A transformer breather discoloration detection method and system

Abstract

The application provides a transformer breather discoloration detection method and system, and belongs to the field of transformer breathers. The transformer breather discoloration detection method comprises the following steps: obtaining a to-be-identified breather image; enhancing the to-be-identified breather image by using a multi-scale Retinex image enhancement algorithm to obtain an enhanced image; converting the enhanced image to an HSV color space to obtain a converted image; extracting a silica gel particle region from the converted image based on a maximum connected domain algorithm; calculating the hue threshold of each column of pixels in the H channel data in the silica gel particle region according to the hue of each pixel in the silica gel particle region; fitting the hue threshold of each column of pixels to determine a segmentation line; and dividing the silica gel particle region into a discoloration sub-region and a non-discoloration sub-region by using the segmentation line. The application can be applied to different field environments and improves the accuracy of transformer breather discoloration detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of transformer breather, in particular to a transformer breather discoloration detection method and system. BACKGROUND

[0002] As the basic unit of power system operation, the operation state of power equipment affects the safety and stability of the power grid and the production efficiency. With the rapid development of the national economy, more and more large-scale generating units have been put into operation, and the capacity of the units has been increasing. As the main component of power supply, power equipment bears a heavy power supply task. The power transformer is an important equipment in the substation, and once a fault occurs, it will have a serious impact on the normal operation of the power system. The breather is an important equipment for purifying the air inhaled by the transformer, which can effectively filter the moisture in the air and reduce the moisture and oxidation of the transformer oil. If the breather is saturated with adsorbed moisture and becomes discolored, it is easy to cause the transformer oil to be damp, which will cause the insulation strength of the transformer oil to decrease and cause the risk of internal failure of the transformer. Therefore, it is very important to find and replace the failed breather silica gel in time to ensure the normal operation of the transformer. The current solution is to replace the intelligent maintenance-free breather or install new devices and equipment such as online monitoring devices for the breather.

[0003] The method of data preprocessing for detection by Xinmingyong et al. is to detect through homomorphic filtering, then perform color statistics on the H and S components in the HSV (Hue-Saturation-Value, hue-saturation-value) color space, and make a decision based on the statistical data. However, the scene in the substation is complex, the background of the equipment image is usually messy, and the outside of the transformer breather is mostly transparent glass material and is placed outdoors, which is easily affected by environmental light and other factors, and the shooting angle also produces different degrees of high reflection area in the image. This situation will cause the homomorphic filtering and histogram statistics method to fail, and the detection effect will be poor, so this method can only be used in specific scenarios, and has great limitations.

[0004] Li Ruisheng et al. collected samples at different times, summarized the HSV prior threshold value by artificial, then extracted the color by using the artificial prior threshold value, and judged the discoloration degree of the breather according to the color proportion. Since the prior threshold value is summarized by artificial in the actual use process of this algorithm, the threshold value is not adaptive, and the effect is poor; at the same time, the existing detection method is to calculate the discoloration degree of the breather by counting the pixel points of the silica gel particles, which does not conform to the discoloration rule of the breather. SUMMARY

[0005] The purpose of the present application is to provide a transformer breather discoloration detection method and system, which can improve the accuracy of transformer breather discoloration detection.

[0006] To achieve the above object, the present application provides the following scheme:

[0007] A transformer breather discoloration detection method, comprising:

[0008] Obtaining a to-be-identified breather image;

[0009] Enhancing the to-be-identified breather image by using a multi-scale Retinex image enhancement algorithm to obtain an enhanced image;

[0010] Converting the enhanced image to an HSV color space to obtain a converted image;

[0011] Extracting a silica gel particle region from the converted image based on a maximum connected domain algorithm; the silica gel particle region includes silica gel H channel data, silica gel S channel data, and silica gel V channel data;

[0012] According to the hue of each pixel in the silica gel particle region, calculating the hue threshold of each column of pixels in the silica gel H channel data;

[0013] Fitting the hue threshold of each column of pixels to determine a segmentation line;

[0014] Dividing the silica gel particle region into a discolored sub-region and a non-discolored sub-region by using the segmentation line.

[0015] Optionally, the transformer breather discoloration detection method further comprises:

[0016] When the color of the discolored sub-region reaches a pre-set color threshold and the area of the discolored sub-region is greater than a pre-set area threshold, generating a pre-warning information of replacing the breather silica gel.

[0017] Optionally, the obtaining of the to-be-identified breather image specifically comprises:

[0018] Shooting the to-be-identified breather image by using a visible light camera.

[0019] Optionally, the converted image includes initial H channel data, initial S channel data, and initial V channel data;

[0020] The extracting of the silica gel particle region from the converted image based on the maximum connected domain algorithm specifically comprises:

[0021] Performing median filtering on the initial H channel data to obtain H channel filtering information;

[0022] According to the H channel filtering information, the initial S channel data, and the initial V channel data, determining a binary image;

[0023] A morphological opening operation is performed on the binary image to obtain a smoothed image;

[0024] The maximum connected component search algorithm is used to perform maximum connected component search on the smoothed image to determine the initial silicone particle region in the smoothed image;

[0025] The initial silicone particle region is ANDed with the initial H-channel filter information to determine the final silicone particle region.

[0026] Optionally, determining the binary image based on the H-channel filtering information, the initial S-channel data, and the initial V-channel data specifically includes:

[0027] The H-channel filtered information is binarized to obtain H-channel binary data;

[0028] The initial S-channel data is binarized to obtain S-channel binary data;

[0029] The initial V-channel data is binarized to obtain V-channel binary data;

[0030] A binary image is determined based on the binary data of the H channel, the binary data of the S channel, and the binary data of the V channel.

[0031] Optionally, the step of using a maximum connected component algorithm to perform a maximum connected component search on the smoothed image to determine the initial silicone grain region in the smoothed image specifically includes:

[0032] The Canny edge detection algorithm is used to detect connected regions in the smoothed image;

[0033] The largest connected region in the smoothed image is determined based on the size of each connected region; the largest connected region is the initial silicone particle region.

[0034] To achieve the above objectives, the present invention also provides the following solution:

[0035] A transformer breather color change detection system, comprising:

[0036] An image acquisition unit is used to acquire an image of the respirator to be identified.

[0037] An enhancement unit, connected to the image acquisition unit, is used to enhance the image of the respirator to be identified using a multi-scale Retinex image enhancement algorithm to obtain an enhanced image;

[0038] A conversion unit, connected to the enhancement unit, is used to convert the enhanced image to the HSV color space to obtain a converted image;

[0039] The region determination unit is connected with the conversion unit and configured to extract a silica gel particle region from the converted image based on a maximum connected domain algorithm; the silica gel particle region includes silica gel H channel data, silica gel S channel data and silica gel V channel data.

[0040] The hue threshold calculation unit is connected with the region determination unit and configured to calculate a hue threshold of each column of pixels in the silica gel H channel data according to the hue of each pixel in the silica gel particle region.

[0041] The split line determination unit is connected with the hue threshold calculation unit and configured to fit the hue threshold of each column of pixels to determine a split line.

[0042] The split unit is connected with the split line determination unit and the region determination unit and configured to divide the silica gel particle region into a color-changing sub-region and a non-color-changing sub-region by using the split line.

[0043] Optionally, the transformer respirator color change detection system further includes:

[0044] The warning unit is connected with the split unit and configured to generate a warning information of replacing the silica gel of the respirator when the color of the color-changing sub-region reaches a pre-set color threshold and the area of the color-changing sub-region is greater than a pre-set area threshold.

[0045] Optionally, the converted image includes initial H channel data, initial S channel data and initial V channel data.

[0046] The region determination unit includes:

[0047] The filtering module is connected with the conversion unit and configured to perform median filtering on the initial H channel data to obtain H channel filtering information.

[0048] The binary module is connected with the filtering module and the conversion unit and configured to determine a binary image according to the H channel filtering information, the initial S channel data and the initial V channel data.

[0049] The smoothing module is connected with the binary module and configured to perform morphological opening operation on the binary image to obtain a smoothed image.

[0050] The connected domain search module is connected with the smoothed image and configured to perform maximum connected domain search on the smoothed image by using a maximum connected domain algorithm to determine an initial silica gel particle region in the smoothed image.

[0051] The final region determination module is connected with the connected domain search module and configured to perform AND operation on the initial silica gel particle region and the initial H channel filtering information to determine a final silica gel particle region.

[0052] Optionally, the binary module comprises:

[0053] An H channel binary sub-module, connected with the filter module, configured to binarize the H channel filter information to obtain H channel binary data;

[0054] An S channel binary sub-module, connected with the conversion unit, configured to binarize the initial S channel data to obtain S channel binary data;

[0055] A V channel binary sub-module, connected with the conversion unit, configured to binarize the initial V channel data to obtain V channel binary data;

[0056] A binary image determination sub-module, connected with the H channel binary sub-module, the S channel binary sub-module and the V channel binary sub-module, configured to determine a binary image according to the H channel binary data, the S channel binary data and the V channel binary data.

[0057] According to the embodiments of the present application, the following technical effects are achieved: the multi-scale Retinex image enhancement algorithm is used to enhance the to-be-identified respirator image, the contrast between colors is enhanced and the influence of high reflection on the image is eliminated, the enhanced image is converted to the HSV color space to obtain a converted image, the silica gel particle region is extracted from the converted image based on the maximum connected domain algorithm, the hue threshold of each column of pixels in the H channel data of the silica gel particle region is calculated, the hue threshold of each column of pixels is fitted to determine a segmentation line, the silica gel particle region is divided into a discolored sub-region and a non-discolored sub-region by using the segmentation line, the adaptive adjustment of the color segmentation threshold is not dependent on prior knowledge, the method can be applied to different field environments, and the accuracy of the transformer respirator discoloration detection is improved. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0059] Figure 1 The flowchart of the transformer respirator discoloration detection method of the present application;

[0060] Figure 2 The schematic diagram of the transformer respirator discoloration detection process;

[0061] Figure 3 The module structure schematic diagram of the transformer respirator discoloration detection system of the present application.

[0062] Symbol explanation:

[0063] Image acquisition unit-1, enhancement unit-2, conversion unit-3, region determination unit-4, tone threshold calculation unit-5, split line determination unit-6, split unit-7, early warning unit-8. DETAILED DESCRIPTION

[0064] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0065] The purpose of the present application is to provide a transformer breather discoloration detection method and system, which combines a multi-scale Retinex image enhancement algorithm and an adaptive threshold segmentation method to improve the accuracy of transformer breather discoloration detection and is suitable for different field environments.

[0066] In order to make the above-mentioned purposes, characteristics and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0067] As shown in Figure 1 and Figure 2 The transformer breather discoloration detection method of the present application comprises:

[0068] S1: Obtain an image of a breather to be identified. In this embodiment, a visible light camera with high definition is used to capture the image of the breather to be identified. The image of the breather to be identified is a visible light image.

[0069] S2: Use a multi-scale Retinex image enhancement algorithm to enhance the image of the breather to be identified to obtain an enhanced image. The enhanced image is an RGB color space image.

[0070] According to the Retinex theory, an image is obtained by multiplying the illumination component and the reflection component. When the illumination component can be represented as the convolution of the image and the Gaussian filter function, a single-scale Retinex algorithm can be obtained. Then, the weighted average of the results of multiple Gaussian filter functions with different variances can be obtained, i.e. a multi-scale Retinex algorithm. The multi-scale Retinex image enhancement algorithm is used to enhance the contrast between the colors of the image of the breather to be identified and eliminate the influence of high reflection on the image, thereby retaining the inherent properties of the target device (breather) to achieve the purpose of weakening the cluttered background of the power equipment.

[0071] S3: convert the enhanced image to HSV color space to obtain a converted image. The converted image includes initial H channel data, initial S channel data, and initial V channel data. The initial H channel data represents the hue of the converted image, the initial S channel data represents the saturation of the converted image, and the initial V channel data represents the brightness of the converted image.

[0072] S4: extract a silica gel particle region from the converted image based on a maximum connected region algorithm. The silica gel particle region includes silica gel H channel data, silica gel S channel data, and silica gel V channel data.

[0073] S5: calculate a hue threshold value of each column of pixels in the silica gel H channel data according to the hue of each pixel in the silica gel particle region.

[0074] S6: fit the hue threshold value of each column of pixels to determine a segmentation line.

[0075] S7: divide the silica gel particle region into a discolored sub-region and an un-discolored sub-region using the segmentation line.

[0076] In addition, the transformer breather discoloration detection method further includes: S8: when the color of the discolored sub-region reaches a pre-set color threshold value and the area of the discolored sub-region is greater than a pre-set area threshold value, generating a pre-warning information of replacing the silica gel of the breather.

[0077] The transformer breather discoloration detection method of the present application adopts multi-scale Retinex and adaptive threshold segmentation, enhances the contrast between colors, eliminates the influence of high reflection on images, extracts the silica gel particle region, and finally determines the segmentation line through adaptive hue threshold value independent of prior knowledge, divides the silica gel particle region into a discolored sub-region and an un-discolored sub-region, and can be applied to different field environments to complete the judgment of the discoloration of the breather. The present application solves the problem that the existing computer vision method is easily affected by the environment, and improves the robustness, adaptability, and accuracy of the breather detection.

[0078] Further, S4 specifically includes:

[0079] S41: median filter the initial H channel data to obtain H channel filter information. Median filtering the H primary color channel can effectively eliminate salt and pepper noise.

[0080] S42: determine a binary image according to the H channel filter information, the initial S channel data, and the initial V channel data. The binary image is a black and white image, making subsequent image processing more convenient.

[0081] Specifically, the H channel filter information is binarized to obtain H channel binary data. The initial S channel data is binarized to obtain S channel binary data. The initial V channel data is binarized to obtain V channel binary data. The H channel binary data, S channel binary data and V channel binary data are used to determine a binary image.

[0082] S43: Morphological opening operation is performed on the binary image to obtain a smoothed image.

[0083] The opening operation is first erosion and then dilation. The erosion is that the highlight part in the binary image is eroded, and the domain is reduced. It is represented by replacing the adjacent region with a minimum value (black). In the dilation, the highlight part in the binary image is dilated, and the domain is expanded. It is represented by replacing the adjacent region with a maximum value (white). The effect of the opening operation is to smooth the object profile, remove noise in the binary image, disconnect narrow connected domains to retain larger connected domains, and remove small protruding parts.

[0084] S44: A maximum connected domain algorithm is used to perform maximum connected domain search on the smoothed image to determine an initial silica gel particle region in the smoothed image.

[0085] Specifically, a canny edge detection algorithm is used to detect the connected regions in the smoothed image. The maximum connected domain in the smoothed image is determined according to the size of each connected region; and the maximum connected domain is the initial silica gel particle region. Preferably, the canny edge detection algorithm is used to find each connected region, eliminate larger non-target regions, and search for the final position of the maximum connected domain, i.e. the silica gel particle region. In addition, the cvFindContours function of opencv can also be used instead of the canny edge detection algorithm to detect the connected regions in the smoothed image.

[0086] S45: The initial silica gel particle region is ANDed with the initial H channel filter information to determine a final silica gel particle region.

[0087] Further, the respirator discoloration process to be identified is always vertical or close to vertical. The silica gel particles in the respirator are always in a cylindrical container, and the discoloration plane is parallel or close to parallel to the ground of the cylinder. The discoloration process of the silica gel particles is continuous and derivable. Given a one-dimensional vector The constructor function f is the two-norm square of the difference between the mean of the sub-vector composed of the first i elements of the one-dimensional vector and the mean of the sub-vector composed of the first i-1 elements. The denominator is the two-norm square sum of the means of the adjacent two parts and the means of the remaining parts in the one-dimensional vector The optimal segmentation position is obtained by the maximum inter-class distance and the minimum intra-class distance. The one-dimensional vector The element in the formula is the hue of any column pixel in the H channel data. The numerator represents the distance between classes, and the denominator represents the intra-class distance of the two parts.

[0088] S5 specifically comprises: for any column pixel in the silica gel H channel data, according to the hue of each pixel, classifying each pixel of the column and determining the class of each pixel. The class includes discoloration and non-discoloration. The hue threshold of the column pixel is determined according to the class of each pixel.

[0089] Retinex is a commonly used image enhancement method based on scientific experiments and scientific analysis, which was first proposed by Edwin. H. Land in 1963. Just like Matlab is composed of Matrix and Laboratory, Retinex is also a word composed of two words, retina and cortex, that is: retina and cortex. Land's retinex model is based on the following three assumptions:

[0090] (1) The real world is colorless, and the color perceived by vision is the result of the interaction of light and matter. The water seen is colorless, but the water film-soap film is colorful, which is the result of thin film surface light interference.

[0091] (2) Each color area is composed of red, green and blue three primary colors of a given wavelength.

[0092] (3) The three primary colors determine the color of each unit area.

[0093] The basic theory of Retinex theory is that the color of an object is determined by the object's ability to reflect long-wave (red), medium-wave (green), and short-wave (blue) light, rather than by the absolute value of reflected light intensity. The color of an object is not affected by the non-uniformity of light, and has consistency, that is, retinex is based on color consistency (color constancy). Unlike traditional linear and nonlinear methods that can only enhance certain features of an image, Retinex can achieve a balance in dynamic range compression, edge enhancement and color constancy, so it can adaptively enhance various types of images.

[0094] The multi-scale Retinex is the reflection property of the object itself and the illumination intensity around the object. The illumination intensity determines the dynamic range size of all pixel points in the original image, and the inherent property of the original image is determined by the reflection coefficient of the object itself. The application adopts the idea of Retinex to remove the influence of illumination, and then retains the inherent property of the target device, so as to weaken the cluttered background of the power equipment. Color enhancement, color constancy, local dynamic range compression and global dynamic range compression can be realized while maintaining high fidelity of the image and compressing the dynamic range of the image.

[0095] The application adopts the multi-scale Retinex and adaptive threshold segmentation to detect the discoloration of the respirator, adopts the multi-scale Retinex to pre-process the transformer respirator image data, enhances the contrast of the detected area, suppresses various interference elements caused by uneven illumination, extracts the silica gel particle area, performs color segmentation on the image in the HSV color space through the adaptive hue threshold, weakens the influence of the cluttered background, concentrates the feature points on the target device, and thus improves the matching quality and efficiency of the image, and completes the judgment on the discoloration degree of the respirator.

[0096] As shown in Figure 3 The transformer respirator discoloration detection system of the application includes an image acquisition unit 1, an enhancement unit 2, a conversion unit 3, a region determination unit 4, a hue threshold calculation unit 5, a segmentation line determination unit 6 and a segmentation unit 7.

[0097] The image acquisition unit 1 is used to acquire the image of the respirator to be identified. In the embodiment, the image acquisition unit 1 adopts a visible light camera to shoot the image of the respirator to be identified.

[0098] The enhancement unit 2 is connected with the image acquisition unit 1, and the enhancement unit 2 is used to enhance the image of the respirator to be identified by using the multi-scale Retinex image enhancement algorithm, to obtain an enhanced image.

[0099] The conversion unit 3 is connected with the enhancement unit 2, and the conversion unit 3 is used to convert the enhanced image to the HSV color space, to obtain a converted image. The converted image includes initial H channel data, initial S channel data and initial V channel data.

[0100] The region determination unit 4 is connected with the conversion unit 3, and the region determination unit 4 is used to extract a silica gel particle region from the converted image based on the maximum connected domain algorithm. The silica gel particle region includes silica gel H channel data, silica gel S channel data and silica gel V channel data.

[0101] The hue threshold value calculation unit 5 is connected with the region determination unit 4, and the hue threshold value calculation unit 5 is used for calculating the hue threshold value of each column of pixels in the silica gel H channel data according to the hue of each pixel in the silica gel particle region.

[0102] The split line determination unit 6 is connected with the hue threshold value calculation unit 5, and the split line determination unit 6 is used for fitting the hue threshold value of each column of pixels to determine a split line.

[0103] The split unit 7 is connected with the split line determination unit 6 and the region determination unit 4, and the split unit 7 is used for dividing the silica gel particle region into a color-changing sub-region and a non-color-changing sub-region by using the split line.

[0104] In addition, the transformer breather color change detection system further comprises a warning unit 8. The warning unit 8 is connected with the split unit 7, and the warning unit 8 is used for generating a warning information of replacing the silica gel of the breather when the color of the color-changing sub-region reaches a pre-set color threshold value and the area of the color-changing sub-region is greater than a pre-set area threshold value.

[0105] Further, the region determination unit 4 comprises a filtering module, a binary module, a smoothing module, a connected domain search module and a final region determination module.

[0106] The filtering module is connected with the conversion unit 3, and the filtering module is used for performing median filtering on the initial H channel data to obtain H channel filtering information.

[0107] The binary module is connected with the filtering module and the conversion unit 3, and the binary module is used for determining a binary image according to the H channel filtering information, the initial S channel data and the initial V channel data.

[0108] Specifically, the binary module comprises an H channel binary sub-module, an S channel binary sub-module, a V channel binary sub-module and a binary image determination sub-module.

[0109] The H channel binary sub-module is connected with the filtering module, and the H channel binary sub-module is used for binarizing the H channel filtering information to obtain H channel binary data.

[0110] The S channel binary sub-module is connected with the conversion unit 3, and the S channel binary sub-module is used for binarizing the initial S channel data to obtain S channel binary data.

[0111] The V channel binary sub-module is connected with the conversion unit 3, and the V channel binary sub-module is used for binarizing the initial V channel data to obtain V channel binary data.

[0112] The binary image determination submodule is connected with the H channel binary submodule, the S channel binary submodule and the V channel binary submodule, and is configured to determine a binary image according to the H channel binary data, the S channel binary data and the V channel binary data.

[0113] The smoothing module is connected with the binary module, and is configured to perform a morphological opening operation on the binary image to obtain a smoothed image.

[0114] The connected domain search module is connected with the smoothed image, and is configured to perform a maximum connected domain search on the smoothed image by using a maximum connected domain algorithm to determine an initial silica gel particle region in the smoothed image.

[0115] The final region determination module is connected with the connected domain search module, and is configured to perform an AND operation on the initial silica gel particle region and the initial H channel filter information to determine a final silica gel particle region.

[0116] Compared with the prior art, the transformer breather discoloration detection system has the same beneficial effects as the transformer breather discoloration detection method, and will not be described here again.

[0117] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments, and the same or similar parts of each embodiment can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the related parts can be referred to the method part.

[0118] The principles and implementation manners of the present application are described by using specific examples in the present application, and the above embodiment description is only used to help understand the method of the present application and its core idea; meanwhile, for the general technical personnel in the art, according to the idea of the present application, the specific implementation manner and application range will be changed. In conclusion, the content of the specification should not be understood as the limitation of the present application.

Claims

1. A method for detecting color change in a transformer breather, characterized in that, The method for detecting color change in the transformer breather includes: Acquire an image of the respirator to be identified; The image of the respirator to be identified is enhanced using the multi-scale Retinex image enhancement algorithm to obtain an enhanced image; The enhanced image is converted to the HSV color space to obtain the converted image; Based on the maximum connected component algorithm, silicone particle regions are extracted from the transformed image; the silicone particle regions include silicone H channel data, silicone S channel data, and silicone V channel data. Based on the hue of each pixel in the silicone particle region, calculate the hue threshold of each column of pixels in the silicone H channel data; specifically, for any column of pixels in the silicone H channel data, classify each pixel in the column according to the hue of each pixel to determine the category of each pixel; the category includes color change and no color change; determine the hue threshold of the column of pixels according to the category of each pixel. The tonal threshold of each column of pixels is fitted to determine the dividing line; specifically, a one-dimensional vector is given. Construct a constructor f, with the numerator being a one-dimensional vector. The square of the second normal form of the subvector consisting of the first i elements minus the square of the mean of the subvector consisting of the first i-1 elements, with the denominator being a one-dimensional vector. The two adjacent parts are summed with the sum of the squares of the second normal form of the means of the remaining parts; the optimal split position is obtained by the maximum inter-class distance and the minimum intra-class distance; where, a one-dimensional vector The elements in the numerator represent the hue of any column of pixels in the H channel data; the numerator represents the inter-class distance, and the denominator represents the intra-class distance between the two parts. The silica particle region is divided into a color-changing region and a non-color-changing region using dividing lines.

2. The method for detecting color change in a transformer breather according to claim 1, characterized in that, The transformer breather color change detection method also includes: When the color of the color-changing sub-region reaches a preset color threshold and the area of ​​the color-changing sub-region is greater than a preset area threshold, an early warning message is generated indicating that the silicone of the respirator needs to be replaced.

3. The method for detecting color change in a transformer breather according to claim 1, characterized in that, The acquisition of the image of the respirator to be identified specifically includes: Images of the respirator to be identified are captured using a visible light camera.

4. The method for detecting color change in a transformer breather according to claim 1, characterized in that, The converted image includes initial H channel data, initial S channel data, and initial V channel data; The extraction of silicone grain regions from the transformed image based on the maximum connected component algorithm specifically includes: The initial H-channel data is subjected to median filtering to obtain H-channel filtering information; The binary image is determined based on the H-channel filtering information, the initial S-channel data, and the initial V-channel data; A morphological opening operation is performed on the binary image to obtain a smoothed image; The maximum connected component search algorithm is used to perform maximum connected component search on the smoothed image to determine the initial silicone particle region in the smoothed image; The initial silicone particle region is ANDed with the initial H-channel filter information to determine the final silicone particle region.

5. The method for detecting color change in a transformer breather according to claim 4, characterized in that, The step of determining the binary image based on the H-channel filtering information, the initial S-channel data, and the initial V-channel data specifically includes: The H-channel filtered information is binarized to obtain H-channel binary data; The initial S-channel data is binarized to obtain S-channel binary data; The initial V-channel data is binarized to obtain V-channel binary data; A binary image is determined based on the binary data of the H channel, the binary data of the S channel, and the binary data of the V channel.

6. The method for detecting color change in a transformer breather according to claim 4, characterized in that, The step of performing a maximum connected component search on the smoothed image using a maximum connected component algorithm to determine the initial silicone grain region in the smoothed image specifically includes: The Canny edge detection algorithm is used to detect connected regions in the smoothed image; The largest connected region in the smoothed image is determined based on the size of each connected region; the largest connected region is the initial silicone particle region.

7. A transformer breather color change detection system, characterized in that, The transformer breather color change detection system includes: An image acquisition unit is used to acquire an image of the respirator to be identified. An enhancement unit, connected to the image acquisition unit, is used to enhance the image of the respirator to be identified using a multi-scale Retinex image enhancement algorithm to obtain an enhanced image; A conversion unit, connected to the enhancement unit, is used to convert the enhanced image to the HSV color space to obtain a converted image; A region determination unit, connected to the transformation unit, is used to extract silicone grain regions from the transformed image based on the maximum connected component algorithm; the silicone grain regions include silicone H channel data, silicone S channel data, and silicone V channel data. A tone threshold calculation unit, connected to the region determination unit, is used to calculate the tone threshold of each column of pixels in the silicone H-channel data based on the tone of each pixel in the silicone particle region; specifically, for any column of pixels in the silicone H-channel data, each pixel in the column is classified according to its tone to determine the category of each pixel; the category includes color change and no color change; the tone threshold of the column of pixels is determined according to the category of each pixel. The dividing line determination unit, connected to the tone threshold calculation unit, is used to fit the tone threshold of each column of pixels to determine the dividing line; specifically, a one-dimensional vector is given. Construct a constructor f, with the numerator being a one-dimensional vector. The square of the second normal form of the subvector consisting of the first i elements minus the square of the mean of the subvector consisting of the first i-1 elements, with the denominator being a one-dimensional vector. The two adjacent parts are summed with the sum of the squares of the second normal form of the means of the remaining parts; the optimal split position is obtained by the maximum inter-class distance and the minimum intra-class distance; where, a one-dimensional vector The elements in the numerator represent the hue of any column of pixels in the H channel data; the numerator represents the inter-class distance, and the denominator represents the intra-class distance between the two parts. A segmentation unit, connected to the segmentation line determination unit and the region determination unit, is used to divide the silicone particle region into a color-changing sub-region and a non-color-changing sub-region using a segmentation line.

8. The transformer breather color change detection system according to claim 7, characterized in that, The transformer breather color change detection system also includes: The warning unit, connected to the segmentation unit, is used to generate a warning message to replace the silicone of the respirator when the color of the color-changing sub-region reaches a preset color threshold and the area of ​​the color-changing sub-region is greater than a preset area threshold.

9. The transformer breather color change detection system according to claim 7, characterized in that, The converted image includes initial H channel data, initial S channel data, and initial V channel data; The region determination unit includes: A filtering module, connected to the conversion unit, is used to perform median filtering on the initial H-channel data to obtain H-channel filtered information. The binary module, connected to the filtering module and the conversion unit, is used to determine the binary image based on the H-channel filtering information, the initial S-channel data, and the initial V-channel data. A smoothing module, connected to the binary module, is used to perform morphological opening operations on the binary image to obtain a smoothed image; A connected component search module, connected to the smoothed image, is used to perform a maximum connected component search on the smoothed image using a maximum connected component algorithm to determine the initial silicone particle region in the smoothed image. The final region determination module, connected to the connected component search module, is used to perform an AND operation between the initial silicone particle region and the initial H-channel filter information to determine the final silicone particle region.

10. The transformer breather color change detection system according to claim 9, characterized in that, The binary module includes: The H-channel binary submodule is connected to the filtering module and is used to binarize the H-channel filtered information to obtain H-channel binary data. The S-channel binary submodule is connected to the conversion unit and is used to binarize the initial S-channel data to obtain S-channel binary data. The V-channel binary submodule is connected to the conversion unit and is used to binarize the initial V-channel data to obtain V-channel binary data. The binary image determination submodule is connected to the H-channel binary submodule, the S-channel binary submodule, and the V-channel binary submodule, and is used to determine a binary image based on the H-channel binary data, the S-channel binary data, and the V-channel binary data.

Images