An infrared thermal imaging detection method for zero-value insulators in distribution lines based on image matching

Abstract

The present invention discloses an infrared thermal image detection method for zero-value insulators of distribution lines based on image matching. First, infrared images of insulators of distribution lines are collected on-site to establish an infrared image library of insulators under typical shooting angles, and the infrared images are preprocessed by grayscale conversion and an improved BM3D denoising algorithm to restore the effective information in the infrared images and reduce the clutter noise. Then, an improved template matching algorithm is used to locate and identify the insulators, the coordinate parameters of the insulators in the infrared images are obtained, and an improved threshold segmentation algorithm is used to segment and extract the insulators. Finally, the least square linear fitting is used to extract the temperature of the insulator images, and the difference in the temperature states among multiple insulators is compared by the same-kind judgment comparison method to realize the detection of zero-value insulators. The present invention can provide a safe and efficient zero-value insulator detection method for operation and maintenance personnel, which helps to realize the intelligent operation and maintenance of insulators of distribution lines.

Description

Technical Field

[0001] The present invention relates to the field of distribution lines, and particularly to an infrared thermal imaging detection method for zero-value insulators of distribution lines based on image matching. Background Art

[0002] As the first level closest to users in the power grid system, the safe and stable operation of distribution lines directly affects the life and production of the general public. Distribution lines are widely distributed and are vulnerable to natural factors such as lightning strikes due to their low insulation level. Lightning strikes on the lines or surrounding buildings are likely to cause accidents of distribution line insulators. Porcelain insulators are commonly used in distribution lines and are an indispensable part of the distribution network. Due to some process defects in the production and manufacturing process, there will be microcracks and small pores inside the porcelain insulators. The porcelain insulators are in long-term live operation and are affected by wire pressure, temperature change, pollution, weather change, strong electric field, etc. When they deteriorate to a certain extent, their mechanical strength and insulation performance will decline, resulting in zero-value insulators. The creepage distance of zero-value insulators decreases compared with normal insulators, which is likely to cause breakdown and flashover phenomena, seriously threatening the safe and stable operation of distribution lines.

[0003] In 2018, multiple lightning strike faults occurred in an important distribution line in the western part of Haining, Zhejiang. When on-site operation and maintenance personnel patrolled the line, it was found that the faults were mainly caused by lightning damage to porcelain insulators. Since it was located near a chemical plant, the surrounding acid-base land had a relatively high corrosion degree on the distribution network, which also accelerated the generation of zero-value insulators. To avoid accidents caused by deteriorated insulators, DL / T393-2010 and DL / T 626-2015 stipulate that regular inspections of porcelain insulators are required. Researchers at home and abroad have proposed various methods such as electric field measurement method, ultraviolet imaging method, and robot detection method to detect deteriorated insulators of transmission and distribution lines. These live detection methods not only have a large workload, poor safety, high detection cost, but also lack the acquisition of real-time information on the external insulation performance of equipment, and cannot make timely and accurate judgments. The infrared thermal imaging method determines whether there are zero-value insulators by comparing the temperature state differences between insulators. Because of its non-contact, safe and efficient advantages, it is a relatively feasible live detection method in the existing technology. Since the voltage level of distribution lines is relatively lower than that of transmission lines and the pole tower height is limited, a large number of infrared images can be taken from the same shooting angle on the ground. The infrared thermal imaging method is more suitable for detecting zero-value insulators of distribution lines. Summary of the Invention

[0004] Aiming at the problems existing in the prior art, the purpose of the present invention is to provide an infrared thermal imaging detection method for zero-value insulators of distribution lines based on image matching, to provide a safe and efficient zero-value insulator detection method for distribution line operation and maintenance personnel, which helps to realize the intelligent operation and maintenance of line insulators and improve the efficiency and safety of detecting deteriorated insulators.

[0005] To achieve the object of the present invention, the present invention adopts the following technical solutions, including the following steps:

[0006] S1: First, collect infrared images of distribution line insulators including N models through on-site shooting to establish an insulator infrared image library; then, perform weighted average graying on each image in the infrared image library, and use an improved BM3D algorithm to denoise the grayed image; finally, crop the insulator regions at the typical shooting angles of the N insulators in the denoised image to establish an insulator infrared image template library;

[0007] S2: Select 1 insulator infrared image from the insulator infrared image template library as the template image T, and select N infrared images of the same model insulators from the infrared image library as the target image set I i (i ∈ 1, 2, 3…, n); construct an insulator recognition model of the Canny-Scale template matching algorithm, locate and recognize the insulators in the target image set, obtain the coordinate parameters of the insulators in the infrared image, and crop the insulator regions according to the coordinate parameters, crop the insulators with the minimum circumscribed rectangle to obtain the cropped image set C i (i ∈ 1, 2, 3…, n);

[0008] S3: Use the Canny-Otsu segmentation algorithm to segment the cropped image set to obtain a binary image set of insulators, where the insulator part is classified as the foreground object and the area outside the insulator is classified as the background; use the weighted average fusion algorithm in pixel-level image fusion to fuse the insulator cropped image and the corresponding insulator binary image respectively to obtain the fused image set F i (i ∈ 1, 2, 3…, n); according to the coordinate parameters of the cropped image set in the original target image set, return the images in the fused image set to the positions of the cropped image set in the original target image set respectively to obtain the insulator segmentation and extraction image S i (i ∈ 1, 2, 3…, n);

[0009] S4: Perform weighted graying on the target image set I i (i ∈ 1, 2, 3…, n) and its temperature width bar, read the gray matrices of the target image set and the temperature width bar, and select 24 groups of gray values and their corresponding temperatures from the gray matrix of the temperature width bar. The gray value is used as the independent variable x, and the temperature is used as the dependent variable y for least squares linear fitting, and obtain the corresponding relationship between the gray value and the insulator temperature in the target image set from the fitting linear expression;

[0010] S5: The target image set I iSubstitute the grayscale values of the insulators in (i ∈ 1, 2, 3…, n) into the fitting linear expression to obtain the temperature of the insulators in the target image set, and realize the detection of zero-value insulators by comparing the temperature state differences among multiple insulators.

[0011] Furthermore, the N typical shooting angles of insulators in S1 include but are not limited to: for P-10 type and PS-15 type pin insulators, they are 30° and 45° below the side, and for suspension insulators such as XP-70 type, they are directly below and 30° below the side.

[0012] Furthermore, the improved BM3D algorithm in S1: For an infrared image of a distribution line insulator, perform preprocessing on the image by grayscale conversion to obtain a grayscale image with the same pixel values in the R, G, and B color channels; use BM3D to perform initial filtering on the R-channel image alone to obtain a grayscale denoised image p, downsample the G-channel image, reduce the G-channel image to 1 / 2 of its original size to obtain a low-scale image d, then perform BM3D filtering on the low-scale image d, and upsample the filtered image to the original image size to obtain a guiding image g. The guiding image g is linearly transformed according to formula (1) to obtain the denoised target image I:

[0013]

[0014] In the formula: G j is the pixel value of the guiding image g, I j is the pixel value of the output image, and the linear coefficients (a k , b k ) are obtained from the following formula:

[0015]

[0016]

[0017]

[0018] In the formula: μ k and σ k 2 are respectively the pixel average value and variance of the small square box ω k centered on pixel k in the guiding image g; n is the number of pixels in ω k ; p j is the pixel value of the low-scale image at point j; p k is the pixel value of the grayscale denoised image p in the small square box ω kThe average pixel value in []. Compared with the original BM3D algorithm, the method of the present invention uses the image after BM3D filtering as the low-scale image d, and then performs a smooth secondary filtering on the low-scale image d, which is more suitable for infrared image denoising, retains the advantages of BM3D filtering, and has better denoising performance and higher peak signal-to-noise ratio.

[0019] Further, in S2, different types of insulator images are used as the template image T to achieve the positioning and recognition of different types of insulators. The template library includes, but is not limited to, P-10 type and PS-15 type pin insulators, and XP-70 type suspension insulators; for insulator images of the same type, the same template image T can simultaneously match M target images.

[0020] Further, for the positioning and recognition of insulators in S2, an insulator recognition model based on the Canny-Scale template matching algorithm is constructed. For a target image I, it is scaled according to the Scale ratio, and the image is scaled in turn according to the ratios of α = 95%, α = 90%, α = 85%, α = 80%, α = 75%, α = 70%, α = 65%, α = 60%, α = 55%, α = 50%, and the aspect ratio is maintained; the Canny operator is used to extract the edge of the template image T to obtain the binary image T b , and the Canny operator is used to extract the edge of the target image I to obtain the binary image I b , and the correlation coefficient matching method is used to perform template matching on the binary image T b and the binary image I b . When the scaling ratio is α = 95%, the binary image T of the template image b first slides on the binary image I of the target image b starting from the upper left corner coordinate origin until the binary image I is traversed b . For each pixel slide, the similarity of the overlapping area between the binary image T b and it is calculated, and the coordinate position of the image when the similarity is the largest is recorded; after the matching of the scaling ratio of α = 95% is completed, the same template matching process as when α = 95% is performed for α = 90%, α = 85%, α = 80%, α = 75%, α = 70%, α = 65%, α = 60%, α = 55%, α = 50% until the target image I is scaled smaller than the template image T, then the scaling is stopped, and the image with the largest similarity obtained therefrom is the result of the Canny-Scale template matching. Finally, the result is marked with the minimum bounding rectangle. Compared with the ordinary template matching algorithm, the algorithm of the present invention uses the Scale ratio for scaling, so that the insulator area can be marked with the minimum bounding rectangle, and the Canny operator is used to extract the edges of the template image and the target image, and the edges are used for template matching, which improves the accuracy of the matching.

[0021] Further, the Canny-Otsu segmentation algorithm in S3: Compared with the threshold segmentation algorithm, the Canny-Otsu segmentation algorithm of the present invention combines the Canny edge detection operator with the Otsu method to achieve more accurate edge segmentation of the insulators recognized by Canny-Scale template matching, and can effectively enhance the contrast of the infrared image and reduce noise. For a cropped image C, first use the Canny edge detection operator for edge detection, calculate the partial derivatives of a certain point x and y axes of the cropped image C, and obtain the gradient value of this point; then perform non-maximum suppression to refine the insulator edge. If the gradient value of the point (x, y) is greater than the gradient values of the adjacent points in the gradient direction, then this point is a candidate edge point; then the candidate edge points are divided into A 1 、A 2 、A 3 a total of 3 categories, A 1 represents non-edge points, A 2 represents points to be determined, A 3 represents edge points; use the Otsu method to calculate the maximum between-class variance, and find the optimal boundary points t 1 、A 2 、A 3 interval of t j and t k ; according to the optimal boundary points t j and t k , determine the high and low thresholds of Canny detection, and segment the image.

[0022] Further, in S4, the temperature width bars of the target image sets I i (i ∈ 1, 2, 3…, n) are all in the same temperature range, that is, when taking infrared images on site, the infrared camera is set to shoot with a fixed temperature range of the temperature width bar; the corresponding relationship between the gray value and the temperature obtained from the fitted linear expression is the corresponding relationship between the gray value and the insulator temperature in the target image sets I i (i ∈ 1, 2, 3…, n).

[0023] Further, in S5, use OpenCV to read the gray value with the highest frequency of occurrence in the insulator, that is, the proportion of the gray value in the insulator gray matrix is higher than 95%, and use this gray value to represent the actual gray value of the insulator in the infrared image.

[0024] Further, in S5, determine zero-value insulators according to the infrared diagnosis application specification standard for live equipment. The heating temperature of zero-value insulators is about 1°C lower than that of normal insulators. By obtaining the temperatures of n insulators, if the insulator with the lowest temperature is more than 1°C lower than the average temperature value of the other insulators, it can be judged as a zero-value insulator.

[0025] Advantages of the present invention:

[0026] The infrared thermal image detection method for zero-value insulators of distribution lines based on image matching provided by the present invention can effectively realize the intelligent detection of zero-value insulators of distribution lines, has high accuracy and good generalization performance, overcomes the problems in the prior art, can provide a safe and efficient zero-value insulator detection method for operation and maintenance personnel, helps to realize the intelligent operation and inspection of insulators of distribution lines, and improves the efficiency and safety of detecting deteriorated insulators. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] FIG Figure 1 is a flowchart of the infrared thermal image detection method for zero-value insulators of distribution lines based on image matching in the present invention.

[0028] FIG Figure 2 is the Canny-Scale template matching algorithm model in the present invention.

[0029] FIG Figure 3 is the actual detection effect diagram of the infrared thermal image detection of insulators of distribution lines in the embodiment of the present invention; (a) original infrared image; (b) target image grayscale denoising diagram I; (c) Canny-Scale template matching to identify insulator image M; (d) cropped image C; (e) insulator segmentation and extraction image S. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] The following further describes the present invention in conjunction with embodiments. It is necessary to point out here that the following embodiments are only used to further illustrate the present invention and cannot be understood as limiting the protection scope of the present invention. Some non-essential improvements and adjustments made by those skilled in the art according to the above-mentioned inventive content still fall within the protection scope of the present invention.

[0031] The following elaborates in detail on the infrared thermal image detection method for zero-value insulators of distribution lines based on image matching. Its flowchart is as Figure 1 shown and includes the following steps:

[0032] S1: Collect 800 infrared images of four types of porcelain insulators for distribution lines, namely P-10, PS-15, SL-15 / 30, and XP-70, respectively, through on-site shooting, and construct an infrared image library of insulators. In this embodiment, the infrared images are taken by an infrared camera model IRI-100C1 produced by Hongxiang Company on the same afternoon. The shooting weather is cloudy, the temperature ranges from 16 to 22 °C, the picture size is 290px × 240px, and the shooting angles are typical shooting angles: for pin insulators, they are 30° and 45° below the side, and for suspension insulators, they are directly below and 30° below the side. The original data set is processed by grayscale conversion, and then the improved BM3D algorithm is used to denoise the data set. The infrared images obtained at typical shooting angles are cropped for the insulator area to obtain a template image T with a size of 30px × 30px, and an infrared image template library of insulators is established.

[0033] In this embodiment, select the infrared image of the P-10 type insulator from the infrared images of 4 types of porcelain insulators for distribution lines of P-10, PS-15, SL-15 / 30, and XP-70 models taken on site as a typical case for analysis. These types of insulators are most widely used in 10kV distribution lines, and selecting them for data set analysis is the most typical. Just replace the template image T with the infrared images of other types of insulators in the template image library to achieve the recognition of the remaining 3 types of insulators.

[0034] S2: Select 1 infrared image of the P-10 type insulator with a size of 30px × 30px from the infrared image template library as the template image T, and select 1 infrared image of the P-10 type insulator with a size of 290px × 240px from the infrared image library as the target image I 1 , construct an insulator recognition model based on the Canny-Scale template matching algorithm, set the lower threshold of the Canny operator to 50 and the upper threshold to 200, and then use the template image T and the target image I 1 as input images. After positioning and recognition by the insulator recognition model, the output image is the insulator image M recognized by the Canny-Scale template matching, as shown in Figure 2 . Obtain the coordinate parameters of the insulator in the target image I 1 , and crop the insulator area according to the coordinate parameters. Crop the insulator with the minimum bounding rectangle to obtain the cropped image C.

[0035] In this embodiment, by combining the template matching algorithm with the Canny algorithm, an insulator recognition model based on the Canny-Scale template matching algorithm is constructed. First, convert the target image I 1 and the template image T into grayscale images and then denoise them. Then, use the Canny operator to extract the template image T and the target image I1 At the edge, template matching is performed using the edge of the template image T. The template image T after edge extraction is slid on the target image I 1 The black rectangular frame is the position of the template image T sliding on the target image I 1 When sliding on, while the target image I 1 The image is scaled according to the Scale ratio while maintaining its aspect ratio. When the scaled image is smaller than the template image, the scaling stops. The similarity between the template image T and the image within the current black rectangular frame is calculated using the correlation coefficient matching method for correlation matching. The image within the black rectangular frame with the highest similarity found from the calculation results is the output image. The principle formula of the correlation coefficient matching method in this embodiment is as follows:

[0036]

[0037]

[0038] In the formula: T represents the template image, I represents the target image, w and h represent the height and width of the template image T, (x, y) is the coordinate of a certain point in the figure relative to the origin of the target image I, (x′, y′) is the coordinate relative to the upper left corner of the sub-image within the black rectangular frame, T'(x', y') represents the reduced gray value of this point in the selected template T, I'(x', y') represents the reduced gray value of this point in the target image I, and the similarity function is calculated as follows:

[0039]

[0040] Among them, the similarity function R(x, y) represents the similarity between the template image T and the sub-image within the black rectangular frame.

[0041] S3: The Canny - Otsu segmentation algorithm is used to segment the cropped image C. The Canny edge detection operator is used for edge detection to calculate the partial derivatives of a certain point x and y axes of the cropped image C, and then the gradient value of this point is obtained; non - maximum suppression is performed to refine the insulator edge. If the gradient value of the point (x, y) is greater than the gradient values of adjacent points in the gradient direction, then this point is a candidate edge point; the candidate edge points are divided into A 1 、A 2 、A 3 a total of 3 categories. A 1 represents a non - edge point, A 2 represents a point to be determined, A 3 represents an edge point; the Otsu method is used to calculate the maximum between - class variance to find the optimal boundary points t 1 、A 2 、A 3 interval of t j and t k ; According to the optimal boundary point tj and t k , determine the high and low thresholds of Canny detection, segment the image, classify the insulator part as the foreground object, classify the area outside the insulator as the background, and obtain the binary image of the insulator; use the weighted average fusion algorithm in pixel-level image fusion to fuse the cropped image C and the binary image to obtain the fused image F; according to the coordinate parameters of the cropped image C in the original target image I 1 , return the fused image F to the position where the cropped image C is located in the original target image I 1 to obtain the insulator segmentation and extraction image S.

[0042] In this embodiment, 4 infrared images of P-10 type insulators are selected to verify the effectiveness of this method. By replacing the target image I 1 with the target image I n (n = 2, 3, 4), the detection of insulators can be realized on multiple target images I, and the actual detection effect is as Figure 3 shown.

[0043] S4: By performing the same weighted grayscale processing on the target image I 1 and its temperature width bar, read the grayscale matrix of the target image I 1 and its temperature width bar. Select 24 groups of grayscale values and their corresponding temperatures from the temperature width bar grayscale matrix. Taking the grayscale value as the independent variable x and the temperature as the dependent variable y, perform least squares linear fitting, and obtain the corresponding relationship between the grayscale value and the insulator temperature in the target image I 1 . The corresponding relationship is shown in Table 1.

[0044] Table 1 Grayscale values and their corresponding temperatures

[0045]

[0046] After linear fitting, the straight line y = 0.1132x - 2.1824 can be obtained as the expression of the corresponding relationship between the grayscale value and the temperature. After error analysis, when x = 112, the y obtained from the fitted straight line is 10.496, which differs from the actual value y = 11 by 0.504, being the maximum error in the fitted data. That is, through least squares linear fitting, the temperature values of each pixel on the infrared image are extracted, with the maximum error within ±0.5 °C and the average error within ±0.20 °C, and its accuracy can meet the requirements of zero-value insulator detection and analysis.

[0047] S5: Substitute the grayscale value of the insulator in the target image I 1 into the fitted linear expression to find the temperature of the insulator in the target image I 1 ; by replacing the target image I 2 , use the same steps to obtain the target image I2 The temperature of the insulator in the middle can realize batch reading of the temperature of the insulators in M infrared images through program design; finally, zero-value insulators are judged according to the infrared diagnosis application specification standard for live equipment. The heating temperature of zero-value insulators is about 1°C lower than that of normal insulators. By obtaining the temperatures of M insulators, if the insulator with the lowest temperature is more than 1°C lower than the average temperature value of the other insulators, it can be judged as a zero-value insulator.

[0048] In this embodiment, the gray value with the highest frequency of occurrence in the cropped image C is read through OpenCV. The proportion of this gray value in the insulator gray matrix is higher than 95%. This gray value represents the actual gray value of the insulator in the infrared image. The insulator temperatures in the 4 infrared images of P-10 type insulators selected in this embodiment are shown in Table 2.

[0049] Table 2 Experimental results of the test data set

[0050]

[0051] As can be seen from Table 2, the actual temperatures of the insulators in the original images are all around 22°C, but the temperature of "Target Image I 4 " is only 21.1368°C, which is significantly lower than the temperatures of the other normal insulators and is 0.9056°C lower than the temperature of "Target Image I 3 ". Therefore, it can be judged that it is suspected to be a zero-value insulator and needs to be replaced by the operation and maintenance personnel. During actual detection, multiple Target Images I n can be input for zero-value insulator detection. The more the number of Target Images I n , the more persuasive the detection result.

[0052] The above only expresses the preferred embodiments of the present invention and does not limit the present invention in other forms. Any person skilled in the art may use the disclosed content to make changes or modifications into equivalent embodiments with equivalent changes. However, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An infrared thermal image detection method for zero-value insulators of distribution lines based on image matching, characterized in that: It includes the following steps: S1: First, collect infrared images of distribution line insulators including N models through on-site shooting to establish an infrared image library of insulators; then, perform weighted average grayscale processing on each image in the infrared image library, and use the improved BM3D algorithm to denoise the grayscale image; finally, crop the insulator regions in the denoised image at the typical shooting angles of N insulators to establish an infrared image template library of insulators; S2: Select 1 insulator infrared image from the insulator infrared image template library as the template image T, and select N infrared images of the same type of insulator from the infrared image library as the target image set I i (i ∈ 1, 2, 3…, n); Construct an insulator recognition model for the Canny-Scale template matching algorithm, locate and recognize the insulators in the target image set, obtain the coordinate parameters of the insulators in the infrared image, and crop the insulator area according to the coordinate parameters. Crop the insulators with the minimum bounding rectangle to obtain the cropped image set C i (i ∈ 1, 2, 3…, n); S3: Use the Canny-Otsu segmentation algorithm to segment the cropped image set to obtain a binary image set of insulators, where the insulator part is classified as the foreground object and the area outside the insulator is classified as the background; use the weighted average fusion algorithm in pixel-level image fusion to fuse the insulator cropped image and the corresponding insulator binary image respectively to obtain the fused image set F i (i ∈ 1, 2, 3…, n); According to the coordinate parameters of the cropped image set in the original target image set, return the images in the fused image set to the positions of the cropped image set in the original target image set respectively to obtain the insulator segmentation and extraction image S i (i ∈ 1, 2, 3…, n); S4: Weight the grayscale of the target image set I i (i ∈ 1, 2, 3…, n) and its temperature-width bar, read the grayscale matrices of the target image set and the temperature-width bar, select 24 groups of grayscale values and their corresponding temperatures from the grayscale matrix of the temperature-width bar, use the grayscale values as the independent variable x and the temperature as the dependent variable y for least squares linear fitting, and obtain the corresponding relationship between the grayscale value and the insulator temperature in the target image set from the fitting linear expression; S5: Substitute the gray values of the insulators in the target image set I i (i ∈ 1, 2, 3…, n) into the fitting linear expression to obtain the temperature of the insulators in the target image set, and realize zero-value insulator detection by comparing the temperature state differences among multiple insulators; The improved BM3D algorithm in S1: For an infrared image of a distribution line insulator, perform preprocessing on the image by grayscaling to obtain a grayscale image with the same pixel values in the R, G, and B color channels; use BM3D to perform initial filtering on the R-channel image alone to obtain a grayscale denoised image p, downsample the G-channel image, reduce the G-channel image to 1 / 2 of its original size to obtain a low-scale image d, then perform BM3D filtering on the low-scale image d, and upsample the filtered image to the original image size to obtain a guiding image g. The guiding image g is linearly transformed according to formula (1) to obtain the denoised target image I: Where: G j is the pixel value of the guiding image g, I j is the pixel value of the output image, and the linear coefficients (a k , b k ) are obtained by the following formula: where: μ k and σ k 2 are respectively the pixel average value and variance of the small square box ω k centered at pixel k in the guiding image g; n is the number of pixels in ω k ; p j is the pixel value of the low-scale image at point j; is the pixel average value of the gray-scale denoised map p in the small square box ω k ; In S2, an insulator recognition model based on the Canny-Scale template matching algorithm is constructed by combining the template matching algorithm with the Canny algorithm; the target image I 1 and the template image T are first converted into grayscale images and then denoised. Next, the Canny operator is used to extract the edges of the template image T and the target image I 1 . The edges of the template image T are used for template matching; the template image T after edge extraction is slid on the target image I 1 . The black rectangular box is the position of the template image T when it slides on the target image I 1 . At the same time, the target image I 1 scales the image according to the Scale ratio and maintains its aspect ratio. When the scaled image is smaller than the template image, the scaling stops; Calculate the similarity between the template image T and the image within the current black rectangular frame according to the correlation coefficient matching method for correlation matching, and the image within the black rectangular frame with the highest similarity found from the calculation results is the output image.

2. An infrared thermal image detection method for zero-value insulators of distribution lines based on image matching as described in claim 1, characterized in that: The typical shooting angles of the N insulators in S1 include: for the pin insulator, it is 30° and 45° below the side, and for the suspension insulator, it is directly below and 30° below the side.

3. An infrared thermal image detection method for zero-value insulators of distribution lines based on image matching as described in claim 1, characterized in that: In S2, different models of insulator images are used as the template image T to achieve the positioning and recognition of different models of insulators. The template library includes P-10 type and PS-15 type pin insulators, and XP-70 type suspension insulators; for insulator images of the same model, the same template image T can simultaneously match M target images.

4. An infrared thermal image detection method for zero-value insulators of distribution lines based on image matching as described in claim 1, characterized in that: For the positioning and recognition of insulators in S2, an insulator recognition model based on the Canny-Scale template matching algorithm is constructed. For a target image I, it is scaled according to the Scale ratio, and the image is scaled in sequence according to the ratios of α = 95%, α = 90%, α = 85%, α = 80%, α = 75%, α = 70%, α = 65%, α = 60%, α = 55%, α = 50%, while maintaining the aspect ratio. The Canny operator is used to extract the edges of the template image T to obtain the binary image T b , and the Canny operator is used to extract the edges of the target image I to obtain the binary image I b . The correlation coefficient matching method is used to perform template matching on the binary image T b and the binary image I b . When the scaling ratio is α = 95%, the binary image T of the template image b first slides and translates on the binary image I of the target image b starting from the upper left corner coordinate origin until the binary image I b is traversed. For each pixel slide, the similarity of the overlapping area between the binary image T b and it is calculated, and the coordinate position of the image when the similarity is the largest is recorded. After the template matching process with a scaling ratio of α = 95% is completed, the same template matching process as when α = 95% is performed for α = 90%, α = 85%, α = 80%, α = 75%, α = 70%, α = 65%, α = 60%, α = 55%, α = 50% until the target image I is scaled smaller than the template image T, then the scaling stops. The image with the maximum similarity obtained therefrom is the result of the Canny-Scale template matching. Finally, the result is marked with the minimum bounding rectangle.

5. An infrared thermal image detection method for zero-value insulators of distribution lines based on image matching as described in claim 1, characterized in that: Canny-Otsu Segmentation Algorithm in S3: The Canny-Otsu segmentation algorithm combines the Canny edge detection operator with the Otsu method to achieve more accurate edge segmentation of the insulators recognized by Canny-Scale template matching, which can effectively enhance the contrast of infrared images and reduce noise. For a cropped image C, first use the Canny edge detection operator for edge detection, calculate the partial derivatives of a certain point in the x and y axes of the cropped image C, and obtain the gradient value of this point. Then perform non-maximum suppression to refine the insulator edge. If the gradient value of the point (x, y) is greater than the gradient values of the adjacent points in the gradient direction, then this point is a candidate edge point. Then the candidate edge points are divided into A 1 、A 2 、A 3 with a total of 3 categories. A 1 represents non-edge points, A 2 represents points to be determined, A 3 represents edge points; use the Otsu method to calculate the maximum between-class variance and find the optimal boundary points t 1 、A 2 、A 3 in the interval. j and t k ; According to the optimal boundary points t j and t k , determine the high and low thresholds for Canny detection and segment the image.

6. An infrared thermal image detection method for zero-value insulators of distribution lines based on image matching as described in claim 1, characterized in that: Target image set I in S4 i (i ∈ 1, 2, 3…, n), the temperature width bars are all within the same temperature range, that is, when taking on-site infrared images, the infrared camera is set to take pictures within a fixed temperature range of the temperature width bars; the corresponding relationship between the gray value and the temperature obtained in the fitting linear expression is the target image set I i (i ∈ 1, 2, 3…, n), the corresponding relationship between the gray value and the insulator temperature 7. An infrared thermal image detection method for zero-value insulators of distribution lines based on image matching as described in claim 1, characterized in that: In S5, use OpenCV to read the grayscale value with the highest frequency of occurrence in the insulator, that is, the proportion of the grayscale value in the insulator grayscale matrix is higher than 95%, and use this grayscale value to represent the actual grayscale value of the insulator in the infrared image.

8. A method for infrared thermal imaging detection of zero-value insulators in distribution lines based on image matching as described in claim 1, characterized in that: In S5, the determination of zero-value insulators is carried out according to the infrared diagnosis application specification standard for live equipment. The heating temperature of zero-value insulators is about 1°C lower than that of normal insulators. By obtaining the temperatures of n insulators, if the insulator with the lowest temperature is more than 1°C lower than the average temperature value of the remaining insulators, it can be determined as a zero-value insulator.

Images