Infrared image composite insulator heating defect identification method and system under complex background

Abstract

The invention discloses an infrared image composite insulator heating defect identification method and system under a complex background. The method comprises the following steps: S1, obtaining a composite insulator infrared image; s2, performing image segmentation on the infrared image of the composite insulator to obtain a composite insulator rod core area mask; s3, performing temperature decoding on the infrared image source data of the composite insulator to obtain a temperature data matrix corresponding to each pixel; s4, extracting corresponding temperature data from the temperature data matrix based on the composite insulator rod core area mask to obtain a rod core mask matrix with temperature; and S5, carrying out temperature rise difference analysis on the rod core based on the rod core mask matrix, and carrying out heating abnormity judgment and temperature rise defect range determination based on the temperature rise difference analysis. According to the scheme, the reference value is determined through the mean value by utilizing the characteristics that the temperature distribution of the rod core is relatively uniform under the normal condition and the temperature distribution is not uniform under the abnormal condition, so that the influence of the environment on temperature rise analysis can be avoided, the influence of individual abnormal data can be overcome, and the judgment precision is ensured.

Description

Technical Field

[0001] The invention belongs to the technical field of power system transmission line monitoring, and in particular relates to a method and system for identifying heating defects of composite insulators using infrared images under complex backgrounds. Background Art

[0002] Composite insulators are special insulating components used in overhead transmission lines. The insulating parts of composite insulators are usually composed of core rods, sheaths and sheds made of organic materials. The core rods are usually composed of glass fiber and resin. Compared with traditional porcelain insulators and glass insulators, composite insulators have many advantages, such as high strength, light weight, and excellent anti-pollution flashover performance, and have become an indispensable part of the power system.

[0003] With the continuous expansion of the scale of power grids and the improvement of the requirements for the safety of power system operation, the performance and stability of composite insulators have become an important concern of the power industry. With the increase of service life, the aging problem of composite insulators has become more and more prominent year by year. When its external materials are exposed to the natural environment for a long time, they are easily affected by external factors such as ultraviolet rays, temperature changes, moisture, pollution, etc., which can lead to performance degradation and even string breakage. Once a string breakage accident occurs, it may cause a serious power outage, resulting in huge economic losses and social impact.

[0004] In order to effectively evaluate and monitor the aging status of composite insulators, infrared analysis technology has gradually been widely used as a non-contact, efficient detection method. By detecting the temperature distribution on the surface of composite insulators, infrared analysis can promptly detect potential internal defects and aging problems. However, in complex background environments, there will be a large number of interference factors in infrared photos, such as mountains, buildings, meteorological conditions, etc., which puts higher requirements on the accuracy of infrared detection.

[0005] At present, there are usually two ways to identify heating defects in composite insulators in response to complex background problems:

[0006] 1) Obtain infrared images and use machine learning or deep learning models to identify defects based on infrared images;

[0007] 2) Obtain an infrared image, obtain temperature characteristics based on infrared characteristics to roughly obtain the temperature value of the rod core, and make a defect judgment based on the difference between the rod core temperature value and a preset value.

[0008] For the first method, a recognition model needs to be trained in advance, which is difficult to implement, and the recognition accuracy depends on the model itself and the model training effect, and the stability is poor; although the second method is simpler and easier to implement than the first method, its method of making simple judgments based on preset values ​​has insufficient recognition accuracy and is very prone to misjudgment. Summary of the invention

[0009] The purpose of the present invention is to provide a method for identifying heating defects of composite insulators in infrared images under complex backgrounds, which has high accuracy, is not affected by complex backgrounds, is simple to implement, and can provide effective technical support for safety assessment and monitoring of composite insulators in complex geographical areas.

[0010] Another object of the present invention is to provide a composite insulator heating defect recognition system using infrared images under complex backgrounds.

[0011] In order to achieve the above object, the present invention adopts the following technical solutions:

[0012] A method for identifying heating defects of composite insulators using infrared images under complex backgrounds, the method comprising:

[0013] S1. Acquire infrared images of composite insulators;

[0014] S2. Perform image segmentation on the composite insulator infrared image to obtain a composite insulator rod core region mask;

[0015] S3. Temperature decoding of the composite insulator infrared image source data is performed to obtain a temperature data matrix corresponding to each pixel;

[0016] S4. Extracting the corresponding temperature data from the temperature data matrix based on the composite insulator rod core region mask to obtain a rod core mask matrix with temperature;

[0017] S5. Performing temperature rise difference analysis on the rod core based on the rod core mask matrix, and determining heating anomaly and temperature rise defect range based on the temperature rise difference analysis.

[0018] In the above-mentioned infrared image composite insulator heating defect recognition method under complex background, step S5 includes:

[0019] S51. Obtaining the temperature value of each pixel in the rod core region based on the rod core mask matrix with temperature;

[0020] S52. Take the average value of all temperature values ​​in the rod core area as the reference temperature;

[0021] S53. Calculate the temperature rise value ΔT(x,y) of each pixel in the rod core area based on the reference temperature to obtain a temperature rise matrix:

[0022] ΔT(x,y)=|T(x,y)-T base |, T(x,y) represents the temperature value of the corresponding pixel, T base is the reference temperature;

[0023] S54. Determine whether the rod core has abnormal heating and determine the temperature rise defect range based on the temperature rise matrix.

[0024] In the above-mentioned infrared image composite insulator heating defect recognition method under complex background, when the temperature rise matrix meets the set conditions, it is judged that the rod core has heating anomaly and the temperature rise defect range is further determined.

[0025] In the above-mentioned method for identifying heating defects of composite insulators using infrared images under complex backgrounds, a temperature rise threshold is included. When there are at least a set number of temperature rise values ​​in the temperature rise matrix that are greater than the temperature rise threshold, the temperature rise defect range is determined.

[0026] In the above-mentioned infrared image composite insulator heating defect recognition method under complex background, the temperature rise defect range is determined by the following method:

[0027] S541. Select all pixels whose temperature rise values ​​are greater than the temperature rise threshold;

[0028] S542. Taking the selected pixel as the center, check several pixels around the center point. If (ΔP0-ΔP i )<ε, the corresponding pixel is included in the heating defect area, ΔP0 represents the temperature rise value of the center pixel, ΔP i represents the temperature rise value of the pixels around the target, and ε is the temperature rise difference threshold;

[0029] S543. All surrounding pixels that meet the conditions are included in the heating defect area to obtain the heating defect area corresponding to the corresponding center point;

[0030] S544. Merge the heating defect areas of all center points to obtain the final temperature rise defect range.

[0031] In the above-mentioned infrared image composite insulator heating defect recognition method under complex background, when there are at least a set number of temperature rise values ​​in the temperature rise matrix that are greater than the temperature rise threshold, it is preliminarily determined that there is heating abnormality;

[0032] If the final temperature rise defect range is greater than the set value, it is confirmed that there is abnormal heating;

[0033] Otherwise, continue to perform the detection of steps S1-S5, and the detection speed is greater than the thermal conduction speed of the unit pixel, and obtain the temperature rise defect range of each detection. If the temperature rise defect range shows a trend of gradually expanding over time, or the final temperature rise defect range of at least one detection during the detection period is greater than the set range, it is confirmed that there is a heating abnormality.

[0034] In the above-mentioned method for identifying heating defects of composite insulators in infrared images under complex backgrounds, in step S2, the Deeplabv3plus model is used to perform image segmentation on the composite insulator infrared image to obtain a composite insulator rod core area mask;

[0035] Step S2 specifically includes:

[0036] S21. Normalizing the infrared image of the composite insulator;

[0037] S22. The Deeplabv3plus model extracts features from the image processed by S21 to obtain a multi-scale feature map;

[0038] S23. Classify each pixel point in the multi-scale feature map, and output a probability matrix S(x, y) that each pixel belongs to the composite insulator rod core region, where (x, y) represents the pixel coordinates in the image, and the probability matrix represents the probability value of each pixel point belonging to the rod core region;

[0039] S24. Filter out pixel points whose probability values ​​are greater than or equal to the probability threshold according to the probability matrix S(x,y), and mark the corresponding pixel points as the rod core area, and obtain a binary mask matrix about the rod core area, one value represents the rod core area, and the other value represents the non-rod core area, thereby obtaining the composite insulator rod core area mask.

[0040] In the above-mentioned infrared image composite insulator heating defect recognition method under complex background, step S3 specifically includes:

[0041] S31. Decode the additional temperature information in the infrared image and extract the temperature data corresponding to each pixel in the image;

[0042] S32. Perform outlier processing on temperature data:

[0043] Calculate the standard deviation distance between each temperature value and the mean of the temperature set;

[0044] If the temperature value of a certain pixel is more than the outlier threshold from the average value, the pixel is corrected according to the temperature values ​​of the neighboring pixels;

[0045] S33. Perform a normal transformation on the temperature data to make it close to a normal distribution.

[0046] In the above-mentioned infrared image composite insulator heating defect recognition method under complex background, in step S4, when extracting the temperature data of the composite insulator rod core area, all temperature data corresponding to the rod core area are extracted from the temperature data matrix using the rod core area mask matrix obtained in step S2:

[0047]

[0048] M core ′(x,y) represents the binary pickled matrix, 1 indicates that the corresponding pixel is in the rod core area, and 0 indicates that the corresponding pixel is not in the rod core area.

[0049] A composite insulator heating defect recognition system with infrared images under complex backgrounds, the system comprising a drone, the drone having an infrared lens and a control module;

[0050] The infrared lens is used to capture the infrared image of the composite insulator and transmit the infrared image of the composite insulator to the control module;

[0051] The control module is used to execute the method described in any one of claims 1 to 9;

[0052] And the control module includes a wireless module for transmitting the defect identification result to the ground background.

[0053] The advantages of the present invention are:

[0054] 1. This scheme identifies heating defects of composite insulators based on rod core mask and temperature source data analysis. The temperature rise difference analysis is performed through the temperature mask matrix, which can accurately obtain the temperature distribution information of the composite insulator. In this process, the interference of complex background is overcome. On the one hand, the ability to identify heating defects is improved, and on the other hand, the versatility of infrared defect identification of composite insulators in non-sky background environments such as mountainous areas and cities can be improved;

[0055] 2. Propose a flexible confirmation benchmark value, and obtain the temperature rise matrix based on the flexible confirmation benchmark value to perform temperature rise analysis, which can effectively avoid the influence of ambient temperature and humidity on temperature rise analysis;

[0056] 3. Taking advantage of the fact that the temperature distribution of the rod core is relatively uniform under normal conditions and uneven under abnormal conditions, the reference value is confirmed by the mean value. On the one hand, it can avoid the influence of the environment on the temperature rise analysis, and on the other hand, it can overcome the influence of individual abnormal data and ensure the judgment accuracy;

[0057] 4. This solution uses the temperature rise threshold to filter out high-temperature pixel points, and uses the high-temperature pixel points as seed points to determine the temperature rise defect range through the regional growing algorithm, which can more accurately determine the temperature rise defect range;

[0058] 5. This solution uses the size of the determined temperature rise defect range to reversely verify the heating anomaly. Only when the defect range is larger than the set range, or the defect range shows a trend of gradual expansion, is it confirmed that there is a heating anomaly problem, which can effectively improve the recognition accuracy and reduce the misrecognition rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1This is a module structure diagram of the infrared image composite insulator heating defect recognition system under complex background in the first embodiment of the present invention;

[0060] Figure 2 This is a method flow chart of a method for identifying heating defects of composite insulators using infrared images under complex backgrounds in Embodiment 1 of the present invention;

[0061] Figure 3 It is a flow chart of a method for performing image segmentation on an infrared image of a composite insulator to obtain a composite insulator rod core region mask in the first embodiment of the present invention;

[0062] Figure 4 It is a flow chart of a method for obtaining a temperature data matrix from infrared image source data of a composite insulator in the first embodiment of the present invention;

[0063] Figure 5 is a flow chart of a method for temperature rise difference analysis in Embodiment 1 of the present invention;

[0064] Figure 6 is a flow chart of a method for determining a temperature rise defect range in Embodiment 1 of the present invention;

[0065] Figure 7 It is a flow chart of the method for temperature rise difference analysis in the second embodiment of the present invention.

[0066] Figure numerals: drone 1, infrared lens 2, control module 3, wireless module 4, ground background 5. DETAILED DESCRIPTION

[0067] Embodiment 1

[0068] like Figure 1 As shown, the present invention provides a composite insulator heating defect recognition system based on infrared images under complex backgrounds, and the system includes a drone 1 and a ground background 5. The drone 1 is equipped with an infrared lens 2, a control module 3, and a wireless module 4. The drone 1 flies to a target area according to the control of the control module 3. The infrared lens 2 takes an infrared image of a target composite insulator according to the control of the control module 3 after the drone 1 flies to the target area and transmits the infrared image of the composite insulator to the control module 3. The control module 3 processes the infrared image to segment the composite insulator, and then recognizes the heating defect thereof, as shown in FIG. Figure 2 As shown, the specific implementation method is as follows:

[0069] S1. Obtain infrared images of composite insulators, using airborne infrared lens 2, flying in the transmission line area to obtain infrared images of composite insulators in real time;

[0070] S2. Perform image segmentation on the composite insulator infrared image to obtain a composite insulator rod core region mask;

[0071] S3. Temperature decoding of the composite insulator infrared image source data is performed to obtain a temperature data matrix corresponding to each pixel;

[0072] S4. Extracting the corresponding temperature data from the temperature data matrix based on the composite insulator rod core region mask to obtain a rod core mask matrix with temperature;

[0073] S5. Perform temperature rise difference analysis on the rod core based on the rod core mask matrix with temperature, and determine heating anomaly and temperature rise defect range based on the temperature rise difference analysis.

[0074] like Figure 3 As shown, step S2 specifically includes:

[0075] S21. The dimension of the infrared image is (H, W, C), where H is the height of the image, w is the width, and C is the number of channels. The infrared image of the composite insulator is normalized and the pixel values ​​are scaled to the interval [0, 1] to obtain the preprocessed image I.

[0076] S22. Input the preprocessed image I into the Deeplabv3plus model, which is an open source optimized version of deeplabv3, and use the convolutional neural network F of the model for feature extraction. In this process, the model extracts multi-scale features in the image through convolution and pooling operations to generate a feature map F(I). The feature map is expanded using the atrous convolution and spatial pyramid pooling (ASPP) module to generate a multi-scale feature map F ASPP (I), this step can enhance the recognition ability of the rod core part of composite insulators of different scales. The hole convolution can expand the receptive field through different hole rates, thereby capturing more feature information at multiple scales. Its expression is:

[0077]

[0078] Where r is the dilation rate, and N represents the number of dilated convolution operations of different scales.

[0079] S23. Multi-scale feature map F ASPP (I) classifies each pixel point and outputs a probability matrix S(x,y), where (x,y) represents the pixel coordinates in the image and the probability matrix represents the probability value of each pixel point belonging to the rod core area;

[0080] S24. Filter out pixels whose probability values ​​are greater than or equal to the probability threshold τ according to the probability matrix S(x,y), and mark the corresponding pixels as the rod core area, and obtain the binary mask matrix about the rod core area, that is, when S(x,y)≥τ, mark the corresponding pixels as the rod core area, and generate the binary mask matrix M core(x, y) represents the rod core area of ​​the composite insulator in the image, where the pixels in the non-rod core part will be marked as 0, and the pixels in the rod core part will be marked as 1.

[0081] Suppose there is an infrared image, the pixel matrix is ​​as follows:

[0082]

[0083] After segmentation, the resulting probability matrix is ​​as follows:

[0084]

[0085] If the probability threshold is 0.6, the following rod core mask matrix can be obtained:

[0086]

[0087] After the mask is generated, morphological processing operations are further applied to optimize the mask. Through opening and closing operations, the edges are smoothed and small noise areas are removed to obtain the optimized mask matrix M. core ′, noise points and isolated pixels can be further removed.

[0088] Through the above method, composite insulators can be accurately segmented from the infrared images with complex backgrounds, such as segmenting composite insulators from the backgrounds of the sky, complex mountainous areas, and densely populated factory buildings, thereby improving the accuracy of subsequent temperature rise analysis.

[0089] In this solution, only the Deeplabv3plus model used needs to be fine-tuned and trained. The model training effect is guaranteed to a certain extent. Moreover, the model is mainly used for the segmentation of insulators, and the recognition is realized by the subsequent process, which has little impact on the recognition accuracy. Therefore, it can overcome the problems of implementation difficulties and poor stability.

[0090] Furthermore, if Figure 4 As shown, step S3 specifically includes:

[0091] S31. Decode the temperature additional information in the infrared image by using software such as TSDK, extract the temperature data corresponding to each pixel in the image, and obtain the temperature data matrix T(x, y), where (x, y) represents the pixel coordinates in the image;

[0092] S32. Then, outlier processing is performed on the temperature data:

[0093] Calculate the standard deviation distance between each temperature value and the mean of the temperature dataset:

[0094]

[0095] μ represents the mean of the temperature data, and σ represents the standard deviation of the temperature data. If the Z(x,y) of a certain pixel exceeds the outlier threshold, the pixel is corrected according to the temperature values ​​of the neighboring pixels. The correction method can be to take the average value of several neighboring pixels, which can be determined by the technician according to the actual situation and is not limited here.

[0096] S33. Perform normal transformation on the temperature data to make it close to normal distribution. Specifically, Box-Cox transformation is used to transform the temperature data. The transformation expression is:

[0097]

[0098] Among them, λ is the transformation parameter. The appropriate λ value is selected by the maximum likelihood estimation method to make the transformed data close to the normal distribution. In this way, the temperature data matrix T_transformed(x,y) has better distribution characteristics, eliminating the influence of extreme values ​​on subsequent analysis, and providing high-quality input data for subsequent analysis and processing with the average value as the benchmark value, which can more accurately analyze the temperature distribution in the infrared image of the composite insulator.

[0099] The temperature data matrix corresponds to the pixel matrix of the infrared image. Assume that the temperature matrix after infrared image decoding and outlier and transformation processing is as follows:

[0100]

[0101] In this matrix, each tuple represents the temperature of a pixel position, which corresponds to the pixel matrix one by one. Combined with the rod core mask matrix, a pixel coordinate matrix with temperature information can be created, which only contains the pixel positions and corresponding temperature values ​​of the rod core area. This matrix is ​​the rod core mask matrix with temperature. For details, see step S4:

[0102] When extracting the temperature data of the composite insulator rod core region, all temperature data corresponding to the rod core region are extracted from the temperature data matrix using the rod core region mask matrix obtained in step S2:

[0103]

[0104] M ccore ′(x,y) represents the binary pickled matrix, 1 means the corresponding pixel is in the rod core area, 0 means the corresponding pixel is not in the rod core area. This method can be used to extract all the temperature data T of the composite insulator rod core area. core (x,y), get the rod core mask matrix with temperature:

[0105]

[0106] "-" means that the pixel at this position does not belong to the rod core area, and the corresponding value in the mask matrix is ​​0, and these positions are not listed in the pixel coordinate matrix. In this way, the pixels and temperature values ​​of the rod core area can be screened out through the rod core mask matrix and the temperature data matrix, and the positions and temperature data of these pixels can be used for further analysis.

[0107] like Figure 5 As shown, step S5 includes:

[0108] S51. Obtaining the temperature value of each pixel in the rod core region based on the rod core mask matrix with temperature;

[0109] S52. Take the average value of all temperature values ​​in the rod core area as the reference temperature;

[0110] S53. Calculate the temperature rise value ΔT(x,y) of each pixel in the rod core area based on the reference temperature to obtain a temperature rise matrix:

[0111] ΔT(x,y)=|T(x,y)-T base |, T(x,y) represents the temperature value of the corresponding pixel, T base is the reference temperature;

[0112] S54. Determine whether the rod core has abnormal heating and determine the temperature rise defect range based on the temperature rise matrix.

[0113] Using the internal temperature rise of the rod core to determine whether the insulator has a heating defect, rather than comparing it with a specific temperature value, can effectively avoid the influence of ambient temperature and humidity on the temperature rise analysis. Moreover, performing temperature rise analysis in the form of average values ​​can overcome the influence of individual abnormal data on the analysis results.

[0114] Specifically, when the temperature rise matrix meets the set conditions, it is judged that the rod core has abnormal heating and the temperature rise defect range is further determined. Specifically, when there are at least a set number of temperature rise values ​​greater than the temperature rise threshold in the temperature rise matrix, it is judged that there is abnormal heating. The set number can be 1, or other values. When it is 1, it means that as long as there is a temperature rise value greater than the temperature rise threshold, it is considered that there is abnormal heating. When it is 3, it means that there must be at least 3 temperature rise values ​​greater than the temperature rise threshold before it is judged that there is abnormal heating, and the temperature rise range is further confirmed.

[0115] Specifically, Figure 6 As shown, the temperature rise defect range is determined by the following method:

[0116] S541. Select all pixels whose temperature rise values ​​are greater than the temperature rise threshold;

[0117] S542. Taking the selected pixel point P0 as the center, check several pixels P around the center point P0. i, if (ΔP0-ΔP i )<ε, the corresponding pixel is included in the heating defect area, ΔP0 represents the temperature rise value of the center pixel, ΔP i represents the temperature rise value of the pixels around the target, and ε is the temperature rise difference threshold;

[0118] S543. All surrounding pixels that meet the conditions are included in the heating defect area to obtain the heating defect area corresponding to the corresponding center point;

[0119] S544. Merge the heating defect areas of all center points to obtain the final temperature rise defect range.

[0120] When there is only one center point, no merging is required. When there are multiple center points, the heating defect area of ​​each center point needs to be determined separately, and the heating defect areas of all center points need to be merged into the final defect range. When a point has been included in the heating defect area of ​​a certain center point, it is no longer necessary to repeatedly determine whether it meets (ΔP0-ΔP i )<ε to avoid repeated counting.

[0121] Furthermore, after identifying the temperature rise defect range, the defect level is evaluated according to the temperature rise intensity and the defect area in the following way:

[0122]

[0123] Among them, α and β are adjustment parameters, A defec t is the area of ​​the defect region, Represents the average temperature rise intensity in the area. This formula combines the temperature rise intensity with the area to comprehensively evaluate the severity of the defect. Finally, based on the comprehensive results of the temperature rise intensity and the defect area, the heating defects are divided into different levels such as slight, moderate, and severe, providing a basis for maintenance decisions and ensuring the reliability of composite insulators and the safe operation of the power system.

[0124] The above analysis results of the control module 3, including the judgment result of whether there is a heating abnormality, and if there is a heating abnormality, also include the determined temperature rise defect range and temperature rise defect level, are transmitted to the ground background 5 through the wireless module 4. The ground background 5 has a corresponding application, which automatically generates an analysis report based on the analysis results, and outputs the analysis results and analysis report to the human-computer interaction screen to complete the process of providing the analysis results to the user.

[0125] Embodiment 2

[0126] like Figure 7As shown, this embodiment is similar to the first embodiment, except that, in this embodiment, when there are at least a set number of temperature rise values ​​in the temperature rise matrix that are greater than the temperature rise threshold, it is preliminarily determined that there is a heating abnormality;

[0127] If the final temperature rise defect range is greater than the set value, it is confirmed that there is a heating anomaly, and the analysis result is transmitted to the ground background 5;

[0128] Otherwise, continue to perform the detection of steps S1-S5, and the detection speed is greater than the thermal conduction speed of the unit pixel, and obtain the temperature rise defect range of each detection. If the temperature rise defect range shows a trend of gradually expanding over time, or the final temperature rise defect range of at least one detection during the detection period is greater than the set range, it is confirmed that there is a heating abnormality.

[0129] The set value is determined by the technicians based on factors such as the size of the insulator and is not limited here.

[0130] In practical applications, it is impossible to uniquely determine the heat conduction speed of a unit pixel. Therefore, in this embodiment, the technicians preset a heat conduction speed for the insulator according to factors such as the insulator material, the environment in which the insulator is located, the shooting distance, and the resolution of the infrared lens. In the process of shooting, the composite insulator is shot with an infrared lens that meets the resolution requirements at a certain shooting distance from the insulator to make the preset heat conduction speed of the unit pixel close to the actual heat conduction speed. In order to ensure that the detection speed is less than the actual heat conduction speed of the unit pixel, the insulator can be shot at 1 / 2 of the preset heat conduction speed or even a smaller shooting speed to obtain an infrared image for temperature rise analysis and heating defect identification.

[0131] Specifically, the temperature rise defect range of each detection is obtained. If the temperature rise defect range shows a trend of gradually expanding over time, it can be confirmed that abnormal heating exists and is in a state of continuous temperature rise, indicating that the abnormal heating problem of the insulator may be in the development stage, and the heating center is still in the process of conducting heat to the outside.

[0132] Alternatively, if the final temperature rise defect range detected at least once during the detection period is greater than the set range, the presence of heating abnormality can also be confirmed.

[0133] In this way, the size of the determined temperature rise defect range can be used to reversely verify the heating anomaly. Only when the defect range is larger than the set range, or the defect range shows a trend of gradual expansion, is it confirmed that there is a heating anomaly problem, which can effectively improve the recognition accuracy and reduce the misrecognition rate.

[0134] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

[0135] Although the terms such as drone 1, infrared lens 2, control module 3, wireless module 4, and ground background 5 are frequently used in this article, the possibility of using other terms is not excluded. These terms are used only to more conveniently describe and explain the essence of the present invention; interpreting them as any additional limitation is contrary to the spirit of the present invention.

Claims

1. A method for identifying heating defects of composite insulators using infrared images under complex backgrounds, characterized in that: The method includes: S1. Acquire infrared images of composite insulators; S2. Perform image segmentation on the composite insulator infrared image to obtain a composite insulator rod core area mask; S3. Temperature decoding of the composite insulator infrared image source data is performed to obtain a temperature data matrix corresponding to each pixel; S4. Extracting the corresponding temperature data from the temperature data matrix based on the composite insulator rod core region mask to obtain a rod core mask matrix with temperature; S5. Performing temperature rise difference analysis on the rod core based on the rod core mask matrix, and determining heating anomaly and temperature rise defect range based on the temperature rise difference analysis.

2. The infrared image composite insulator heating defect recognition method under complex background according to claim 1 is characterized in that: Step S5 includes: S51. Obtaining the temperature value of each pixel in the rod core region based on the rod core mask matrix with temperature; S52. Take the average value of all temperature values ​​in the rod core area as the reference temperature; S53. Based on the reference temperature, the temperature rise value △T(x, y) of each pixel in the rod core area is calculated to obtain a temperature rise matrix: △T(x, y)=|T(x, y)-T base |, T(x, y) represents the temperature value of the corresponding pixel, T base is the reference temperature; S54. Determine whether the rod core has abnormal heating and determine the temperature rise defect range based on the temperature rise matrix.

3. The infrared image composite insulator heating defect recognition method under complex background according to claim 2 is characterized in that: When the temperature rise matrix meets the set conditions, it is judged that the rod core has heating abnormality and the temperature rise defect range is further determined.

4. The method for identifying heating defects of composite insulators using infrared images under complex backgrounds according to claim 3 is characterized in that: A temperature rise threshold is included. When there are at least a set number of temperature rise values ​​in the temperature rise matrix that are greater than the temperature rise threshold, the temperature rise defect range is determined.

5. The method for identifying heating defects of composite insulators using infrared images under complex backgrounds according to claim 4 is characterized in that: Determine the temperature rise defect range by: S541. Select all pixels whose temperature rise values ​​are greater than the temperature rise threshold; S542. Taking the selected pixel as the center, check several pixels around the center point. If (△P0-△P i )<ε, the corresponding pixel is included in the heating defect area, △P0 represents the temperature rise value of the center pixel, △P i represents the temperature rise value of the pixels around the target, and ε is the temperature rise difference threshold; S543. All surrounding pixels that meet the conditions are included in the heating defect area to obtain the heating defect area corresponding to the corresponding center point; S544. Merge the heating defect areas of all center points to obtain the final temperature rise defect range.

6. The method for identifying heating defects of composite insulators using infrared images under complex backgrounds according to claim 5 is characterized in that: When there are at least a set number of temperature rise values ​​in the temperature rise matrix that are greater than the temperature rise threshold, it is preliminarily determined that there is a heating anomaly; If the final temperature rise defect range is greater than the set value, it is confirmed that there is abnormal heating; Otherwise, continue to perform the detection of steps S1-S5, and the detection speed is greater than the thermal conduction speed of the unit pixel, and obtain the temperature rise defect range of each detection. If the temperature rise defect range shows a trend of gradually expanding over time, or the final temperature rise defect range of at least one detection during the detection period is greater than the set range, it is confirmed that there is a heating abnormality.

7. The method for identifying heating defects of composite insulators using infrared images under complex backgrounds according to claim 1 is characterized in that: In step S2, the Deeplabv3plus model is used to perform image segmentation on the composite insulator infrared image to obtain a composite insulator rod core area mask; Step S2 specifically includes: S21. Normalizing the infrared image of the composite insulator; S22. The Deeplabv3plus model extracts features from the image processed by S21 to obtain a multi-scale feature map; S23. Classify each pixel in the multi-scale feature map, and output a probability matrix S(x, y) that each pixel belongs to the composite insulator rod core region, where (x, y) represents the pixel coordinates in the image, and the probability matrix represents the probability value of each pixel belonging to the rod core region; S24. Pixel points whose probability values ​​are greater than or equal to the probability threshold are screened out according to the probability matrix S(x, y), and the corresponding pixel points are marked as the rod core area, and a binary mask matrix about the rod core area is obtained, where one value represents the rod core area and the other value represents the non-rod core area, thereby obtaining the composite insulator rod core area mask.

8. The method for identifying heating defects of composite insulators using infrared images under complex backgrounds according to claim 1 is characterized in that: Step S3 specifically includes: S31. Decode the additional temperature information in the infrared image and extract the temperature data corresponding to each pixel in the image; S32. Perform outlier processing on temperature data: Calculate the standard deviation distance between each temperature value and the mean of the temperature set; If the temperature value of a certain pixel is more than the outlier threshold from the average value, the pixel is corrected according to the temperature values ​​of the neighboring pixels; S33. Perform a normal transformation on the temperature data to make it close to a normal distribution.

9. The method for identifying heating defects of composite insulators using infrared images under complex backgrounds according to claim 8, characterized in that: In step S4, when extracting the temperature data of the composite insulator rod core region, all temperature data corresponding to the rod core region are extracted from the temperature data matrix using the rod core region mask matrix obtained in step S2: M core ′(x, y) represents a binary pickled matrix, 1 represents that the corresponding pixel is in the rod core region, and 0 represents that the corresponding pixel is not in the rod core region.

10. A composite insulator heating defect recognition system based on infrared images under complex backgrounds, characterized in that: The system includes a drone, wherein the drone is provided with an infrared lens and a control module; The infrared lens is used to capture the infrared image of the composite insulator and transmit the infrared image of the composite insulator to the control module; The control module is used to execute the method described in any one of claims 1 to 9; And the control module includes a wireless module for transmitting the defect identification result to the ground background.

Images