Pollution level assessment method based on visible light image of insulator

Abstract

The present invention provides a method for evaluating the pollution level based on the visible light image of an insulator, comprising: S1. collecting the visible light image of the insulator under natural light conditions, and graying the visible light image; S2. edge detection is performed on the graying pre-processed image to output the edge detection image, and the edge detection image is morphologically processed to segment the insulator disk image; S3. feature extraction is performed on the insulator judgment image to obtain the insulator disk color feature component, and the Fisher criterion function is used to filter out the feature quantity corresponding to the pollution level of the insulator; S4. the feature quantity obtained in step S3 is input into a support vector machine classifier for training, and the support vector machine classifier outputs the evaluation result of the insulator pollution level. Through the above method, the pollution level of the insulator can be accurately judged, thereby effectively ensuring the detection accuracy, and the detection can be completed without the conditions such as the insulator itself heating or discharging, so that the insulation pollution level detection is real-time, efficient, safer and more reliable.

Description

Technical Field

[0001] The invention relates to an insulator contamination assessment method, and in particular to a contamination level assessment method based on an insulator visible light image. Background Art

[0002] Insulators in transmission lines operate in outdoor environments all year round. Dust particles in the air will continue to deposit on the surface of the insulators, and eventually adhere to the surface of the insulators to form a dirty layer, which is easy to cause pollution flashover accidents and seriously threaten the safety of the power system. At present, targeted cleaning according to the degree of pollution accumulation on the insulators is the most effective pollution flashover prevention measure. In order to provide accurate pollution status information for cleaning, operating experience shows that a simple, economical, safe and accurate insulator pollution detection method is urgently needed to provide the most important technical guarantee for anti-pollution work.

[0003] In the existing technologies, such as leakage current detection method, infrared or ultraviolet imaging method, the environment requirements are strict, and the detection can only be performed when the insulator is heated or discharged, which leads to untimely detection and low accuracy of detection results, especially the detection efficiency is difficult to guarantee.

[0004] Therefore, in order to solve the above technical problems, it is urgent to propose a new technical means. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a pollution level assessment method based on visible light images of insulators. By collecting images of insulators under visible light, the pollution level of the insulators can be accurately judged, thereby effectively ensuring the detection accuracy, and the detection can be completed without the need for the insulators to heat up or discharge themselves, thereby making the insulation pollution level detection real-time, efficient, safer and more reliable.

[0006] The present invention provides a method for evaluating the pollution level of an insulator based on a visible light image, comprising the following steps:

[0007] S1. Collect visible light images of insulators under natural light conditions, and perform grayscale preprocessing on the visible light images;

[0008] S2. Perform edge detection on the grayscale preprocessed image to output an edge detection image, perform morphological processing on the edge detection image, and segment the insulator disk image;

[0009] S3. Extract the features of the insulator judgment image to obtain the color feature components of the insulator disk surface, and use the Fisher criterion function to filter out the feature quantities corresponding to the insulator contamination levels; the insulator contamination levels are 0, I, II, III and IV in order from small to large according to the degree of contamination accumulation;

[0010] S4. Input the feature quantity obtained in step S3 into the support vector machine classifier for training, and the support vector machine classifier outputs the insulator contamination level evaluation result.

[0011] Further, in step S1, the grayscale preprocessing of the visible light image is determined by the following method:

[0012] V gray =0.3R+0.59G+0.11B; where: R, G and B are the RGB values ​​of the visible light image, V gray is the calculated gray value.

[0013] Furthermore, the canny edge detection method is used to detect the edge of the image after grayscale preprocessing, which specifically includes:

[0014] S21. Use a Gaussian kernel function to perform Gaussian smoothing on the grayscale image, where the Gaussian kernel function is:

[0015] Where, σ is the standard deviation of the Gaussian kernel function;

[0016] S22. Calculate the first-order derivatives in the horizontal and vertical directions of the image after Gaussian smoothing using the Sobel operator to obtain the gradient value and gradient direction of each pixel of the image;

[0017] S23. Add sub-pixel points to the image according to the angle by linear interpolation, perform non-maximum suppression, and retain the maximum value between pixels with the same gradient direction;

[0018] S24. Use dual threshold detection to determine whether the pixel belongs to the edge part. Set a low threshold θ1 and a high threshold θ2, where: θ1 = 0.5θ2;

[0019] S25. Mark the points in the image with grayscale values ​​less than θ1 as 0, and the points with grayscale values ​​greater than θ2 as 1. The points with grayscale values ​​greater than θ1 and less than θ2 are determined using 8-connected regions. When there is a point with a grayscale value greater than θ2 in the 8-connected region of the point, it can be considered as an edge point and marked as 1.

[0020] Further, the high threshold θ2 is determined by the following method:

[0021] S241. Calculate the number of each gray level of the image and the probability of each gray level pixel appearing;

[0022] S242. Calculate the grayscale mean of the background class and the target class and the occurrence probability of the background class and the target class;

[0023] S243. Calculate the class separation index of the background class and the target class, that is, the variance between the two classes; and find the point where the variance has the maximum value as the high threshold θ2.

[0024] Further, step S3 includes:

[0025] The source image a and the reference image b are converted from RGB space to LMS space, and then to Lab space. The mean and standard deviation of the three channels of the image are calculated respectively. The color attribute of the reference image is added to the source image to obtain the three channels of the Lab space of the target image. Then the target image is reversely converted to RGB space to obtain the pixel values ​​of the three channels, which are recorded as R′, G′, and B′.

[0026]

[0027]

[0028]

[0029] Where: l′ a , α′ a , β′ a and a , α a , β a They are the three channels of the Lab space of the target image and the source image respectively. are the means of the three channels of the source image and the reference image respectively, is the standard deviation;

[0030] The R′, G′, and B′ components are corrected to new R, G, and B components, and similarly, new H, S, and V components are obtained to compensate for the effect of illumination changes on pixel values; the correction formula is:

[0031]

[0032]

[0033]

[0034] The insulator disk binary image is multiplied by the R, G, B components and H, S, V components obtained by compensating for illumination changes to obtain a color image containing only the insulator disk surface;

[0035] Extract the characteristic parameters of RGB and HSV color spaces respectively, including the mean, median, maximum, minimum, range and variance of each component of R, G, B and H, S, V;

[0036] The Fisher criterion function is used to identify the component characteristics with obvious numerical differences between different pollution conditions of each insulator from the characteristic parameters, which is used as an intuitive description of the pollution condition of the insulator. The Fisher criterion function value J is taken i Larger feature quantity; among which:

[0037] Fisher criterion function value J i The expression is:

[0038]

[0039] Where: N ai is the intra-class variance, N bi is the between-class variance.

[0040] For the insulator image, the intra-class variance N of the i-th dimension feature ai The expression is:

[0041]

[0042] Between-class variance N bi The expression is:

[0043]

[0044] Where: k is the number of dirty categories of the sample, is the eigenvalue of the i-th dimension of a single sample of the dirty category k; ω k The sample set of dirty category k; is the average value of the eigenvalue of the i-th dimension of a single sample of the dirty category k, y i is the average value of the eigenvalue of the i-th dimension of all dirty category samples; is the number of eigenvalues ​​of the i-th dimension of a single sample of the pollution category k; n i is the total number of eigenvalues ​​of the i-th dimension of all dirty category samples.

[0045] Further, step S4 includes:

[0046] The input of the support vector machine classifier is the feature parameter obtained by the Fisher criterion function, and the input data is normalized;

[0047] The support vector machine kernel function selects the radial basis kernel function and determines the optimal parameters of the penalty coefficient c and the kernel function parameter g;

[0048] The image feature quantities processed in step S3 are divided into a training set and a test set. The data of the training set is input into a multi-value classifier for training to obtain a model, and then the test set samples are input into the multi-value classifier to obtain the insulator contamination level classification result.

[0049] The beneficial effects of the present invention are as follows: through the present invention, the pollution level of the insulator can be accurately judged by collecting images of the insulator under visible light, thereby effectively ensuring the detection accuracy, and the detection can be completed without the need for the insulator itself to generate heat or discharge, thereby making the insulation pollution level detection real-time, efficient, safer and more reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0051] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0052] The present invention is further described in detail below:

[0053] The present invention provides a method for evaluating the pollution level of an insulator based on a visible light image, comprising the following steps:

[0054] S1. Collect visible light images of insulators under natural light conditions, and perform grayscale preprocessing on the visible light images;

[0055] S2. Perform edge detection on the grayscale preprocessed image to output an edge detection image, perform morphological processing on the edge detection image, and segment the insulator disk image;

[0056] S3. Extract the features of the insulator judgment image to obtain the color feature components of the insulator disk surface, and use the Fisher criterion function to filter out the feature quantities corresponding to the insulator contamination levels; the insulator contamination levels are 0, I, II, III and IV in order from small to large according to the degree of contamination accumulation;

[0057] S4. The feature quantity obtained in step S3 is input into the support vector machine classifier for training, and the support vector machine classifier outputs the evaluation result of the insulator pollution level. Through the above method, the pollution level of the insulator can be accurately judged, thereby effectively ensuring the detection accuracy, and the detection can be completed without the conditions such as heating or discharge of the insulator itself, so that the insulation pollution level detection is real-time, efficient, safer and more reliable.

[0058] In this embodiment, in step S1, the grayscale preprocessing of the visible light image is determined by the following method:

[0059] V gray =0.3R+0.59G+0.11B; where: R, G and B are the RGB values ​​of the visible light image, V gray is the calculated gray value.

[0060] In this embodiment, the canny edge detection method is used to perform edge detection on the image after grayscale preprocessing, which specifically includes:

[0061] S21. Use a Gaussian kernel function to perform Gaussian smoothing on the grayscale image, where the Gaussian kernel function is:

[0062] Where, σ is the standard deviation of the Gaussian kernel function;

[0063] S22. Calculate the first-order derivatives in the horizontal and vertical directions of the image after Gaussian smoothing using the Sobel operator to obtain the gradient value and gradient direction of each pixel of the image;

[0064] S23. Add sub-pixel points to the image according to the angle by linear interpolation, perform non-maximum suppression, and retain the maximum value between pixels with the same gradient direction;

[0065] S24. Use dual threshold detection to determine whether the pixel belongs to the edge part. Set a low threshold θ1 and a high threshold θ2, where: θ1 = 0.5θ2;

[0066] S25. Mark the points in the image with grayscale values ​​less than θ1 as 0, and the points with grayscale values ​​greater than θ2 as 1. The points with grayscale values ​​greater than θ1 and less than θ2 are determined using 8-connected regions. When there is a point with a grayscale value greater than θ2 in the 8-connected region of the point, it can be considered as an edge point and marked as 1.

[0067] Specifically: the high threshold θ2 is determined by the following method:

[0068] S241. Calculate the number of each gray level of the image and the probability of each gray level pixel appearing;

[0069] S242. Calculate the grayscale mean of the background class and the target class and the occurrence probability of the background class and the target class;

[0070] S243. Calculate the class separation index of the background class and the target class, that is, the variance between the two classes; and find the point where the variance has the maximum value as the high threshold θ2. Among them, each calculation process in S241-S242 can use existing algorithms, which belong to the prior art and are not described here; the image after edge detection is also subjected to existing algorithms such as expansion operation, corrosion operation, opening operation, closing operation, etc. to remove noise in the binary image, and smooth the edge and fill the hole. These processing processes are all prior art.

[0071] In this embodiment, step S3 includes:

[0072] The source image a and the reference image b are converted from RGB space to LMS space, and then to Lab space. The mean and standard deviation of the three channels of the image are calculated respectively. The color attribute of the reference image is added to the source image to obtain the three channels of the Lab space of the target image. Then the target image is reversely converted to RGB space to obtain the pixel values ​​of the three channels, which are recorded as R′, G′, and B′.

[0073]

[0074]

[0075]

[0076] Where: l′ a , α′ a , β′ a and a , α a , β a They are the three channels of the Lab space of the target image and the source image respectively. are the means of the three channels of the source image and the reference image respectively, is the standard deviation;

[0077] The R′, G′, and B′ components are corrected to new R, G, and B components, and similarly, new H, S, and V components are obtained to compensate for the effect of illumination changes on pixel values; the correction formula is:

[0078]

[0079]

[0080]

[0081] The insulator disk binary image is multiplied by the R, G, B components and H, S, V components obtained by compensating for illumination changes to obtain a color image containing only the insulator disk surface;

[0082] Extract the characteristic parameters of RGB and HSV color spaces respectively, including the mean, median, maximum, minimum, range and variance of each component of R, G, B and H, S, V;

[0083] The Fisher criterion function is used to identify the component characteristics with obvious numerical differences between different pollution conditions of each insulator from the characteristic parameters, which is used as an intuitive description of the pollution condition of the insulator. The Fisher criterion function value J is taken i Larger feature quantity; among which:

[0084] Fisher criterion function value J i The expression is:

[0085]

[0086] Where: N ai is the intra-class variance, N bi is the between-class variance.

[0087] For the insulator image, the intra-class variance N of the i-th dimension feature ai The expression is:

[0088]

[0089] Between-class variance N bi The expression is:

[0090]

[0091] Where: k is the number of dirty categories of the sample, is the eigenvalue of the i-th dimension of a single sample of the dirty category k; ω i The sample set of dirty category k; is the average value of the eigenvalue of the i-th dimension of a single sample of the dirty category k, y i is the average value of the eigenvalue of the i-th dimension of all dirty category samples; is the number of eigenvalues ​​of the i-th dimension of a single sample of the pollution category k; n i is the total number of eigenvalues ​​of the i-th dimension of all dirty category samples.

[0092] In this embodiment, step S4 includes:

[0093] The input of the support vector machine classifier is the feature parameter obtained by the Fisher criterion function, and the input data is normalized;

[0094] The support vector machine kernel function selects the radial basis kernel function and determines the optimal parameters of the penalty coefficient c and the kernel function parameter g;

[0095] The image feature quantities processed in step S3 are divided into a training set and a test set. The data of the training set is input into a multi-value classifier for training to obtain a model, and then the test set samples are input into the multi-value classifier to obtain the insulator contamination level classification result.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A method for evaluating the pollution level of an insulator based on a visible light image, characterized in that: The following steps are involved: S1. Collect visible light images of insulators under natural light conditions, and perform grayscale preprocessing on the visible light images; S2. Perform edge detection on the grayscale preprocessed image to output an edge detection image, perform morphological processing on the edge detection image, and segment the insulator disk image; S3. Extract the features of the insulator judgment image to obtain the color feature components of the insulator disk surface, and use the Fisher criterion function to filter out the feature quantities corresponding to the insulator contamination levels; the insulator contamination levels are 0, I, II, III and IV in order from small to large according to the degree of contamination accumulation; S4. The feature quantity obtained in step S3 is input into the support vector machine classifier for training, and the support vector machine classifier outputs the insulator contamination level evaluation result; Step S3 includes: The source image a and the reference image b are converted from RGB space to LMS space, and then to Lab space. The mean and standard deviation of the three channels of the image are calculated respectively. The color attribute of the reference image is added to the source image to obtain the three channels of the target image in Lab space. Then the target image is reversely converted to RGB space to obtain the pixel values ​​of the three channels, which are recorded as R′, G′, and B′. Where: l′ a , α′ a , β′ a and a , α a , β a They are the three channels of the Lab space of the target image and the source image respectively. are the means of the three channels of the source image and the reference image respectively, is the standard deviation; The R′, G′, and B′ components are corrected to new R, G, and B components, and similarly, new H, S, and V components are obtained to compensate for the effect of illumination changes on pixel values; the correction formula is: The insulator disk binary image is multiplied by the R, G, B components and H, S, V components obtained by compensating for illumination changes to obtain a color image containing only the insulator disk surface; Extract the characteristic parameters of RGB and HSV color spaces respectively, including the mean, median, maximum, minimum, range and variance of each component of R, G, B and H, S, V; The Fisher criterion function is used to identify the component characteristics with obvious numerical differences between different pollution conditions of each insulator from the characteristic parameters, which is used as an intuitive description of the pollution condition of the insulator. The Fisher criterion function value J is taken i Larger feature quantity; among which: Fisher criterion function value J i The expression is: Where: N ai is the intra-class variance, N bi is the between-class variance; For the insulator image, the intra-class variance N of the i-th dimension feature ai The expression is: Between-class variance N bi The expression is: Where: k is the number of dirty categories of the sample, is the eigenvalue of the i-th dimension of a single sample of the dirty category k; ω k The sample set of dirty category k; is the average value of the eigenvalue of the i-th dimension of a single sample of the dirty category k, y i is the average value of the eigenvalue of the i-th dimension of all dirty category samples; is the number of eigenvalues ​​of the i-th dimension of a single sample of the pollution category k; n i is the total number of eigenvalues ​​of the i-th dimension of all dirty category samples.

2. The method for evaluating the contamination level based on the visible light image of an insulator according to claim 1 is characterized in that: In step S1, the grayscale preprocessing of the visible light image is determined by the following method: V gray =0.3R+0.59G+0.11B; where: R, G and B are the RGB values ​​of the visible light image, V gray is the calculated gray value.

3. The method for evaluating the contamination level based on the visible light image of an insulator according to claim 1 is characterized in that: The canny edge detection method is used to detect the edge of the image after grayscale preprocessing, which includes: S21. Use a Gaussian kernel function to perform Gaussian smoothing on the grayscale image, where the Gaussian kernel function is: Where, σ is the standard deviation of the Gaussian kernel function; S22. Calculate the first-order derivatives in the horizontal and vertical directions of the image after Gaussian smoothing using the Sobel operator to obtain the gradient value and gradient direction of each pixel of the image; S23. Add sub-pixel points to the image according to the angle by linear interpolation, perform non-maximum suppression, and retain the maximum value between pixels with the same gradient direction; S24. Use dual threshold detection to determine whether the pixel belongs to the edge part; set a low threshold θ1 and a high threshold θ2, where: θ1 = 0.5θ2; S25. Mark the points in the image with grayscale values ​​less than θ1 as 0, and the points with grayscale values ​​greater than θ2 as 1. The points with grayscale values ​​greater than θ1 and less than θ2 are determined using 8-connected regions. When there is a point with a grayscale value greater than θ2 in the 8-connected region of the point, it is considered to be an edge point and marked as 1.

4. The method for evaluating the contamination level based on the visible light image of an insulator according to claim 3 is characterized by: The high threshold θ2 is determined by the following method: S241. Calculate the number of each gray level of the image and the probability of each gray level pixel appearing; S242. Calculate the grayscale mean of the background class and the target class and the occurrence probability of the background class and the target class; S243. Calculate the class separation index of the background class and the target class, that is, the variance between the two classes; and find the point where the variance has the maximum value as the high threshold θ2.

5. The method for evaluating the contamination level based on the visible light image of an insulator according to claim 1 is characterized in that: Step S4 includes: The input of the support vector machine classifier is the feature parameter obtained by the Fisher criterion function, and the input data is normalized; The support vector machine kernel function selects the radial basis kernel function and determines the optimal parameters of the penalty coefficient c and the kernel function parameter g; The image feature quantities processed in step S3 are divided into a training set and a test set. The data of the training set is input into a multi-value classifier for training to obtain a model, and then the test set samples are input into the multi-value classifier to obtain the insulator contamination level classification result.

Images