High-voltage insulator damaged image denoising method, device, equipment and medium
Through U-Net neural network and smoothing processing technology, the problem of image blurring of high-voltage insulator damage is solved, and the image denoising is effectively improved, which is suitable for analysis and support for complex insulator damage.
Patent Information
- Application Number
- CN202411625576.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-14
- Publication Date
- 2025-02-14
AI Technical Summary
The image blurring caused by damage during operation of high-voltage insulators is difficult to effectively denoise in the prior art, which affects the safety and stability of the power system.
U-Net neural network is used for denoising, and deep images are extracted through multi-layer convolution operations, combined with smoothing processing and pooling layer framework, and adaptively adjust the number of iterations until the model convergence or maximum iterations are reached.
It effectively improves the quality of high-voltage insulator damage images, can adapt to complex insulator damage situations, and provides strong image analysis support.
Smart Images
Figure CN119444612A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing technology, and in particular to a method, device, equipment and medium for denoising damaged images of high-voltage insulators. Background Art
[0002] High-voltage insulators are inevitably damaged during operation, which can lead to power system failures and seriously affect the safety and stability of the power system. In order to detect and repair the damage of high-voltage insulators in a timely manner, researchers usually use image processing technology to analyze and identify the images of high-voltage insulators. However, due to external factors such as noise and interference, many images are blurred. Therefore, it is necessary to improve the existing technology.
[0003] The above information is presented as background information only to assist with understanding the present disclosure and no determination or admission is made as to whether any of the above may be used as prior art with respect to the present disclosure. Summary of the invention
[0004] The present invention provides a method, device, equipment and medium for denoising a damaged image of a high-voltage insulator, so as to solve the problems existing in the prior art.
[0005] To achieve the above object, the present invention provides the following technical solutions:
[0006] A high-voltage insulator damage image denoising method, comprising:
[0007] The damaged image of the high-voltage insulator is input into the U-Net neural network for denoising to obtain a preliminary image;
[0008] Performing smoothing on the preliminary image to obtain a smoothed image;
[0009] Constructing a U-Net neural network model with a pooling layer framework, and using the U-Net neural network model for training the smoothed image;
[0010] Calculate the loss value of the U-Net neural network based on the loss function, compare the loss value with a preset standard value, and continue if the standard is met; otherwise, return to the smoothing processing step;
[0011] The number of iterations is adaptively adjusted according to the change of the loss function, and the denoised image is output when the termination condition is met; the termination condition includes reaching the maximum number of iterations or the model convergence.
[0012] Optionally, the step of inputting the high-voltage insulator damage image into a U-Net neural network for denoising includes:
[0013] The deep features of the image are extracted through the multi-layer convolution operation of the convolution layer, the input and output values are determined to construct the excitation function, the ReLU function is used to solve the overfitting problem, and the regional maximum value is extracted and the average value is calculated through the pooling layer to form a feature operation process.
[0014] Optionally, the smoothing of the preliminary image includes:
[0015] Calculate the mean and variance of image data using the following formula:
[0016]
[0017] Among them, p m is the mean of the data; f m 2 represents the variance of the data; x i is the i-th data; m is the number of data.
[0018] Optionally, the smoothing of the preliminary image further includes:
[0019] The data state is changed by displacement transformation, and the calculation formula is:
[0020]
[0021] Among them, y i Indicates the image data of the damaged insulator after displacement change; x i represents the insulator damage image data before the change; m represents the parameters learned by the network; k p represents the transformation reconstruction coefficient; B n Represents the normalization factor, which is used to control and adjust the output of the image data after the displacement change.
[0022] Optionally, the constructing a U-Net neural network model with a pooling layer framework, and using the U-Net neural network model for training the smoothed image, includes:
[0023] Calculate the loss value of the neural network model, the calculation formula is:
[0024]
[0025] Among them, L k Represents the loss function value; y abc represents the noise of the original insulator damage image; y abcp represents the noise of the processed insulator damage image; w i and h i Represents the pixel parameters of the insulator damage image.
[0026] Optionally, the basis for judging whether the model has converged is:
[0027] During the continuous iteration process, the loss function value remains unchanged or changes very little for a preset number of times.
[0028] Optionally, the U-Net neural network can capture both local details and global context information when extracting image features;
[0029] The denoising autoencoder realizes image smoothing processing by introducing noise into input data and reconstructing low-noise data.
[0030] The present invention also provides a high-voltage insulator damaged image denoising device, which is used to implement the high-voltage insulator damaged image denoising method as described in any one of the above items, comprising:
[0031] An image input module is used to input the high-voltage insulator damage image into the U-Net neural network for denoising to obtain a preliminary image;
[0032] A smoothing processing module, used for smoothing the preliminary image obtained by the image input module, so as to obtain a smoothed image;
[0033] A model building and training module, used to build a U-Net neural network model with a pooling layer framework, and use the U-Net neural network model to train the smoothed image obtained by the smoothing processing module;
[0034] The loss value calculation and comparison module is applied to calculate the loss value of the U-Net neural network based on the loss function, and compare the loss value with the preset standard value, and decide whether to continue the subsequent operation according to the comparison result. If the standard is met, the next operation is triggered, and if the standard is not met, the operation of returning to the smoothing step is triggered;
[0035] The iteration number adjustment module is used to adaptively adjust the iteration number according to the change of the loss function, and trigger the operation of outputting the denoised image when the termination condition is met; the termination condition includes reaching the maximum number of iterations or model convergence.
[0036] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the high-voltage insulator damage image denoising method as described in any one of the above items is implemented.
[0037] The present invention also provides a storage medium containing computer executable instructions, wherein the computer executable instructions are executed by a computer processor to implement the high-voltage insulator damage image denoising method as described in any one of the above items.
[0038] Compared with the prior art, the present invention has the following beneficial effects:
[0039] The present invention provides a high-voltage insulator damage image denoising method, device, equipment and medium. Multiple iterative training can enable the U-Net neural network to gradually adapt to various complex insulator damage conditions, effectively improve the quality of insulator damage images, and provide strong support for subsequent image analysis.
[0040] The present invention has other features and advantages, which will be apparent from or will be described in detail in the accompanying drawings and the following detailed description incorporated herein, which together serve to explain the specific principles of the invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0042] Figure 1 It is a flow chart of a high-voltage insulator damage image denoising method provided by the present invention;
[0043] Figure 2 It is a structural diagram of a U-Net neural network in a high-voltage insulator damage image denoising method provided by the present invention;
[0044] Figure 3 It is a feature operation flow chart of a network insulator damage image in a high-voltage insulator damage image denoising method provided by the present invention;
[0045] Figure 4 It is a structural diagram of a denoising autoencoder in a high-voltage insulator damage image denoising method provided by the present invention;
[0046] Figure 5 It is a flow chart of an insulator damage image denoising algorithm in a high-voltage insulator damage image denoising method provided by the present invention;
[0047] Figure 6 It is a high-voltage insulator image before denoising in a high-voltage insulator damaged image denoising method provided by the present invention.
[0048] Figure 7 is another high-voltage insulator image before denoising in the high-voltage insulator damage image denoising method provided by the present invention;
[0049] Figure 8It is a high-voltage insulator image after denoising in a high-voltage insulator damage image denoising method provided by the present invention;
[0050] Fig. 9 It is another high-voltage insulator image after denoising in the high-voltage insulator damage image denoising method provided by the present invention;
[0051] Fig.10 This is a structural block diagram of a high-voltage insulator damage image denoising device provided by the present invention.
[0052] Attached figure numbers: 10, image input module; 20, smoothing processing module; 30, model building and training module; 40, loss value calculation and comparison module; 50, iteration number adjustment module. DETAILED DESCRIPTION
[0053] In order to explain in detail the possible application scenarios, technical principles, specific schemes that can be implemented, and the purposes and effects that can be achieved, the following is a detailed description of the specific embodiments listed in conjunction with the accompanying drawings. The embodiments described herein are only used to more clearly illustrate the technical solutions of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.
[0054] Reference to "embodiment" herein means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. The term "embodiment" appearing in various places in the specification does not necessarily refer to the same embodiment, nor does it particularly limit its independence or association with other embodiments. In principle, in the present application, as long as there is no technical contradiction or conflict, the various technical features mentioned in the embodiments can be combined in any way to form a corresponding implementable technical solution.
[0055] Unless otherwise defined, the technical terms used in this document have the same meanings as those generally understood by those skilled in the art to which this application belongs; the use of relevant terms in this document is only for describing specific embodiments and is not intended to limit this application.
[0056] In the description of this application, the term "and / or" is an expression used to describe the logical relationship between objects, indicating that three relationships may exist, for example, A and / or B, which means: A exists, B exists, and A and B exist at the same time. In addition, the character " / " in this article generally indicates that the objects before and after are in an "or" logical relationship.
[0057] In the present application, terms such as “first” and “second” are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship of quantity, priority or sequence between these entities or operations.
[0058] Without further limitations, in this application, the words "include", "comprises", "has" or other similar expressions used in the sentences are intended to cover non-exclusive inclusion. These expressions do not exclude the presence of additional elements in the process, method or product including the elements, so that the process, method or product including a series of elements may include not only those limited elements, but also other elements not explicitly listed, or also include elements inherent to such process, method or product.
[0059] Similar to the understanding in the Examination Guidelines, in this application, expressions such as "greater than", "less than", "exceed" and the like are understood to exclude the number itself; expressions such as "above", "below", "within" and the like are understood to include the number itself. In addition, in the description of the embodiments of this application, "multiple" means more than two (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups", "multiple times", etc., unless otherwise clearly and specifically limited.
[0060] In the description of the embodiments of the present application, space-related expressions used, such as "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "vertical", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicate the orientation or position relationship based on the orientation or position relationship shown in the specific embodiments or drawings, and are only for the convenience of describing the specific embodiments of the present application or facilitating the reader's understanding, and do not indicate or imply that the referred device or component must have a specific position, a specific orientation, or be constructed or operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of the present application.
[0061] Unless otherwise expressly specified or limited, in the description of the embodiments of the present application, the terms such as "install", "connect", "connect", "fix", "set", etc. used should be understood in a broad sense. For example, the "connection" can be a fixed connection, a detachable connection, or an integrated setting; it can be a mechanical connection, an electrical connection, or a communication connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For technicians in the technical field to which the present application belongs, the specific meanings of the above terms in the embodiments of the present application can be understood according to the specific circumstances.
[0062] Please refer to Figure 1 The embodiment of the present invention provides a method for denoising a high-voltage insulator damage image, comprising:
[0063] S1. Input the high-voltage insulator damage image into the U-Net neural network for denoising to obtain a preliminary image.
[0064] Among them, the U-Net neural network can capture local details and global contextual information simultaneously when extracting image features; the denoising autoencoder achieves image smoothing by introducing noise into the input data and reconstructing low-noise data.
[0065] Due to shooting conditions, equipment performance, environmental factors and other reasons, images usually contain large noise. In order to effectively process insulator damage images, it is first necessary to obtain the features of insulator damage images. The U-Net neural network is a deep learning network structure. Its U-shaped structure enables it to capture local details and global contextual information of the image at the same time, realize multi-scale feature extraction, and help to more accurately identify insulator damage. The U-Net neural network structure is relatively simple. It consists of two parts, an encoder and a decoder. It can effectively extract the feature information of the image and perform image reconstruction. At the same time, U-Net also uses jump connection technology, which can simultaneously retain the local and global information of the image, thereby improving the denoising performance. Therefore, after determining the loss of the U-Net neural network, the real insulator damage image can be input into the network feature structure for denoising. The U-Net neural network structure is as follows Figure 2 shown.
[0066] In this step, the high-voltage insulator damage image is input into the U-Net neural network for denoising, including:
[0067] The deep features of the image are extracted through the multi-layer convolution operation of the convolution layer, the input and output values are determined to construct the excitation function, the ReLU function is used to solve the overfitting problem, and the regional maximum value is extracted and the average value is calculated through the pooling layer to form a feature operation process.
[0068] Specifically, in Figure 2 In the example, the U-Net neural network loss is:
[0069] (1)
[0070] In the formula, Dop represents the U-Net neural network loss function; ki represents the convolution layer function; MP24029559 represents the initial value of the input; H(x) represents the real labeled insulator damage image; Nm represents the number of insulator damage images. By comparing the rich texture features and information, the pixels at the same position can be located and the number of paths to the center pixel can be calculated. By comparing the perceptual features of one actual scene with the features generated by the other algorithm, the definitions of the two can be obtained:
[0071]
[0072] Among them, d(t) is the one-dimensional signal of the pixel; p(t) is the frequency domain after conversion; Lf is the convolution path length; m(t) is the receptive field strength; t is the convolution kernel radius. Combined with the frequency domain, the standard network structure can be obtained through the inverse transformation and expansion of the Fourier transform. In the convolution layer of U-Net, the deep features in the image can be extracted layer by layer through multi-layer convolution operations. The input value and output value are:
[0073]
[0074] In the formula, Kn(p) and Kn(f) represent the input value and output value of the convolution layer respectively; Kn-1 (fw) is the specific parameter of the weight in the n-1th layer; fw is the data depth of the convolution layer; hp is the height of the convolution kernel; wd is the width of the convolution kernel; fs is the bias parameter in the convolution kernel. This function can be used to construct the excitation function of the convolution layer. In the excitation layer, the ReLU function is generally used to solve the overfitting problem of the network. The input value of this function can be expressed as:
[0075]
[0076] Where f(x) represents the input value of the function in the activation layer; xp represents the corrected input value. Without affecting the receptive field, the sample features in the insulator damage image are concentrated to obtain the pooling layer, and the maximum value of the selected area is extracted, such as:
[0077]
[0078] Among them, ft represents the maximum feature sample in the 2×2 filter; the values in (a, b, c) are the coefficients in the filter. The average value of the selected area is further calculated as follows:
[0079]
[0080] Where yt is the average pooling value. Combining the above pooling characteristics, the feature operation flow of the network insulator damage image can be obtained, as follows: Figure 3 shown.
[0081] S2, smoothing the preliminary image to obtain a smoothed image;
[0082] Although the U-Net neural network can extract rich feature information, it may contain redundant or irrelevant features, which may interfere with the subsequent damage detection task. Therefore, combined with the above mapping features, the network is first smoothed to improve the visual effect of the image, making the insulator damage area more prominent and easy to identify. The insulator damage image data is trained, and the preprocessed data is input to calculate its mean and variance:
[0083]
[0084] Among them, pm is the mean of the data; fm 2 represents the variance of the data; xi is the ith data; and m is the number of data. After normalization, it is necessary to further force the data to be passed to the next layer of the network to change its displacement.
[0085]
[0086] In the formula, yi represents the damaged insulator image data after displacement change; xi represents the damaged insulator image data before change; λm represents the parameter learned by the network; kp represents the transformation reconstruction coefficient. Bn represents the normalization factor, which is used to control and adjust the output of the image data after displacement change. The displacement transformation can diversify the high-voltage insulator damaged image data, thereby improving the performance of the denoising network. That is, when the high-voltage insulator damaged image is affected by noise, distortion, etc., the position of the pixel may change slightly. By introducing displacement changes in the training process, the diversity of the training set can be increased, and at the same time, the denoising autoencoder (Denoising Autoencoder, DAE) is introduced, such as Figure 4 As shown. It reduces noise while making the image edges smooth and sharp. Denoising Autoencoder is a neural network model for unsupervised learning. Unlike ordinary autoencoders, the goal of denoising autoencoders is to achieve a smoothing effect by introducing noise into the input data and then trying to reconstruct low-noise data from the noisy input.
[0087] After the above transformation, the data is distributed in a relatively stable range, which means that the input value and output value can remain unchanged, solving the problem of low generalization ability and slow training speed of the U-net network, and obtaining a smooth insulator damage image processing result.
[0088] S3, constructing a U-Net neural network model with a pooling layer framework, and using the U-Net neural network model for training smooth images;
[0089] S4, calculating the loss value of the U-Net neural network based on the loss function, comparing the loss value with the preset standard value, and continuing if it meets the standard, otherwise returning to the smoothing processing step;
[0090] S5. Adaptively adjust the number of iterations according to the change of the loss function, and output the denoised image when the termination condition is met; the termination condition includes reaching the maximum number of iterations or the model convergence.
[0091] In this embodiment, the basis for judging whether the model has converged is:
[0092] During the continuous iteration process, the loss function value remains unchanged or changes very little for a preset number of times.
[0093] It is understandable that in the process of smoothing, in order to remove noise, some image edge and detail information will inevitably be lost, and this information is crucial for the accurate identification and positioning of insulator damage. Therefore, based on the smoothing of high-voltage insulator damage images, the U-Net neural network is trained. Multiple iterations of training can make the U-Net neural network gradually adapt to various complex insulator damage situations and continuously improve its denoising performance. As the number of iterations increases, the network will gradually converge to the best denoising effect, thereby obtaining a more accurate insulator damage image denoising result. The overall process of image denoising is as follows: Figure 5 shown.
[0094] like Figure 5 As shown in the figure, the process of insulator damage image denoising algorithm is summarized as follows:
[0095] Step 1: Input the original high-voltage insulator damage image;
[0096] Step 2: normalize and grayscale the high-voltage insulator damage image, and fill the pixels of the insulator damage image;
[0097] Step 3: Smoothing the processed image;
[0098] Step 4: Build a U-Net neural network model with a pooling layer framework for model training;
[0099] Step 5: Calculate the loss value of the U-Net neural network model and determine whether the calculated loss function value is lower than the standard value (evaluate the performance of the model on the data set through cross-validation, and set the standard value according to the validation result. In this invention, the standard value obtained by validation = 0.05 is used). If it is not lower than the standard value, return to step 3; if it is lower, execute step 6. The calculation formula of the loss function is:
[0100]
[0101] Among them, L k Represents the loss function value; y abc represents the noise of the original insulator damage image; y abcp represents the noise of the processed insulator damage image; w i and h i Represents the pixel parameters of the insulator damage image.
[0102] Step 6: Monitor the change of the loss function value, adaptively adjust the number of iterations, and determine whether the maximum number of iterations of the algorithm has been reached. If the loss function value has hardly decreased or decreased very little during multiple iterations, it means that the model has converged and the iteration can be stopped to avoid overfitting. If the maximum number of iterations has not been reached, return to step 3; if the maximum number of iterations has been reached, terminate the iteration, output the denoised insulator damage image, and obtain the denoising result under this method, completing the denoising of the insulator damage image.
[0103] The experimental results of the insulator damage image denoising method provided in this embodiment are as follows:
[0104] 600 images were randomly selected from the high-voltage insulator image dataset. Training and testing were performed in a ratio of 4:2. The initial parameters of the algorithm were set as follows: learning rate was 0.05, decay coefficient was 0.1, and weight decay was 0.0001. Training strategy: The initial weights of the neural network converged. After pre-training, the encoding structure was improved, all network weights were initialized, and the size and number of convolution kernels were obtained.
[0105] The comparison results before and after denoising are shown in the figure. Figure 6 , Figure 7 This is the high voltage insulator image before denoising; Figure 8 , Fig. 9 This is the denoised high-voltage insulator image.
[0106] Please refer to Fig.10 The present invention also provides a high-voltage insulator damage image denoising device, which is used to implement any of the above high-voltage insulator damage image denoising methods, comprising:
[0107] An image input module 10 is used to input the high-voltage insulator damage image into a U-Net neural network for denoising to obtain a preliminary image;
[0108] A smoothing processing module 20, used for smoothing the preliminary image obtained by the image input module, so as to obtain a smoothed image;
[0109] A model building and training module 30 is used to build a U-Net neural network model with a pooling layer framework, and use the U-Net neural network model to train the smoothed image obtained by the smoothing processing module;
[0110] The loss value calculation and comparison module 40 is applied to calculate the loss value of the U-Net neural network based on the loss function, and compare the loss value with the preset standard value, and decide whether to continue the subsequent operation according to the comparison result. If the standard is met, the next operation is triggered, and if the standard is not met, the operation of returning to the smoothing step is triggered;
[0111] The iteration number adjustment module 50 is used to adaptively adjust the iteration number according to the change of the loss function, and trigger the operation of outputting the denoised image when the termination condition is met; the termination condition includes reaching the maximum iteration number or model convergence.
[0112] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the high-voltage insulator damage image denoising method as described above is implemented.
[0113] The present invention also provides a storage medium containing computer executable instructions, and the computer executable instructions are executed by a computer processor to implement any of the above methods for denoising a damaged image of a high-voltage insulator.
[0114] Finally, it should be noted that although the above embodiments have been described in the specification and drawings of this application, this does not limit the scope of patent protection of this application. All technical solutions generated by replacing or modifying equivalent structures or equivalent processes based on the essential concept of this application using the contents recorded in the specification and drawings of this application, as well as directly or indirectly implementing the technical solutions of the above embodiments in other related technical fields, are included in the scope of patent protection of this application.
Claims
1. A method for denoising a high-voltage insulator damage image, characterized in that: include: The damaged image of the high-voltage insulator is input into the U-Net neural network for denoising to obtain a preliminary image; Performing smoothing on the preliminary image to obtain a smoothed image; Constructing a U-Net neural network model with a pooling layer framework, and using the U-Net neural network model for training the smoothed image; Calculate the loss value of the U-Net neural network based on the loss function, compare the loss value with a preset standard value, and continue if the standard is met; otherwise, return to the smoothing processing step; The number of iterations is adaptively adjusted according to the change of the loss function, and the denoised image is output when the termination condition is met; the termination condition includes reaching the maximum number of iterations or the model convergence.
2. The high-voltage insulator damage image denoising method according to claim 1, characterized in that: The step of inputting the high-voltage insulator damage image into a U-Net neural network for denoising includes: The deep features of the image are extracted through the multi-layer convolution operation of the convolution layer, the input and output values are determined to construct the excitation function, the ReLU function is used to solve the overfitting problem, and the regional maximum value is extracted and the average value is calculated through the pooling layer to form a feature operation process.
3. The high-voltage insulator damage image denoising method according to claim 1, characterized in that: The smoothing process of the preliminary image comprises: Calculate the mean and variance of image data using the following formula: ； Among them, p m is the mean of the data; f m 2 represents the variance of the data; x i is the i-th data; m is the number of data.
4. The high-voltage insulator damage image denoising method based on the basic U-Net neural network according to claim 1 is characterized in that: The smoothing process of the preliminary image further includes: The data state is changed by displacement transformation, and the calculation formula is: ； Among them, y i Insulator damage image data after displacement change; x i represents the insulator damage image data before the change; m represents the parameters learned by the network; k p represents the transformation reconstruction coefficient; B n Represents the normalization factor, which is used to control and adjust the output of the image data after the displacement change.
5. The high-voltage insulator damage image denoising method according to claim 1, characterized in that: The step of constructing a U-Net neural network model with a pooling layer framework and using the U-Net neural network model for training the smoothed image includes: Calculate the loss value of the neural network model, the calculation formula is: ； Among them, L k Represents the loss function value; y abc represents the noise of the original insulator damage image; y abcp represents the noise of the processed insulator damage image; w i and h i Represents the pixel parameters of the insulator damage image.
6. The high-voltage insulator damage image denoising method according to claim 1, characterized in that: The basis for judging whether the model converges is: During the continuous iteration process, the loss function value remains unchanged or changes very little for a preset number of times.
7. The high voltage insulator damage image denoising method according to claim 1, characterized in that: The U-Net neural network can capture both local details and global context information when extracting image features; The denoising autoencoder realizes image smoothing processing by introducing noise into input data and reconstructing low-noise data.
8. A high-voltage insulator damage image denoising device, characterized in that: The method for denoising a damaged image of a high-voltage insulator according to any one of claims 1 to 7 comprises: An image input module is used to input the high-voltage insulator damage image into the U-Net neural network for denoising to obtain a preliminary image; A smoothing processing module, used for smoothing the preliminary image obtained by the image input module, so as to obtain a smoothed image; A model building and training module, used to build a U-Net neural network model with a pooling layer framework, and use the U-Net neural network model to train the smoothed image obtained by the smoothing processing module; The loss value calculation and comparison module is applied to calculate the loss value of the U-Net neural network based on the loss function, and compare the loss value with the preset standard value, and decide whether to continue the subsequent operation according to the comparison result. If the standard is met, the next operation is triggered, and if the standard is not met, the operation of returning to the smoothing step is triggered; The iteration number adjustment module is used to adaptively adjust the iteration number according to the change of the loss function, and trigger the operation of outputting the denoised image when the termination condition is met; the termination condition includes reaching the maximum number of iterations or model convergence.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the high-voltage insulator damage image denoising method according to any one of claims 1 to 7 is implemented.
10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions are executed by a computer processor to implement the high-voltage insulator damage image denoising method according to any one of claims 1 to 7.