image, directly reflecting the brightness information of the image; G defines a local region covered by the filter kernel, used to analyze the correlation between adjacent pixels in the image; M and N define the size of the filter kernel, affecting the neighborhood range considered during processing; g(x,y) is the new grayscale value calculated through the mean filtering operation, which aims to smooth the image; Factor, as an influencing factor, can adjust the filtering intensity; res represents the enhanced image result after a series of processing steps.

[0072] In one alternative implementation, the enhancement process can also include histogram equalization, sharpening filtering, etc.

[0073] For example, histogram equalization is used to enhance the channel image. By adjusting the distribution range of the image's grayscale values, the overall contrast is improved, making the grayscale difference between the originally blurred defect area and the background more significant.

[0074] For example, Laplacian sharpening filtering is used for enhancement. By calculating the second derivative of the image pixels, edge details are highlighted and the contour features of defects are strengthened. In the identification of minor defects such as cracks on the surface of equipment, the crack edges are clearer, which makes it easier to capture the local feature changes of defects when calculating grayscale differences.

[0075] In this embodiment of the invention, step S102 uses Fast Fourier Transform to convert the grayscale image into a spectral image, constructs a Gaussian filter function to perform convolution and inverse transform on the spectral image to obtain a spatial domain image, and also includes sub-steps B1-B3:

[0076] B1: The fast Fourier transform method is used to separate low-frequency defect information and high-frequency texture noise in grayscale images, converting them from spatial domain to spectral images;

[0077] B2: Construct a Gaussian filter function, and perform a convolution operation between the Gaussian filter function and the spectrum image to attenuate high-frequency components and retain low-frequency components;

[0078] B3: Perform an inverse fast Fourier transform on the convolutional spectral image to convert the spectral image back to the spatial domain, resulting in a smoothed spatial domain image that highlights defect features.

[0079] Specifically, taking a grayscale image f(x,y) of size M×N pixels as an example, it is first subjected to a Fast Fourier Transform to obtain a spectral image F(u,v). The points on the spectral image do not correspond one-to-one with the original image. Points of different brightness represent the intensity of the difference between the grayscale value and the neighborhood in the original image. Low-frequency components tend to cluster in the center of the spectrum and gradually diffuse towards the edge as the frequency increases. This can easily lead to isolated noise points appearing in the edge region.

[0080] When optimizing the spectrogram, attention should be paid to unusual bright spots in the high-frequency region. These are usually visual manifestations of noise and can be effectively filtered out or weakened through technical means to improve the accuracy of signal analysis. Texture details with significant gray-level differences from the surrounding areas appear as high-frequency components (i.e., parts far from the center of the spectrum) and sporadic noise in the spectrogram image.

[0081] For example, the mathematical formula for frequency domain filtering using Fourier transform is shown below:

[0082]

[0083] F(u,v) describes the spatial domain information of the image. Its size is defined by the number of horizontal and vertical pixels M and N. It is a spectral image obtained by fast Fourier transform and reflects the characteristics of the image.

[0084] To reduce the impact of texture details on subsequent processing, a Gaussian function is constructed as the transfer function G(u,v), which is convolved with the spectral image to attenuate high-frequency components, resulting in the filtered spectral image H(u,v), as shown in the following equation:

[0085]

[0086] H(u,v)=G(u,v)F(u,v)

[0087] Where σ is the standard deviation of the Gaussian distribution, which affects the smoothness of the filter.

[0088] The smoothed frequency domain image is processed by inverse fast Fourier transform to convert the processed image back to the spatial domain, thereby achieving image smoothing and enhanced feature representation.

[0089] For example, an image processed in the frequency domain converted back to its spatial domain representation is as follows:

[0090]

[0091] Among them, f new(x,y) is the spatial domain image processed by inverse fast Fourier transform. The parameter σ controls the degree of dispersion of the weight distribution and directly determines the final smoothness of the spectral image.

[0092] If the value of σ is too small, the texture details of the grid lines cannot be completely removed; if the value of σ is too large, the spatial image after the inverse fast Fourier transform will be blurred and cannot be further segmented. Through multiple experiments, it was found that the processing effect is best when σ = 2, which can remove high-frequency texture details while preserving the structure of the defect area.

[0093] In this embodiment of the invention, step S103, which involves adaptive threshold segmentation of the spatial domain image to divide the image into defective and non-defective regions and obtain a segmented image, further includes sub-steps C1-C2:

[0094] C1: Extract the gray values ​​of pixels within the target region of the spatial domain image, and perform statistical analysis and grouping of the pixels based on the gray values;

[0095] C2: Calculate and select the threshold that maximizes the inter-class variance of different gray values ​​based on the candidate thresholds as the optimal segmentation threshold. Based on the optimal segmentation threshold, divide the spatial domain image into defective and non-defective regions to obtain the segmented image.

[0096] It should be noted that the target region refers to a specific analytical area in the spatial domain image that focuses on potential defects in the power equipment; that is, it contains the key area on the surface of the power equipment to be inspected. This region is selected from the spatial domain image after frequency domain processing and covers the parts of the equipment surface where defects may occur.

[0097] Specifically, extract the grayscale values ​​of N pixels within the target area and construct an array G = (g1, g2, ..., g...) covering L grayscale levels. L For each grayscale value g (0 < g < L), its frequency of occurrence is counted as n(g), and the proportion of this grayscale value in the target area is calculated as p(g) = n(g) / N; a threshold t is set (0 ≤ t ≤ L-1), and the grayscale values ​​are divided into two groups G1 and G2, containing N1(t) and N2(t) pixels respectively, as shown in the following formula, thereby realizing grayscale division:

[0098]

[0099] The proportions of pixels in G1 and G2 to the total pixels in the target area are P1(t) and P2(t), respectively, expressed as:

[0100]

[0101] The sums of the pixel grayscale values ​​in G1 and G2 are S1(t) and S2(t), respectively, and are expressed as:

[0102]

[0103] The average gray values ​​in G1 and G2 are M1(t) and M2(t), respectively. The average gray value of the image is M, expressed as:

[0104]

[0105] The final segmentation threshold can be obtained by iterating through t∈(0,L-1) and maximizing the inter-class variance.

[0106] For example, the formula for calculating inter-class variance is:

[0107] σ 2 =P1(t)[M1(t)-M] 2 +P2(t)[M2(t)-M] 2

[0108] In one alternative implementation, the segmented image can be obtained by an iterative segmentation method based on local adaptive thresholds. First, the target region of the spatial domain image is divided into multiple overlapping local windows of size 10×10 pixels, with the overlap set to 50%, to ensure that each pixel is covered by at least two windows.

[0109] For each local window, extract the grayscale values ​​of all pixels within it, calculate the mean μ and standard deviation σ of the grayscale values ​​within the window, and use μ-kσ as the local threshold of the window, where k takes a value of 1.2-1.5 and is adjusted according to the device texture complexity. Traverse all pixels within the target area, count the local threshold of the window in which it is located, and take the average value as the final judgment threshold of the pixel.

[0110] The pixel grayscale value is compared with the corresponding judgment threshold. Pixels below the threshold are marked as defective pixels, and pixels above the threshold are marked as non-defective pixels. After the initial segmentation is completed, 8-neighborhood connectivity analysis is performed on the defective pixel region. Isolated regions with fewer than 30 pixels are judged as noise and removed. Then, a 3×3 structuring element dilation operation is performed on the remaining defective regions to fill the tiny holes in the regions. Finally, the original contour of the defective regions is restored through an erosion operation to obtain a complete segmented image.

[0111] In another alternative implementation, a segmented image can be obtained by a segmentation method based on gray-scale distribution clustering, and the gray values ​​of all pixels in the target region can be extracted to construct a gray-scale dataset. The gray values ​​are then clustered using the K-means clustering algorithm, with K set to 2, corresponding to defective and non-defective regions respectively, and two initial cluster centers are randomly selected.

[0112] Calculate the Euclidean distance between each gray value and the two cluster centers, and assign it to the cluster with the closer distance; recalculate the mean of the two clusters as the new cluster centers, and repeat the assignment and update process until the change in the cluster centers is less than the preset threshold of 0.5; in the final clustering result, the region corresponding to the class with the lower gray value is determined as the defect region, and the region corresponding to the class with the higher gray value is determined as the non-defect region.

[0113] Morphological closing operations are performed on the defective regions obtained by clustering. First, a 3×3 structuring element is expanded, and then erosion of the same size is performed to eliminate small gaps in the region. The contour of the defective region is extracted by the boundary tracking algorithm to obtain the segmented image.

[0114] In this embodiment of the invention, step S103, where the segmented image undergoes morphological processing to obtain the power equipment defect identification result, specifically includes:

[0115] After segmentation, morphological processing is performed on the image. Morphological processing is a commonly used technique in black-and-white binary image processing, which can effectively enhance region boundaries and remove burrs and noise. A closing operation is then performed on the segmented regions to smooth the boundaries without increasing the region area. The closing operation can be viewed as first performing a dilation operation on the image and then an erosion operation, expressed as:

[0116]

[0117] Here, A represents the data file of the binary image, and its internal features can be analyzed in detail through the guidance of the structuring element S; S represents the structuring element of morphological operations, which interacts with A to promote the recognition and extraction of specific morphological structures in the image, and finally obtains the defect extraction image.

[0118] In one alternative implementation, the defect identification results of power equipment can be obtained by a combination of opening and closing operations using a morphological processing method.

[0119] For example, the segmented binary image is processed by opening, using a 3×3 square structuring element. First, an erosion operation is performed to remove burrs and small noise points at the edges of the defect area. Then, a dilation operation is performed to restore the main outline of the defect area, avoiding excessive shrinkage of the area due to erosion. Subsequently, the image after opening is processed by closing, also using a 3×3 square structuring element. First, dilation is performed to fill the tiny holes in the defect area. Then, erosion is performed to maintain the original size of the defect area, finally obtaining a defect recognition result with smooth boundaries and complete interior.

[0120] In another alternative implementation, the defect identification results of power equipment are obtained through a morphological processing method based on multiple structural elements;

[0121] For example, for the segmented binary image, a 5×5 circular structuring element is first used for closing operation. The isotropic property of the circular structuring element is used to smooth the irregular edges of the eroded area at the corner of the equipment, while filling in large-area holes. Then, a 2×2 cross-shaped structuring element is used for opening operation to accurately remove false edges around the defect area caused by segmentation errors. The boundary contour of the processed defect area is extracted by morphological gradient operation. The area and perimeter of the contour are combined to select areas with an area greater than 20 pixels as the final defect. Noise areas with too small an area are removed to obtain the power equipment defect recognition result.

[0122] It should be noted that this invention effectively removes salt-and-pepper noise through median filtering, and then highlights defect features through image enhancement; it employs Fast Fourier Transform and Gaussian filtering to separate high-frequency texture and low-frequency defect information in the frequency domain, solving the problem of aliasing between the two in traditional spatial domain processing; through adaptive segmentation using the Otsu's method and morphological processing, it accurately extracts defect regions, improving segmentation accuracy. Simultaneously, this invention enhances the perception of defect evolution trends in dynamic scenes, effectively distinguishes between short-term interference and real defects, improves robustness and generalization ability in complex environments, reduces dependence on hardware performance, and provides an efficient and reliable solution for intelligent inspection of power equipment.

[0123] Example 3 illustrates a schematic scheme for an intelligent identification method for power equipment defects. It should be noted that the technical solution of this intelligent identification system for power equipment defects belongs to the same concept as the technical solution of the aforementioned intelligent identification method for power equipment defects. Details not described in detail in this embodiment can be found in the description of the technical solution of the aforementioned intelligent identification method for power equipment defects.

[0124] This embodiment also provides an intelligent identification system for power equipment defects, including:

[0125] The image acquisition module is used to acquire raw images of power equipment;

[0126] The image processing module is used to filter the original images of power equipment to obtain channel images; after enhancing the channel images, the gray-level difference is calculated to obtain gray-level images; the gray-level images are converted into spectrum images using fast Fourier transform, and a Gaussian filter function is constructed to perform convolution and inverse transform on the spectrum images to obtain spatial domain images.

[0127] The image recognition module is used to perform adaptive threshold segmentation on spatial domain images, dividing the images into defective and non-defective regions, obtaining segmented images, and then performing morphological processing on the segmented images to obtain the power equipment defect recognition results.

[0128] This embodiment also provides an electronic device suitable for intelligent identification of power equipment defects, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the intelligent identification method for power equipment defects as proposed in the above embodiment.

[0129] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the intelligent identification method for power equipment defects as proposed in the above embodiments.

[0130] The storage medium proposed in this embodiment belongs to the same inventive concept as the intelligent identification method for power equipment defects proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0131] Based on the above description of the implementation methods, those skilled in the art will clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0132] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An intelligent identification method for defects of power equipment, characterized in that, The method comprises the following steps: Collecting an original image of the power equipment, screening the original image of the power equipment to obtain a channel image; Performing enhancement processing on the channel image to calculate a gray value difference and obtain a gray image; Converting the gray image into a frequency spectrum image by using fast Fourier transform, constructing a Gaussian filter function, performing convolution operation on the Gaussian filter function and the frequency spectrum image to attenuate high-frequency components and retain low-frequency components, and performing inverse fast Fourier transform on the convolution-processed frequency spectrum image to convert the frequency spectrum image back to a spatial domain image, thereby obtaining a spatial domain image with smoothed processing and highlighted defect features. Performing adaptive threshold segmentation on the spatial domain image to divide the image into a defect region and a non-defect region, thereby obtaining a segmented image, which is processed by morphological processing to obtain a power equipment defect recognition result.

2. The power equipment defect intelligent identification method of claim 1, wherein, The method for obtaining the gray image comprises the following steps: Performing mean filtering on the channel image, calculating a difference between the gray value after mean filtering and the original gray value of the channel image to obtain a gray value difference, and fusing the gray value difference with an influence factor and the original gray value to calculate and generate a gray image. The method for obtaining the spatial domain image comprises the following steps:

3. The power equipment defect intelligent identification method of claim 2, wherein, Separating low-frequency defect information and high-frequency texture noise in the gray image by using fast Fourier transform, converting the spatial domain into a frequency spectrum image; Constructing a Gaussian filter function, performing convolution operation on the Gaussian filter function and the frequency spectrum image to attenuate high-frequency components and retain low-frequency components; Performing inverse fast Fourier transform on the convolution-processed frequency spectrum image to convert the frequency spectrum image back to a spatial domain image, thereby obtaining a spatial domain image with smoothed processing and highlighted defect features. The method for performing adaptive threshold segmentation on the spatial domain image to divide the image into a defect region and a non-defect region, thereby obtaining a segmented image, comprises the following steps:

4. The power equipment defect intelligent identification method of claim 3, wherein, Extracting the gray value of a pixel in a target region in the spatial domain image, and statistically grouping the pixels based on the gray value; Calculating and selecting a threshold value that makes the inter-class variance of different groups of gray values reach a maximum value as an optimal segmentation threshold value, and dividing the spatial domain image into a defect region and a non-defect region according to the optimal segmentation threshold value to obtain a segmented image. The method for statistically grouping the pixels comprises the following steps:

5. The power equipment defect intelligent identification method of claim 4, wherein, Counting the frequency of occurrence of each gray value and the proportion of each gray value in the target region; Dividing the gray values into two groups by setting a candidate threshold value, calculating the proportion of the pixels in the target region, the sum of the gray values and the average gray value of the two groups of pixels, and obtaining the overall average gray value of the spatial domain image. The method for screening the original image of the power equipment to obtain a channel image comprises the following steps:

6. The power equipment defect intelligent identification method of claim 1, wherein, Performing preprocessing on the collected original image of the power equipment, removing noise in the image by using median filtering, and correcting the geometric distortion and color deviation of the original image of the power equipment; Decomposing the preprocessed original image of the power equipment into R, G and B color channels, comparing and analyzing the definition and prominence of the defect region in each channel, and screening an R channel image as a channel image. The calculation formula for fusing the gray value difference with the influence factor and the original gray value to calculate and generate a gray image is as follows: res=[f(x,y)-g(x,y)]*Factor+f(x,y) 7. The power equipment defect intelligent identification method of claim 2, wherein ​ ​ Wherein, f(x, y) represents the original gray value of each point on the image, G represents the local area covered by the filter kernel, M and N represent the size of the filter kernel respectively, g(x, y) is the new gray value calculated by the mean filtering operation, Factor represents the influence factor, which can adjust the filtering strength, and res represents the gray image obtained after a series of processing.

8. An intelligent power equipment defect identification system, applying the method of any one of claims 1-7, characterized in that, It comprises: An image acquisition module is configured to acquire an original image of the power equipment. An image processing module is configured to filter and acquire a channel image from the original image of the power equipment. After enhancement processing of the channel image, a gray difference value is calculated to obtain a gray image; the gray image is converted into a frequency spectrum image by using fast Fourier transform; a Gaussian filter function is constructed to perform convolution and inverse transformation on the frequency spectrum image to obtain a spatial domain image. An image recognition module is configured to perform adaptive threshold segmentation on the spatial domain image to divide the image into a defect region and a non-defect region, and obtain a segmented image; the segmented image is processed by morphological processing to obtain a power equipment defect recognition result.

9. An electronic device, comprising: It comprises: A memory and a processor; the memory is configured to store computer executable instructions; the processor is configured to execute the computer executable instructions; when the computer executable instructions are executed by the processor, the steps of the power equipment defect intelligent identification method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that, The computer executable instructions stored therein, when executed by the processor, implement the steps of the power equipment defect intelligent identification method according to any one of claims 1 to 7.

Images