Power transmission line hardware corrosion state evaluation method and device, terminal equipment and computer readable storage medium

Abstract

The invention discloses a transmission line fitting corrosion state assessment method and device, terminal equipment and a computer readable storage medium, and the method comprises the steps: obtaining an image of a to-be-assessed fitting, carrying out the feature extraction of the image of the to-be-assessed fitting, and obtaining the corrosion feature quantity of the to-be-assessed fitting; the corrosion characteristic quantity comprises a colorimetric characteristic quantity, a texture characteristic quantity, a wavelet characteristic quantity and a fractal characteristic quantity; inputting the corrosion characteristic quantity into a preset corrosion assessment model, so that the model determines the volume corrosion rate and the maximum corrosion pit depth of the to-be-assessed hardware fitting according to the corrosion characteristic quantity of the to-be-assessed hardware fitting; and finally, according to the predicted volume corrosion rate and the maximum corrosion pit depth, determining the corrosion state of the to-be-evaluated hardware fitting. By implementing the method, the corrosion state of the hardware fitting can be evaluated more comprehensively according to the volume corrosion rate and the maximum corrosion pit depth.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method, a device, a terminal device and a computer-readable storage medium for evaluating the corrosion state of hardware of a power transmission line. Background Art

[0002] In recent years, the scale of power transmission lines has grown rapidly. Hardware plays a major role as a structural component in power transmission lines, and its impact on the safe operation of the power grid is critical. In some areas of southern China, due to the humid climate, transmission line hardware is more susceptible to corrosion than in other areas. Severely corroded hardware may cause power grid operation failures. Therefore, inspection personnel are required to promptly detect corroded hardware and determine the degree of corrosion based on some corrosion status indicators, so as to replace severely corroded hardware. With the development of smart grids, it has become common to use drones for intelligent inspections. Therefore, it is very meaningful to identify target hardware from hardware images taken by drones, extract regional features of corroded hardware, and finally quantitatively predict the corrosion status indicators of hardware and give its corrosion status.

[0003] At present, the corrosion evaluation index for corroded hardware in power inspection images is often based on the area corrosion rate obtained by the ratio of the number of pixels in the hardware corrosion area to the number of pixels in the hardware body in the image to judge the degree of corrosion of the hardware. Although the current drone inspection is fixed and shoots the transmission line from multiple angles, it is difficult to judge the corrosion area of ​​the entire hardware through computer vision technology, so it is not comprehensive enough to judge the degree of corrosion of the hardware only by the area corrosion rate. Summary of the invention

[0004] The embodiments of the present invention provide a method, an apparatus, a terminal device and a computer-readable storage medium for evaluating the corrosion state of transmission line hardware, which can more comprehensively evaluate the corrosion state of hardware.

[0005] An embodiment of the present invention provides a method for evaluating the corrosion status of transmission line hardware, comprising:

[0006] Obtain an image of the hardware to be evaluated;

[0007] Extract features from the image of the hardware to be evaluated to obtain the corrosion feature quantity of the hardware to be evaluated; the corrosion feature quantity includes: colorimetric feature quantity, texture feature quantity, wavelet feature quantity and fractal feature quantity;

[0008] The corrosion characteristic quantity is input into a preset corrosion assessment model, so that the corrosion assessment model determines the volume corrosion rate and the maximum corrosion pit depth of the hardware to be assessed according to the corrosion characteristic quantity of the hardware to be assessed;

[0009] Determine the corrosion status of the hardware to be evaluated based on the volume corrosion rate and the maximum corrosion pit depth;

[0010] Among them, the corrosion assessment model takes the corrosion characteristics of the training hardware as input, and the predicted volume corrosion rate and predicted maximum pit depth of the training hardware as output for training. During the training process, the network parameters of the corrosion assessment model are adjusted according to the predicted volume corrosion rate and the actual volume corrosion rate of the training hardware, as well as the predicted maximum pit depth and the actual maximum pit depth of the training hardware.

[0011] Furthermore, the corrosion status of the hardware to be evaluated is determined based on the volume corrosion rate and the maximum corrosion pit depth, including:

[0012] According to the preset corrosion status evaluation grade classification table, the predicted volume corrosion rate and maximum pit depth are matched with the corresponding grades to obtain the corrosion status of the hardware to be evaluated.

[0013] Furthermore, the preset corrosion assessment model is trained in the following way:

[0014] Obtain the volume corrosion rate and maximum pit depth of several training hardware;

[0015] Acquire several training hardware images, and perform grayscale conversion, Gaussian filtering, and piecewise linear enhancement on each training hardware image to obtain preprocessed images of each training hardware;

[0016] Perform feature extraction on the preprocessed images of each training hardware to obtain the corrosion feature value of each training hardware;

[0017] The corrosion feature values ​​of each training hardware are input into the corrosion assessment model to be trained for iterative training until the sum of squares of the model error is less than the prediction threshold, thereby obtaining a preset corrosion assessment model.

[0018] Furthermore, the volume corrosion rate and maximum pit depth of several training hardware are obtained, including:

[0019] For each training hardware, obtain the initial mass of the uncorroded training hardware;

[0020] Derusting the rusted training hardware to obtain the quality of the derusted training hardware;

[0021] According to the initial mass and the mass of the training hardware after rust removal, the volume corrosion rate of each training hardware is calculated;

[0022] Perform three-dimensional scanning on the corroded training hardware to obtain scanning data;

[0023] The scanned data is binarized, and the area below the binarization threshold is regarded as the etch pit area;

[0024] The maximum depth of all pit areas in the training hardware is calculated by flood filling algorithm;

[0025] The maximum depths of all pit areas were compared to obtain the maximum pit depth of the training hardware.

[0026] Furthermore, the colorimetric feature includes: the mean of the R component in the RGB color model, the standard deviation of the R component in the RGB color model, the mean of the H component in the HSI color model, the standard deviation of the H component in the HSI color model, the mean of the S component in the HSI color model, and the standard deviation of the S component in the HSI color model;

[0027] The colorimetric features are extracted in the following way:

[0028] Convert the image into RGB color model and calculate the mean and standard deviation of the R component in the RGB color model;

[0029] The RGB color model is converted into the HSI color model, and the mean of the H component in the HSI color model, the standard deviation of the H component in the HSI color model, the mean of the S component in the HSI color model, and the standard deviation of the S component in the HSI color model are calculated.

[0030] Furthermore, the texture feature quantity includes: energy, a measurement index reflecting the coarseness of the image texture, and randomness, a measurement index reflecting the amount of information contained in the image;

[0031] The texture features are extracted in the following way:

[0032] Convert the image into a grayscale image, and calculate the grayscale co-occurrence matrix based on the grayscale image;

[0033] According to the probability of each gray value in the gray image appearing in the gray level co-occurrence matrix, the energy, a metric reflecting the coarseness of the image texture, and the randomness, a metric reflecting the amount of information contained in the image, are calculated.

[0034] Furthermore, the wavelet feature includes: anisotropic energy value, and energy sum of high frequency components of secondary wavelet decomposition sub-images;

[0035] The wavelet features are extracted in the following way:

[0036] Performing a secondary wavelet transform on the image to obtain an approximate detail coefficient, a horizontal detail coefficient, a vertical detail coefficient, and a diagonal detail coefficient of the secondary wavelet transform;

[0037] According to the approximate detail coefficient, horizontal detail coefficient, vertical detail coefficient, and diagonal detail coefficient of the secondary wavelet transform, the anisotropic energy value and the energy sum of the high-frequency components of the secondary wavelet decomposition sub-image are calculated by the following calculation formula:

[0038] E t =E H2 +E V2 +E G2 ;

[0039]

[0040] Among them, E t represents the energy sum of high frequency components of the sub-image decomposed by secondary wavelet, E H2 Represents the energy percentage of the secondary wavelet transform approximate detail coefficient, E V2 Represents the energy percentage of the diagonal detail coefficients of the secondary wavelet transform, E G2 Represents the energy percentage of the secondary wavelet transform vertical detail coefficient, E O Anisotropic energy value.

[0041] Based on the above method embodiment, the present invention provides a corresponding device embodiment, including: a hardware image acquisition module to be evaluated, a corrosion feature extraction module, an index prediction module, and a state matching module;

[0042] A hardware image acquisition module to be evaluated, used to acquire images of the hardware to be evaluated;

[0043] The rust feature extraction module is used to extract features from the image of the hardware to be evaluated, and obtain the rust feature quantity of the hardware to be evaluated; the rust feature quantity includes: colorimetric feature quantity, texture feature quantity, wavelet feature quantity and fractal feature quantity;

[0044] An indicator prediction module is used to input the corrosion feature quantity into a preset corrosion assessment model, so that the corrosion assessment model determines the volume corrosion rate and maximum pit depth of the hardware to be assessed according to the corrosion feature quantity of the hardware to be assessed; wherein the corrosion assessment model is trained with the corrosion feature quantity of the training hardware as input and the predicted volume corrosion rate and predicted maximum pit depth of the training hardware as output, and during the training process, the network parameters of the corrosion assessment model are adjusted according to the predicted volume corrosion rate and actual volume corrosion rate of the training hardware, and the predicted maximum pit depth and actual maximum pit depth of the training hardware;

[0045] The state matching module is used to determine the corrosion state of the hardware to be evaluated based on the volume corrosion rate and the maximum corrosion pit depth.

[0046] Based on the above-mentioned method embodiment, the present invention provides a corresponding terminal device embodiment, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the steps of the method for evaluating the corrosion status of transmission line fittings as described in the present invention are implemented.

[0047] Based on the above method embodiment, the present invention provides a corresponding computer-readable storage medium embodiment, including: a stored computer program, which controls the device where the computer-readable storage medium is located to execute the steps of the method for evaluating the corrosion status of transmission line hardware as described in the present invention when the computer program is running.

[0048] Compared with the prior art, the beneficial effects of the embodiment of this solution are:

[0049] The present invention obtains a large number of corrosion features of the corroded hardware image by extracting features of the hardware image to be evaluated. According to a preset corrosion evaluation model, the volume corrosion rate and the maximum corrosion pit depth of the hardware can be predicted through the corrosion features of the corroded hardware image. Compared with judging the corrosion degree of the hardware by the area corrosion rate, the area corrosion rate only reflects the damaged part of the surface, while the volume corrosion rate takes into account the depth and area of ​​the corrosion, and can more comprehensively reflect the actual damage of the hardware. In addition, since local deep corrosion may have a significant impact on the bearing capacity and safety of the hardware, the local weaknesses and potential failure points of the hardware can be identified through the maximum corrosion pit depth. Finally, combined with the preset corrosion status evaluation grade classification table, the predicted volume corrosion rate and the corrosion grade corresponding to the maximum corrosion pit depth are matched, so as to more comprehensively evaluate the corrosion status of the hardware. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a flow chart of a method for evaluating the corrosion status of transmission line hardware provided by an embodiment of the present invention;

[0051] Figure 2 It is a flowchart of a corrosion assessment model training process provided by an embodiment of the present invention;

[0052] Figure 3 It is a structural schematic diagram of a device for evaluating the corrosion status of transmission line hardware provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] like Figure 1 As shown, an embodiment of the present invention provides a method for evaluating the corrosion status of transmission line hardware, the method comprising at least the following steps:

[0055] Step S1: Acquire an image of the hardware to be evaluated;

[0056] For step S1, a high-resolution camera is used to obtain an image of the hardware to be evaluated. The image needs to meet the following requirements: the captured image should contain only one hardware, and the hardware should occupy the main part. To ensure the accuracy of the evaluation, the image should be as close to the hardware as possible, and the background space should be minimized to help highlight the characteristics of the hardware and make the evaluation process more effective.

[0057] Step S2: extracting features from the image of the hardware to be evaluated to obtain the corrosion feature quantity of the hardware to be evaluated; the corrosion feature quantity includes: colorimetric feature quantity, texture feature quantity, wavelet feature quantity and fractal feature quantity;

[0058] In a preferred embodiment, the colorimetric feature includes: the mean of the R component in the RGB color model, the standard deviation of the R component in the RGB color model, the mean of the H component in the HSI color model, the standard deviation of the H component in the HSI color model, the mean of the S component in the HSI color model, and the standard deviation of the S component in the HSI color model;

[0059] The colorimetric features are extracted in the following way:

[0060] Convert the image into RGB color model and calculate the mean and standard deviation of the R component in the RGB color model;

[0061] The RGB color model is converted into the HSI color model, and the mean of the H component in the HSI color model, the standard deviation of the H component in the HSI color model, the mean of the S component in the HSI color model, and the standard deviation of the S component in the HSI color model are calculated.

[0062] In a preferred embodiment, the texture feature includes: energy, a measurement index reflecting the coarseness of the image texture, and randomness, a measurement index reflecting the amount of information contained in the image;

[0063] The texture features are extracted in the following way:

[0064] Convert the image into a grayscale image, and calculate the grayscale co-occurrence matrix based on the grayscale image;

[0065] According to the probability of each gray value in the gray image appearing in the gray level co-occurrence matrix, the energy, a metric reflecting the coarseness of the image texture, and the randomness, a metric reflecting the amount of information contained in the image, are calculated.

[0066] In a preferred embodiment, the wavelet feature includes: anisotropic energy value, and energy sum of high frequency components of secondary wavelet decomposition sub-images;

[0067] The wavelet features are extracted in the following way:

[0068] Performing a secondary wavelet transform on the image to obtain an approximate detail coefficient, a horizontal detail coefficient, a vertical detail coefficient, and a diagonal detail coefficient of the secondary wavelet transform;

[0069] According to the approximate detail coefficient, horizontal detail coefficient, vertical detail coefficient, and diagonal detail coefficient of the secondary wavelet transform, the anisotropic energy value and the energy sum of the high-frequency components of the secondary wavelet decomposition sub-image are calculated by the following calculation formula:

[0070] E t =E H2 +E V2 +E G2 ;

[0071]

[0072] Among them, E t represents the energy sum of high frequency components of the sub-image decomposed by secondary wavelet, E H2 Represents the energy percentage of the secondary wavelet transform approximate detail coefficient, E V2 Represents the energy percentage of the diagonal detail coefficients of the secondary wavelet transform, E G2 Represents the energy percentage of the secondary wavelet transform vertical detail coefficient, E O Anisotropic energy value.

[0073] For step S2, first, the hardware image obtained in step S1 is preprocessed. The main purpose of the preprocessing step is to improve the effect of subsequent feature extraction and analysis and reduce calculation complexity.

[0074] In the first step of preprocessing, the classic weighted average algorithm is used for grayscale conversion. Specifically, the red, green and blue color components in the image are weighted calculated using weights 0.299, 0.587 and 0.114 respectively. In this way, the final grayscale image can be obtained. The main function of this conversion is to reduce the dimension of the data, thereby reducing the amount of calculation in the feature extraction link, making subsequent processing more efficient. In the second step of preprocessing, the image is Gaussian filtered, using a Gaussian function as the convolution kernel. Gaussian filtering can effectively remove Gaussian noise in the image, smooth the image, and retain important edge information. In the third step of preprocessing, piecewise linear transformation is used for image enhancement, which is a commonly used technique for improving the contrast of the image. Since the grayscale range of the rust image is usually more concentrated, in order to better extract the red-like features of the rust, the present invention chooses to stretch the pixel range of the grayscale image. The specific stretching formula is as follows:

[0075]

[0076] In the formula, g(x,y) represents the pixel value after linear transformation, f(x,y) represents the pixel value before linear transformation, and r min Represents the minimum pixel value of the grayscale image, r max Represents the maximum pixel value of a grayscale image.

[0077] After a series of preprocessing of the hardware image, the features of the hardware image are extracted. These features mainly include colorimetric features, texture features, wavelet features and fractal features. First, in terms of colorimetric features, the two-dimensional hardware image is converted into RGB color space so that the color information in the image can be analyzed and the mean value M of the R component in the RGB color model can be calculated. R and standard deviation σ R , these quantities can reflect the main characteristics and distribution of the red component in the image; in addition, the present invention also converts the RGB color model into the HSI color model, where H represents hue, S represents saturation, and I represents brightness. Through this conversion, the color characteristics can be analyzed more intuitively, and the mean value M of the hue characteristics in the HSI color model can be further calculated. H and standard deviation σ H , and the mean M of the saturation feature in the HSI color model S and standard deviation σ S ,These colorimetric features provide important color information basis for subsequent ,image analysis.

[0078] Secondly, in the process of extracting texture features, the gray level co-occurrence matrix method is used to analyze the texture structure of the image. By calculating the gray level co-occurrence matrix of the image, the relative frequency distribution of different gray values ​​can be obtained, and then the energy E and randomness metric S can be calculated using these data. The calculation formula is as follows:

[0079]

[0080] In the formula, S represents randomness, P(i,j) represents the probability of a certain gray value appearing in the grayscale co-occurrence matrix, and E represents energy. Among them, energy reflects the coarseness of the texture in the image, that is, the similarity of the grayscale values ​​of adjacent pixels. The higher the energy, the coarser the texture, and the lower the energy, the finer the texture. The randomness metric represents the amount of information contained in the image, that is, the irregularity between the grayscale values ​​of adjacent pixels. The higher the randomness metric index, the more complex the texture of the image, and the lower the randomness metric index, the simpler the texture of the image. These two features can be used to analyze the details and complexity of the hardware image.

[0081] In terms of wavelet feature extraction, by applying high-pass filters (H) and low-pass filters (L) of different bandwidths to the horizontal and vertical directions of the corroded hardware image, four wavelet coefficient sub-images are obtained: L / L, L / H, H / L, H / H, which correspond to the approximate detail coefficient A1, horizontal detail coefficient H1, vertical detail coefficient G1 and diagonal detail coefficient V1 of the first-level wavelet transform, i.e., first-level decomposition. In order to deeply analyze the details of the image, the first-level decomposition sub-image L / L is further subjected to a second-level wavelet transform, and four sub-images of the next level are obtained, i.e., second-level decomposition. These four sub-images correspond to the second-level wavelet transform approximate detail coefficient A2, horizontal detail coefficient H2, vertical detail coefficient G2 and diagonal detail coefficient V2. The obtained second-level wavelet decomposition sub-image is a coefficient matrix of the corresponding direction. The energy formula can be used to calculate the energy percentage of each sub-image, which represents the approximate detail coefficient E A2 , horizontal detail coefficient E H2 , vertical detail factor E V2 and the diagonal detail factor E G2 These wavelet features can effectively capture the multi-scale information of the image and provide rich details for the structure of the hardware image. At the same time, the anisotropic energy value E is introduced. O and the high frequency component energy of the sub-image decomposed by the second-level wavelet and E t As the wavelet feature of the rust image, the calculation formula is as follows:

[0082]

[0083] In the formula, E t represents the energy sum of high frequency components of the sub-image decomposed by secondary wavelet, E H2 represents the horizontal detail coefficient of the secondary wavelet decomposition, E V2 represents the vertical detail coefficient of the secondary wavelet decomposition, E G2 represents the diagonal detail coefficient of the second-level wavelet decomposition, E O The anisotropic energy value measures the energy distribution of the image in different directions. It is obtained by calculating the energy mean of each sub-image and represents the directional and anisotropic characteristics of the image texture. The energy sum of the high-frequency components of the secondary wavelet decomposition sub-image reflects the detailed information of the image. It is obtained by calculating the energy sum of the high-frequency components in the secondary wavelet decomposition sub-image and is used to represent the richness of the local details in the rust image. Through these wavelet feature quantities, richer image feature information can be obtained.

[0084] Finally, in the extraction of fractal features, the existing DBC (Differential Box-counting) "box dimension method" is used to obtain the fractal feature FD to evaluate the morphological irregularity of the corroded hardware image. Fractal features provide a way to describe the complexity and self-similarity of images, which helps to further analyze the geometric characteristics of hardware images.

[0085] Combining the extraction of the above features, we can obtain a series of quantitative features of corroded hardware images [M R , σ R 、M H 、M S , σ H , σ S , E, S, E t 、E O , FD], in this embodiment, the corrosion feature values ​​obtained are [138.84, 57.12, 0.33, 0.11, 0.26, 0.10, 1.91, 0.22, 3.34, 0.001, 2.21].

[0086] Step S3: inputting the corrosion characteristic quantity into a preset corrosion assessment model, so that the corrosion assessment model determines the volume corrosion rate and the maximum corrosion pit depth of the hardware to be assessed according to the corrosion characteristic quantity of the hardware to be assessed;

[0087] The corrosion assessment model is trained with the corrosion characteristics of the training hardware as input and the predicted volume corrosion rate and predicted maximum pit depth of the training hardware as output. During the training process, the network parameters of the corrosion assessment model are adjusted according to the predicted volume corrosion rate and actual volume corrosion rate of the training hardware, as well as the predicted maximum pit depth and actual maximum pit depth of the training hardware.

[0088] In a preferred embodiment, the preset corrosion assessment model is trained in the following manner:

[0089] Obtain the volume corrosion rate and maximum pit depth of several training hardware;

[0090] Acquire several training hardware images, and perform grayscale conversion, Gaussian filtering, and piecewise linear enhancement on each training hardware image to obtain preprocessed images of each training hardware;

[0091] Perform feature extraction on the preprocessed images of each training hardware to obtain the corrosion feature value of each training hardware;

[0092] The corrosion feature values ​​of each training hardware are input into the corrosion assessment model to be trained for iterative training until the sum of squares of the model error is less than the prediction threshold, thereby obtaining a preset corrosion assessment model.

[0093] In a preferred embodiment, obtaining the volume corrosion rate and maximum pit depth of several training hardware includes:

[0094] For each training hardware, obtain the initial mass of the uncorroded training hardware;

[0095] Derusting the rusted training hardware to obtain the quality of the derusted training hardware;

[0096] According to the initial mass and the mass of the training hardware after rust removal, the volume corrosion rate of each training hardware is calculated;

[0097] Perform three-dimensional scanning on the corroded training hardware to obtain scanning data;

[0098] The scanned data is binarized, and the area below the binarization threshold is regarded as the etch pit area;

[0099] The maximum depth of all pit areas in the training hardware is calculated by flood filling algorithm;

[0100] The maximum depths of all pit areas were compared to obtain the maximum pit depth of the training hardware.

[0101] For step S3, the corrosion characteristic quantity obtained in step S2 is input into a preset corrosion assessment model, and the model is allowed to determine the volume corrosion rate and the maximum corrosion pit depth of the hardware to be assessed according to the corrosion characteristic quantity of the hardware to be assessed.

[0102] The preset corrosion assessment model is a trained neural network model. Next, the training process of the corrosion assessment model is described in detail:

[0103] like Figure 2 As shown, an embodiment of the present invention provides a training process of a corrosion assessment model, and the training process includes at least the following steps:

[0104] Step S301: obtaining the volume corrosion rate and maximum corrosion pit depth of a number of training hardware;

[0105] In a preferred embodiment, obtaining the volume corrosion rate and maximum pit depth of several training hardware includes:

[0106] For each training hardware, obtain the initial mass of the uncorroded training hardware;

[0107] Derusting the rusted training hardware to obtain the quality of the derusted training hardware;

[0108] According to the initial mass and the mass of the training hardware after rust removal, the volume corrosion rate of each training hardware is calculated;

[0109] Perform three-dimensional scanning on the corroded training hardware to obtain scanning data;

[0110] The scanned data is binarized, and the area below the binarization threshold is regarded as the etch pit area;

[0111] The maximum depth of all pit areas in the training hardware is calculated by flood filling algorithm;

[0112] The maximum depths of all pit areas were compared to obtain the maximum pit depth of the training hardware.

[0113] For step S301, sufficient initial uncorroded training hardware is obtained, and their initial mass W0 is measured in batches. These hardware are placed in the laboratory accelerated corrosion test environment for a period of time to obtain power hardware with different degrees of corrosion. Subsequently, a three-dimensional scanner with an accuracy of ten microns or more is used to scan the rusted area on the surface of the hardware, and the scanning step is set to 50 microns to obtain scanning data of a regular grid. Next, the hardware is derusted using some rust removal methods that do not damage the substrate to obtain the mass W1 of the hardware after derusting. Finally, the volume corrosion rate of each training hardware is calculated by the following formula:

[0114]

[0115] In the formula, η represents the volume corrosion rate of the hardware, W0 represents the initial mass of the uncorroded hardware, and W1 represents the mass of the corroded hardware after rust removal;

[0116] Next, with the help of a program written on the MATLAB platform, a binary slice image is created from top to bottom along the height direction of the data scanned by the 3D scanner. The isolated areas (i.e., pits) are determined and marked and numbered through the classic flood filling algorithm. In each pit, the maximum depth is found along the height direction as h. j After calculating the maximum depth of all pits, the maximum pit depth h of each training hardware can be obtained by statistical calculation.

[0117] Step S302: acquiring images of a plurality of training hardware, and performing grayscale conversion, Gaussian filtering, and piecewise linear enhancement on each training hardware image to obtain preprocessed images of each training hardware;

[0118] For step S302, a high-resolution camera is used to obtain images of each training hardware, so that the frames around the hardware are close to the hardware, so as to facilitate the preprocessing operation of the training hardware image. For the preprocessing stage, the classic weighted average algorithm is used for grayscale conversion, and then the image is Gaussian filtered to remove Gaussian noise in the image, smooth the image, and retain important edge information. Finally, the piecewise linear transformation is used for image enhancement to improve the contrast of the image and make the details clearer. Through the above preprocessing operations, the images of each training hardware after preprocessing can be obtained. These images have better quality, reduced noise, and enhanced contrast, which is conducive to subsequent image analysis and feature extraction.

[0119] Step S303: extracting features from the preprocessed images of each training hardware to obtain corrosion feature quantities of each training hardware;

[0120] In step S303, the feature quantity of each pre-processed training hardware image is extracted. First, for the colorimetric feature, the mean value M of the R component in the RGB color model is extracted. R and standard deviation σ R , and after converting the RGB color model to the HSI color model, extract the mean value M of the hue feature in the HSI color model H and standard deviation σ H , and the mean M of the saturation feature in the HSI color model S and standard deviation σ S ; Then, for texture features, by calculating the gray-level co-occurrence matrix of the image, the energy E, a metric reflecting the coarseness of the image texture, and the randomness S, a metric reflecting the amount of information contained in the image, are calculated; for wavelet features, high-pass filters (H) and low-pass filters (L) with different bandwidths are used in the horizontal and vertical directions of the rust image to obtain four wavelet coefficient sub-images; L / L, L / H, H / L, H / H, and the above process is repeated for L / L to obtain four sub-images of the next level, corresponding to the secondary wavelet transform approximate detail coefficient A2, horizontal detail coefficient H2, vertical detail coefficient G2, and diagonal detail coefficient V2. Using the energy formula, the energy percentage E of the sub-image can be calculated. A2 、E H2 、E V2 and E G2 At the same time, the anisotropic energy value E is introduced O and the high frequency component energy of the sub-image decomposed by the second-level wavelet and E t As the wavelet feature of the rust image; Finally, for the fractal feature, the existing DBC "box dimension method" is used to obtain the fractal feature FD. After the above feature extraction, the rust feature of each training hardware can be obtained [M R , σ R 、M H、M S , σ H , σ S , E, S, E t 、E O , FD] i .

[0121] Step S304: inputting the corrosion feature values ​​of each training hardware into the corrosion assessment model to be trained for iterative training until the sum of squares of the model error is less than a prediction threshold, thereby obtaining a preset corrosion assessment model.

[0122] In step S304, the volume corrosion rate η of each training hardware obtained in step S301 is calculated. i and the maximum pit depth h i , and the corrosion feature value of each training hardware obtained in step S303 [M R , σ R 、M H 、M S , σ H , σ S , E, S, E t 、E O , FD] i , is input into the corrosion assessment model to be trained to predict the volume corrosion rate and maximum pit depth of the hardware. The assessment model adopts a BP neural network model, which is a multi-layer feedforward neural network based on the BP algorithm. The model consists of an input layer, a hidden layer and an output layer. Each layer has multiple nodes. In this embodiment, the input of the BP neural network is set to the feature value of the corroded hardware image, the number of nodes is set to 11, the output is set to the corrosion state evaluation index, the number of nodes is set to 2, and in order to set the number of nodes in the hidden layer, the following empirical formula is used:

[0123]

[0124] In the formula, n represents the number of nodes in the hidden layer of the model, m represents the number of nodes in the input layer of the model, k represents the number of nodes in the output layer of the model, and a is any integer between 1 and 10. In this embodiment, the value of a is set to 5, and the number of hidden layer nodes calculated according to the formula is 9. By inputting the image feature quantity of the corroded hardware, the BP neural network model can learn the relationship between the feature and the volume corrosion rate and the maximum pit depth, and make predictions. In this way, the volume corrosion rate and maximum pit depth of the hardware can be obtained through the corrosion assessment model, thereby evaluating the corrosion condition of the hardware.

[0125] Step S4: Determine the corrosion state of the hardware to be evaluated based on the volume corrosion rate and the maximum corrosion pit depth.

[0126] In a preferred embodiment, the corrosion state of the hardware to be evaluated is determined according to the volume corrosion rate and the maximum corrosion pit depth, including:

[0127] According to the preset corrosion status evaluation grade classification table, the predicted volume corrosion rate and maximum pit depth are matched with the corresponding grades to obtain the corrosion status of the hardware to be evaluated.

[0128] For step S4, according to actual needs, a corrosion status evaluation grade classification table (Table 1) is defined in advance, which divides the volume corrosion rate and the maximum pit depth into different grades. Each grade can represent different degrees of corrosion status, such as weak corrosion, moderate corrosion, strong corrosion and severe corrosion.

[0129] Table 1 Corrosion status evaluation level classification table

[0130]

[0131] The predicted volume corrosion rate and maximum pit depth are matched with the preset corrosion status evaluation grade classification table. According to the matching results, the corrosion status of the hardware to be evaluated is determined. For example, if the volume corrosion rate and maximum pit depth match a certain grade range in the classification table, the corrosion status of the hardware to be evaluated can be determined as the corresponding grade.

[0132] Compared with the previous intelligent power inspection method which only focuses on the corrosion area index, the method for measuring the corrosion status index and image prediction of the transmission line hardware proposed in the present invention focuses on the volume corrosion rate and the maximum pit depth index, so that the inspection personnel can know the corrosion condition inside the hardware. At the same time, the operation and maintenance inspection personnel can also use the method provided by the present invention to train a correlation model between the image features of the corroded hardware and its corrosion status index suitable for different needs, so that after the power inspection image is processed by other methods, the correlation model can directly predict the specific value of the corrosion status index of the hardware in the image, and then the corrosion status of the hardware can be obtained by judging the specific interval to which its corrosion status index belongs. It is no longer necessary to observe the qualitative corrosion status of the corroded hardware only by the naked eye, but it is possible to use multiple indicators to quantitatively obtain the corrosion status.

[0133] like Figure 3 As shown, based on the above method embodiment, a corresponding device embodiment is provided;

[0134] An embodiment of the present invention provides a load action fault warning device based on a Bayesian network, comprising: a hardware image acquisition module to be evaluated, a corrosion feature extraction module, an index prediction module, and a state matching module;

[0135] A hardware image acquisition module to be evaluated, used to acquire images of the hardware to be evaluated;

[0136] The rust feature extraction module is used to extract features from the image of the hardware to be evaluated, and obtain the rust feature quantity of the hardware to be evaluated; the rust feature quantity includes: colorimetric feature quantity, texture feature quantity, wavelet feature quantity and fractal feature quantity;

[0137] An indicator prediction module is used to input the corrosion feature quantity into a preset corrosion assessment model, so that the corrosion assessment model determines the volume corrosion rate and maximum pit depth of the hardware to be assessed according to the corrosion feature quantity of the hardware to be assessed; wherein the corrosion assessment model is trained with the corrosion feature quantity of the training hardware as input and the predicted volume corrosion rate and predicted maximum pit depth of the training hardware as output, and during the training process, the network parameters of the corrosion assessment model are adjusted according to the predicted volume corrosion rate and actual volume corrosion rate of the training hardware, and the predicted maximum pit depth and actual maximum pit depth of the training hardware;

[0138] The state matching module is used to determine the corrosion state of the hardware to be evaluated based on the volume corrosion rate and the maximum corrosion pit depth.

[0139] It can be understood that the above-mentioned device item embodiments correspond to the method item embodiments of the present invention, and can implement the method for evaluating the corrosion status of transmission line hardware provided by any one of the above-mentioned method item embodiments of the present invention.

[0140] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art may understand and implement the present invention without creative work.

[0141] Based on the above-mentioned embodiment of the method for assessing the corrosion status of transmission line fittings, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the method for assessing the corrosion status of transmission line fittings according to any embodiment of the present invention is implemented.

[0142] Exemplarily, in this embodiment, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to implement the present invention. The one or more module elements may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the terminal device.

[0143] The terminal device may be a computing device such as a desktop computer, a notebook, a palmtop computer, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0144] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0145] Based on the above method embodiment, another embodiment is provided: another embodiment of the present invention provides a computer-readable storage medium, including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for evaluating the corrosion status of transmission line hardware described in any one of the above method embodiments of the present invention.

[0146] Wherein, the module / unit integrated in the transmission line hardware corrosion status assessment device / terminal equipment, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0147] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for evaluating the corrosion status of transmission line hardware, characterized in that: include: Obtain an image of the hardware to be evaluated; Extracting features from the image of the hardware to be evaluated to obtain corrosion feature quantities of the hardware to be evaluated; The corrosion feature quantity includes: colorimetric feature quantity, texture feature quantity, wavelet feature quantity and fractal feature quantity; Inputting the corrosion characteristic quantity into a preset corrosion assessment model, so that the corrosion assessment model determines the volume corrosion rate and the maximum corrosion pit depth of the hardware to be assessed according to the corrosion characteristic quantity of the hardware to be assessed; Determine the corrosion state of the hardware to be evaluated according to the volume corrosion rate and the maximum corrosion pit depth; The corrosion assessment model is trained with the corrosion characteristics of the training hardware as input and the predicted volume corrosion rate and predicted maximum pit depth of the training hardware as output, and the network parameters of the corrosion assessment model are adjusted according to the predicted volume corrosion rate and actual volume corrosion rate of the training hardware, as well as the predicted maximum pit depth and actual maximum pit depth of the training hardware during the training process.

2. The method for evaluating the corrosion status of transmission line hardware according to claim 1, characterized in that: According to the volume corrosion rate and the maximum pit depth, determine the corrosion state of the hardware to be evaluated, including: According to the preset corrosion status evaluation grade classification table, the predicted volume corrosion rate and maximum pit depth are matched with the corresponding grades to obtain the corrosion status of the hardware to be evaluated.

3. The method for evaluating the corrosion status of transmission line hardware according to claim 2, characterized in that: The preset corrosion assessment model is trained in the following way: Obtain the volume corrosion rate and maximum pit depth of several training hardware; Acquire the images of the plurality of training hardware, and perform grayscale conversion, Gaussian filtering, and piecewise linear enhancement on each training hardware image to obtain a preprocessed image of each training hardware; Performing feature extraction on the preprocessed images of each training hardware to obtain a corrosion feature value of each training hardware; The corrosion feature quantities of the training hardware are input into the corrosion assessment model to be trained for iterative training until the sum of squares of the model error is less than a prediction threshold, thereby obtaining the preset corrosion assessment model.

4. The method for evaluating the corrosion status of transmission line hardware according to claim 3, characterized in that: Obtain the volume corrosion rate and maximum pit depth of several training hardware, including: For each training hardware, obtain the initial mass of the uncorroded training hardware; Derusting the rusted training hardware to obtain the quality of the derusted training hardware; Calculating the volume corrosion rate of each training hardware according to the initial mass and the mass of the training hardware after rust removal; Perform three-dimensional scanning on the corroded training hardware to obtain scanning data; Binarization is performed on the scanned data, and an area below a binarization threshold is regarded as an etch pit area; The maximum depth of all pit areas in the training hardware is calculated by flood filling algorithm; The maximum depths of all pit areas were compared to obtain the maximum pit depth of the training hardware.

5. The method for evaluating the corrosion status of transmission line hardware according to claim 4, characterized in that: The colorimetric feature includes: the mean of the R component in the RGB color model, the standard deviation of the R component in the RGB color model, the mean of the H component in the HSI color model, the standard deviation of the H component in the HSI color model, the mean of the S component in the HSI color model, and the standard deviation of the S component in the HSI color model; The colorimetric feature is extracted by the following method: Convert the image into RGB color model and calculate the mean and standard deviation of the R component in the RGB color model; The RGB color model is converted into an HSI color model, and a mean value of an H component in the HSI color model, a standard deviation of the H component in the HSI color model, a mean value of an S component in the HSI color model, and a standard deviation of the S component in the HSI color model are calculated.

6. The method for evaluating the corrosion status of transmission line hardware according to claim 5, characterized in that: The texture feature includes: energy, a measurement index reflecting the coarseness of the image texture, and randomness, a measurement index reflecting the amount of information contained in the image; The texture feature is extracted by: Converting the image into a grayscale image, and calculating a grayscale co-occurrence matrix based on the grayscale image; According to the probability of each gray value in the gray image appearing in the gray level co-occurrence matrix, the energy, a metric reflecting the coarseness of the image texture, and the randomness, a metric reflecting the amount of information contained in the image, are calculated.

7. The method for evaluating the corrosion status of transmission line hardware according to claim 6, characterized in that: The wavelet feature includes: anisotropic energy value and high frequency component energy sum of secondary wavelet decomposition sub-image; The wavelet feature is extracted by the following method: Performing a secondary wavelet transform on the image to obtain an approximate detail coefficient, a horizontal detail coefficient, a vertical detail coefficient, and a diagonal detail coefficient of the secondary wavelet transform; According to the approximate detail coefficient, horizontal detail coefficient, vertical detail coefficient, and diagonal detail coefficient of the secondary wavelet transform, the anisotropic energy value and the energy sum of the high-frequency components of the secondary wavelet decomposition sub-image are calculated by the following calculation formula: AND t =And H2 +E V2 +E G2 ; Among them, E t represents the energy sum of high frequency components of the sub-image decomposed by secondary wavelet, E H2 Represents the energy percentage of the secondary wavelet transform approximate detail coefficient, E V2 Represents the energy percentage of the diagonal detail coefficients of the secondary wavelet transform, E G2 Indicates the energy percentage of the secondary wavelet transform vertical detail coefficient, E O Anisotropic energy value.

8. A device for evaluating the corrosion status of power transmission line hardware, characterized in that: include: Hardware image acquisition module to be evaluated, corrosion feature extraction module, index prediction module, and state matching module; The hardware to be evaluated image acquisition module is used to acquire the image of the hardware to be evaluated; The corrosion feature extraction module is used to extract features from the image of the hardware to be evaluated to obtain the corrosion feature value of the hardware to be evaluated; The corrosion feature quantity includes: colorimetric feature quantity, texture feature quantity, wavelet feature quantity and fractal feature quantity; The indicator prediction module is used to input the corrosion characteristic quantity into a preset corrosion assessment model, so that the corrosion assessment model determines the volume corrosion rate and maximum pit depth of the hardware to be assessed according to the corrosion characteristic quantity of the hardware to be assessed; wherein the corrosion assessment model is trained with the corrosion characteristic quantity of the training hardware as input and the predicted volume corrosion rate and predicted maximum pit depth of the training hardware as output, and during the training process, the network parameters of the corrosion assessment model are adjusted according to the predicted volume corrosion rate and actual volume corrosion rate of the training hardware, and the predicted maximum pit depth and actual maximum pit depth of the training hardware; The state matching module is used to determine the corrosion state of the hardware to be evaluated according to the volume corrosion rate and the maximum corrosion pit depth.

9. A terminal device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for evaluating the corrosion status of transmission line hardware according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: include: A stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for evaluating the corrosion status of transmission line hardware according to any one of claims 1 to 7.

Images