Transformer bushing fault infrared image identification method based on Gaussian filtering algorithm

Abstract

The invention discloses a transformer bushing fault infrared image identification method based on a Gaussian filtering algorithm, and relates to the technical field of power equipment fault identification, and the method comprises the steps: obtaining a transformer bushing infrared image; smoothing the infrared image of the transformer bushing through a Gaussian filtering method; converting the smoothed infrared image of the transformer bushing into a grey-scale map by adopting an average value method; extracting feature vectors from the grey-scale map by using a gray-scale co-occurrence matrix; and inputting the feature vector into a neural network to identify a fault. According to the transformer bushing fault infrared image recognition method based on the Gaussian filtering algorithm, Gaussian filtering is introduced in the transformer bushing fault infrared image recognition process, an original infrared image is preprocessed to prevent sensor noise from affecting the image quality, meanwhile, the effect of smoothing details can be achieved, and the recognition accuracy of the transformer bushing fault infrared image is improved. And the fault of the transformer bushing is identified through the neural network, so that accurate judgment on different fault types is realized.

Description

Technical Field

[0001] The invention relates to the technical field of power equipment fault identification, and in particular to a transformer bushing fault infrared image identification method based on a Gaussian filtering algorithm. Background Art

[0002] In the power system, the transformer is a vital electrical equipment, and the transformer bushing plays a key role as the outlet device for leading the high-voltage wire inside the transformer. The transformer bushing operates in a high-voltage and high-current environment for a long time, and is subjected to strong electricity, heat and mechanical forces. Transformer bushing failures are mainly divided into discharge failures and overheating failures. These failures directly affect the safety of the power grid and may even cause chain accidents, explosions, etc., causing serious harm to equipment and personnel. Therefore, it is necessary to conduct timely safety monitoring to ensure the safe and stable operation of equipment and power grids.

[0003] In the early stage of bushing failure, heating is one of the main characteristics of the fault point, so bushing failure can be detected through infrared technology. Infrared thermal imagers have the advantages of non-contact, high efficiency and high accuracy, and are widely used in power systems. Through infrared imaging technology, temperature anomalies of electrical equipment can be monitored online, and thermal faults can be quickly located and diagnosed through intelligent identification and diagnosis methods.

[0004] High-bitwidth infrared detectors have the characteristics of high precision and high sensitivity, and therefore have become one of the important directions for the development of infrared images. However, high-bitwidth image data needs to be compressed before it can be displayed and processed on ordinary display devices. In order to preserve the details of the image and enhance the display effect, appropriate compression methods need to be adopted. In practical applications, the dynamic range of the original infrared image varies greatly, which poses a challenge to the design of the dynamic range algorithm. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is to solve the problem that the dynamic range of the original infrared image has a large variation range and retain the dynamic details of the original image.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: a transformer bushing fault infrared image recognition method based on Gaussian filtering algorithm, comprising: Acquire infrared images of transformer bushings; The infrared image of transformer bushing is smoothed by Gaussian filtering method; The average value method is used to convert the smoothed infrared image of the transformer bushing into a grayscale image. Use gray-level co-occurrence matrix to extract feature vectors from gray-level images; The feature vector is input into the neural network to identify faults.

[0008] As a preferred solution of the transformer bushing fault infrared image recognition method based on Gaussian filtering algorithm described in the present invention, the transformer bushing infrared image is an infrared image of the transformer bushing taken at any time of the day by an infrared thermal imager.

[0009] As a preferred solution of the infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm described in the present invention, wherein: the Gaussian filtering method is expressed as: , Where x and y are the coordinates of the pixel in the image, and σ is the standard deviation.

[0010] As a preferred solution of the infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm described in the present invention, the average value method is expressed as: , Where M is the grayscale value, R is the pixel value of the R channel of the grayscale image, G is the pixel value of the G channel of the grayscale image, and B is the pixel value of the B channel of the grayscale image.

[0011] As a preferred solution of the infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm described in the present invention, wherein: the gray level co-occurrence matrix is ​​expressed as: , In the formula, GLCM(i,j,d,θ) is the element in the gray-level co-occurrence matrix G, d is the distance in the gray-level image, and θ is the direction in the gray-level image. is the frequency at which pixels with gray level i and pixels with gray level j appear simultaneously.

[0012] As a preferred solution of the infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm described in the present invention, wherein: the feature vector is energy; The energy calculation formula is expressed as, , Where G(i,j) represents the element in the i-th row and j-th column of the gray-level co-occurrence matrix.

[0013] As a preferred solution of the infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm described in the present invention, the neural network refers to a neural network with a data set, and the data set includes grayscale images of different fault types.

[0014] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm are implemented.

[0015] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for infrared image recognition of transformer bushing faults based on a Gaussian filtering algorithm.

[0016] Beneficial effects of the present invention: The infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm provided by the present invention introduces Gaussian filtering in the process of infrared image recognition of transformer bushing fault, pre-processes the original infrared image to avoid the influence of sensor noise on the image quality, and at the same time can play the role of smoothing details, making the image easier to observe and analyze. After being processed by Gaussian filtering, it is converted into a grayscale image to facilitate the subsequent extraction of feature vectors. In the extraction process, the energy features of the image are extracted using the grayscale co-occurrence matrix, and finally the energy features are input into the trained neural network to identify the fault, and finally the rapid and efficient recognition of the transformer bushing fault is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 The present invention provides an overall flow chart of a method for infrared image recognition of transformer bushing faults based on a Gaussian filtering algorithm according to an embodiment of the present invention.

[0019] Figure 2 This is an experimental signal-to-noise ratio comparison diagram of a transformer bushing fault infrared image recognition method based on a Gaussian filtering algorithm provided by an embodiment of the present invention.

[0020] Figure 3 This is a comparison chart of experimental accuracy of a transformer bushing fault infrared image recognition method based on a Gaussian filtering algorithm provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0023] Example 1 Reference Figure 1 , is an embodiment of the present invention, and provides a transformer bushing fault infrared image recognition method based on a Gaussian filtering algorithm, comprising: S1: Obtain an infrared image of the transformer bushing taken by an infrared thermal imager; In this embodiment, the infrared image of the transformer bushing taken by the infrared thermal imager includes an infrared image of the transformer bushing taken by an infrared camera at any time of the day.

[0024] S2: Smoothing the infrared image of the transformer bushing; In this embodiment, the method used for smoothing the infrared image of the transformer bushing is Gaussian filtering, and the specific formula is as follows: , In the formula, x and y are the coordinates of the pixel in the image, and σ is the standard deviation. The row-column separability of the two-dimensional Gaussian function is used for acceleration, that is, a one-dimensional Gaussian filter is first performed on each row, and then the same one-dimensional Gaussian filter is performed on each column of the result, and the mean of the pixels in the filter window is taken as the output. After filtering, the noise in the infrared image of the transformer bushing can be effectively removed and the details of the infrared image of the transformer bushing can be made smoother.

[0025] S3: convert the infrared image of the transformer bushing into a grayscale image; In this embodiment, the transformer bushing infrared image is converted into a grayscale image, and the grayscale value calculation formula adopted is the average value method as follows: , Where M is the grayscale value, R is the pixel value of the R channel of the grayscale image, G is the pixel value of the G channel of the grayscale image, and B is the pixel value of the B channel of the grayscale image. The infrared image is converted to a grayscale image by adding and averaging the pixel values ​​of the red, green, and blue channels. Add the red, green, and blue channel values ​​of each pixel and divide by 3 to get the grayscale value of the pixel. Repeat this operation until the entire image is processed. Converting an image to a grayscale image can simplify the information representation of the image. Compared with the three channels of a color image, it is more concise while reducing computational complexity and retaining important features. S4: Extract feature vector from grayscale image In this embodiment, the method for extracting the feature vector is to use the gray level co-occurrence matrix for extraction, and the feature vector is the energy feature. The gray level matrix generation formula is as follows: , In the formula, GLCM(i,j,d,θ) is the element in the gray-level co-occurrence matrix G, d is the distance in the gray-level image, and θ is the direction in the gray-level image. is the frequency of the simultaneous occurrence of pixels with gray level i and pixels with gray level j, x and y are the coordinates of the pixels in the image, and the gray level co-occurrence matrix can be generated by traversing all the elements in the image. The gray level co-occurrence matrix is ​​used to extract energy features to enhance grayscale image recognition and classification, and to extract texture detail information.

[0026] After obtaining the gray-level co-occurrence matrix, its energy characteristics can be extracted. The energy calculation formula is as follows: , Where G(i,j) represents the element in the i-th row and j-th column of the gray-level co-occurrence matrix, and the energy feature reflects the uniformity of the grayscale distribution and the coarseness of the texture of the image.

[0027] It should be noted that when a fault occurs, it will be accompanied by mechanical damage to the system, which will cause a significant change in the energy characteristics. Therefore, the energy characteristics can well distinguish between the normal state and the fault state. When the energy of the signal is lower than a certain threshold, it can be determined that it is mainly noise. Under conditions with more noise, the accuracy of fault detection can be improved. Moreover, the energy characteristics are simple to calculate, and only square sum is required.

[0028] In many fault scenarios, the texture of the signal will change significantly. The gray-level co-occurrence matrix can reflect the spatial dependency between pixels, thereby revealing the texture features of the image or signal to identify the fault. For example, during normal operation, the texture of the vibration signal of the equipment is relatively stable; when a fault occurs, the signal texture becomes complex and may contain more high-frequency components or change patterns. GLCM extracts texture features and can effectively identify these changes.

[0029] S5: Input the feature vector into the neural network to identify the fault.

[0030] In this embodiment, the input of feature vectors into a neural network to identify faults refers to a convolutional neural network. The data set in the training process includes images of different fault types, and the infrared images are labeled and given corresponding energy feature values ​​through manual annotation. During the training process, the energy features of the input transformer bushing infrared image are provided to the neural network, and a loss function is used to measure the difference between the output result of the neural network and the expected result. The loss function includes the mean square error, and the mean square error calculation formula is as follows: , In the formula, For the The predicted value of samples, For the The actual value of samples, n is the number of samples. Then the parameters of the neural network are updated through the back propagation algorithm to reduce the value of the loss function and optimize the performance of the network. After the neural network training is completed, new samples can be input to achieve fault identification.

[0031] The present invention focuses on preprocessing infrared images through a series of methods such as Gaussian filtering, which is suitable for scenes that require noise reduction and detail smoothing in the image preprocessing stage, especially for scenes with large background noise interference. The traditional method uses a convolutional neural network to directly identify the infrared image of the casing to judge the fault without preprocessing the image, and focuses more on the construction and training of the neural network model.

[0032] Adjustments made to adapt to specific scenarios: The present invention makes the following adjustments: Infrared images processed by Gaussian filtering are used as training data to enhance the recognizability of image features. Gray-level co-occurrence matrix is ​​introduced to extract feature vectors to enhance the texture and detail information of the image. The standard deviation σ in the Gaussian filtering method is adjusted to adapt to images with different noise levels.

[0033] When the original infrared image is noisy and the details are unclear, the accuracy of fault recognition will be affected. Directly identifying the fault without processing will lead to insufficient recognition accuracy. The present invention uses a Gaussian filtering method to preprocess the infrared image and uses a gray-level co-occurrence matrix to extract the features of the image, thereby improving the signal-to-noise ratio of the image and the recognizability of the features, and can also achieve a higher accuracy rate under the condition of large noise interference.

[0034] Since the σ value affects the balance between smoothing effect and image details, a larger σ value will lead to a stronger smoothing effect and cleaner noise removal, but at the same time may blur important features of the image. A smaller σ value has a weaker smoothing effect, but can retain more image details, while the noise removal effect may not be sufficient.

[0035] Therefore, the σ value can be preliminarily set through experiments or expert experience. Use the cross-validation method to evaluate the filtering effect and fault identification accuracy under different σ values ​​and select the optimal σ value. The process written into the steps is as follows Specifically, the infrared image of the transformer bushing is smoothed; In this embodiment, the method used for smoothing the infrared image of the transformer bushing is Gaussian filtering, and the specific formula is as follows: , Where x and y are the coordinates of the pixel points in the image, and σ is the standard deviation. First, we can set the initial value range of σ based on expert experience, such as σ1, σ2, …σn, and prepare a data set of infrared images of transformer bushings with different types of faults and normal states. The data set is annotated so that each image has the correct fault type label. For each σ value, the Gaussian filtering algorithm is used to process the entire data set to obtain a series of smoothed images.

[0036] Cross-validation selects the optimal σ value; Divide the labeled transformer bushing infrared image dataset into 5 parts. For each part of the data: use all the data except this part as the training set to train the neural network model. Use this data as the test set to evaluate the performance of the model and record the accuracy of each part of the data. For each σ value, calculate the average value of the performance index of each data to obtain the average performance of the model under this σ value, compare the average performance indicators under different σ values, and find the σ value that performs best in the fault identification task.

[0037] The method of this embodiment fully considers the dynamic change range of the original infrared image, uses Gaussian filtering to smooth the original infrared image, retains the dynamic details of the original image, and can provide powerful assistance for transformer bushing fault identification.

[0038] Example 2 An embodiment of the present invention is different from the previous embodiment in that: If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0039] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0040] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0041] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0042] Example 3 Reference Figure 2-Figure 3 , which is an embodiment of the present invention, provides a transformer bushing fault infrared image recognition method based on Gaussian filtering algorithm. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0043] like Figure 2 As shown in the figure, this chart shows the signal-to-noise ratio (SNR) performance of the two methods under different Gaussian filter standard deviation values. The horizontal axis represents the standard deviation of the Gaussian filter (σ), and the vertical axis represents the signal-to-noise ratio (dB). The SNR value of the method in this embodiment is generally higher than that of the traditional method, indicating that the method in this embodiment is more effective in noise reduction. In particular, when the standard deviation value of the Gaussian filter increases, the SNR improvement of the method in this embodiment is more significant, indicating that Gaussian filtering can effectively smooth the image and reduce noise, thereby improving the signal-to-noise ratio of the image. A higher SNR means that there are more useful signal components and fewer noise components in the image, which is very beneficial for subsequent feature extraction and fault identification. The method in this embodiment preprocesses the image through Gaussian filtering, retains more image details, removes noise, and improves image quality.

[0044] like Figure 3 As shown, when the standard deviation value of the Gaussian filter increases, the accuracy of the method of this embodiment first increases rapidly and then tends to stabilize. This shows that the accuracy of fault identification can be significantly improved by appropriate Gaussian filtering. The method of this embodiment can more accurately capture the texture and structural information of the image through the feature vector extracted by the gray-level co-occurrence matrix, which is crucial for the training of neural networks and the identification of fault modes. In contrast, the traditional method directly uses the noisy image for feature extraction, resulting in inaccurate features, which affects the performance of fault identification. Robustness: The high accuracy of the method of this embodiment shows that it has better robustness to images with different noise levels. This means that even in the case of large noise interference, the method of this embodiment can maintain a high recognition accuracy, which is particularly important for infrared image recognition of transformer bushing faults in practical applications.

[0045] It should be noted that the above embodiments are only used to illustrate the technical solutions 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 solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A transformer bushing fault infrared image recognition method based on Gaussian filtering algorithm, characterized in that: include: Acquire infrared images of transformer bushings; The infrared image of transformer bushing is smoothed by Gaussian filtering method; The average value method is used to convert the smoothed infrared image of the transformer bushing into a grayscale image. Use gray-level co-occurrence matrix to extract feature vectors from gray-level images; Input the feature vector into the neural network to identify the fault; The gray-level co-occurrence matrix is ​​expressed as, ， In the formula, GLCM(i,j,d,θ) is the element in the gray-level co-occurrence matrix G, d is the distance in the gray-level image, and θ is the direction in the gray-level image. is the frequency at which pixels with gray level i and pixels with gray level j appear simultaneously.

2. The infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm according to claim 1 is characterized in that: The transformer bushing infrared image is an infrared image of the transformer bushing taken at any time of the day by an infrared thermal imager.

3. The infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm as claimed in claim 2 is characterized in that: The Gaussian filtering method is expressed as, ， Where x and y are the coordinates of the pixel in the image, and σ is the standard deviation.

4. The method for identifying transformer bushing fault infrared images based on Gaussian filtering algorithm according to claim 3, characterized in that: The mean value method is expressed as, ， Where M is the grayscale value, R is the pixel value of the R channel of the grayscale image, G is the pixel value of the G channel of the grayscale image, and B is the pixel value of the B channel of the grayscale image.

5. The method for identifying transformer bushing fault infrared images based on Gaussian filtering algorithm according to claim 4, characterized in that: The eigenvector is energy; The energy calculation formula is expressed as, ， Where G(i,j) represents the element in the i-th row and j-th column of the gray-level co-occurrence matrix.

6. The infrared image recognition method for transformer bushing fault based on Gaussian filtering algorithm as claimed in claim 5, characterized in that: The neural network refers to a neural network with a data set, and the data set includes grayscale images of different fault types.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the transformer bushing fault infrared image recognition method based on Gaussian filtering algorithm as described in any one of claims 1 to 6 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for identifying infrared images of transformer bushing faults based on a Gaussian filtering algorithm as described in any one of claims 1 to 6 are implemented.

Images