A smart detection method for power system distribution lines

Abstract

This invention belongs to the field of image processing technology, specifically relating to an intelligent detection method for power system distribution lines. The method includes: performing multi-scale analysis on infrared images of power towers to obtain the gradient amplitude and direction of pixels; calculating the degree to which a pixel belongs to a temperature-altering edge based on local fluctuations in gradient amplitude and the stability of gradient direction clustering, and classifying pixels into temperature-altering, gradually changing, and non-edge pixels; calculating the significance coefficient of each edge pixel by combining pixel temperature value, temperature change trend, and edge type; finally, weighting the area proportion of different types of abnormal regions based on the significance coefficients to assess the overall abnormal temperature rise of the power tower and even the distribution line, and dynamically adjusting the maintenance cycle interval. This invention can accurately distinguish different types of temperature edges, effectively identify key fault points, and improve the accuracy and reliability of detection.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to an intelligent detection method for power system distribution lines. Background Technology

[0002] Power distribution lines are a core component of the power system, and their operating status directly affects the reliability of power supply. Traditional manual inspection methods have significant drawbacks such as low efficiency and numerous blind spots. Drone inspection technology, equipped with high-definition cameras, infrared thermal imagers, and other sensing devices, can effectively cover complex terrain areas such as mountainous regions, efficiently collect infrared image data of power distribution lines, and capture the temperature distribution characteristics of various components of the towers, especially fuses, making it easier for technicians to detect potential risks in a timely manner.

[0003] Inspecting the towers in power distribution lines is a crucial step in ensuring the safe operation of the power system. When the internal resistance of the fuses on the towers increases abnormally due to aging, poor contact, or overload, the current flowing through them will generate a significant Joule heating effect, causing an abnormal increase in local temperature and affecting the stable operation of the transmission line. Therefore, infrared thermography can be used to detect whether there are any abnormalities in the power distribution lines.

[0004] In infrared thermal images, temperature distribution is typically a continuous, gradual change. This can cause the edges of locally overheated areas to appear blurred in the image, such as the faulty area of ​​a fuse that is locally overheated. Traditional edge detection algorithms struggle to accurately distinguish between different types of temperature edges: firstly, abrupt edges with drastic temperature changes, such as the sharp boundary of an overheated fault core area; and secondly, gradually changing edges with gentle temperature changes, such as the diffuse boundary of an overheated area. This leads traditional edge detection algorithms to treat edges of different intensities and types indiscriminately, resulting in inaccurate edge localization and consequently reducing the accuracy of detection of power distribution lines in power systems. Summary of the Invention

[0005] To address the technical problem that traditional edge detection algorithms indiscriminately process edges of different intensities and types, resulting in inaccurate edge localization and consequently reducing the accuracy of power distribution line detection in power systems, this invention provides an intelligent detection method for power distribution lines.

[0006] In a first aspect, the present invention provides an intelligent detection method for power system distribution lines, employing the following technical solution: Otsu threshold segmentation is performed on the infrared image of the towers within the distribution line to obtain tower region images; Gaussian pyramid decomposition is performed on the tower region images to obtain image layers of different scales; based on the gradient magnitude and gradient direction of pixels in the image layers of different scales, the gradient magnitude and gradient direction of each pixel in the tower region image at different scales are obtained; based on the degree of local fluctuation of the gradient magnitude and the degree of clustering stability of the gradient direction between different scales, the degree to which a pixel belongs to a temperature-abrupt edge is obtained. The system categorizes pixels into temperature-abrupt edge pixels, temperature-gradient edge pixels, and non-edge pixels. Based on the pixel's temperature magnitude, temperature change, and degree of belonging to a temperature-abrupt edge, it obtains the significance coefficients corresponding to all temperature-abrupt edge pixels and temperature-gradient edge pixels. These coefficients are then used to weight the area proportions of temperature-abrupt and temperature-gradient regions composed of temperature-abrupt and temperature-gradient edge pixels, respectively, to obtain the degree of abnormal temperature rise in the tower area image. Based on the degree of abnormal temperature rise in all tower area images, the fixed cycle for power distribution line maintenance is adjusted.

[0007] This invention accurately distinguishes between temperature-abrupt and temperature-gradient edge features through multi-scale gradient analysis, and dynamically calculates pixel saliency coefficients by combining temperature time-series data. Finally, it comprehensively evaluates the degree of anomaly by weighting area proportion and saliency. This method effectively solves the problem of low positioning accuracy caused by the inability of traditional global edge detection algorithms to distinguish between different types of temperature edges. It can accurately identify key fault points such as poor fuse contact, providing a scientific basis for preventive maintenance of power facilities and significantly improving the detection accuracy and power supply reliability of distribution lines.

[0008] Preferably, obtaining the tower area image includes: constructing a grayscale histogram of the denoised infrared image; using the maximum inter-class variance method to determine a threshold for binarization segmentation; morphologically filling holes and then using connected component analysis to label the tower area to obtain a binary image, including foreground pixels and background pixels; and combining all pixels in the infrared image that are at the same position as the foreground pixels in the binary image to form the tower area image.

[0009] This method uses denoised grayscale histogram analysis and the maximum inter-class variance method to automatically determine the optimal segmentation threshold. Combined with morphological filling and connected component analysis, it effectively separates the tower area from the background, significantly improving the accuracy and completeness of tower area segmentation. This provides precise target areas for intelligent detection of power system distribution lines, enhancing detection efficiency and reliability.

[0010] Preferably, the determination of the degree to which a pixel belongs to a temperature-change-type edge includes: in the first... In an image at various scales, a 3×3 neighborhood is constructed centered on the i-th pixel, serving as the neighborhood of the i-th pixel at the i-th pixel in the i-th image. The neighborhood at each scale of the image is quantified by the following formula: ;in, It represents the degree to which the i-th pixel belongs to a temperature-abrupt edge; Is the i-th pixel at the i-th position? Within a neighborhood at a given scale, all pixels at the th... The entropy value corresponding to the gradient magnitude at each scale; Is the i-th pixel at the i-th position? Within a neighborhood at a given scale, all pixels at the th... The variance of the corresponding gradient direction at each scale; It is the total number of scales. It is a normalization of maximum and minimum values.

[0011] This invention utilizes multi-scale gradient magnitude entropy With directional variance The collaborative computing of [the technology] solves the problem of distinguishing between abrupt and gradual edges in infrared images, thereby reducing positioning errors.

[0012] Preferably, obtaining the gradient magnitude and gradient direction of each pixel in the tower area image at different scales includes: denoting the size of the tower area image as... Then the first The size of the layer image is For the tower area image with coordinates as The pixel in the first position The coordinates of the mapped pixels in the layer image are , It is the floor function, used to obtain the coordinates in the tower area image. The pixels are mapped to the pixels in the image layer at different scales; then the gradient features of the mapped pixels of each pixel in the tower area image at different scales are used as the gradient features of each pixel in the image layer at different scales. The gradient features include the gradient magnitude and gradient direction.

[0013] This invention obtains image layers of different scales through Gaussian pyramid decomposition, enabling the acquisition of multi-scale information of the image. This lays the foundation for subsequent analysis of the gradient features of pixels at different scales. Furthermore, it establishes a precise mapping relationship between pixels at each scale layer, accurately acquiring the gradient magnitude and gradient direction information of each pixel in the tower area image in multi-scale space. This provides a multi-dimensional gradient feature basis for subsequent determination of whether a pixel belongs to a temperature-abrupt edge or a temperature-gradient edge, thereby effectively improving the accuracy and reliability of edge detection in abnormally heated areas of power distribution lines.

[0014] Preferably, dividing pixels into temperature-abrupt edge pixels, temperature-gradual edge pixels, and non-edge pixels includes: setting two edge probability thresholds. ,Require When probability When the i-th pixel belongs to the edge pixel of temperature change type; when the probability When the i-th pixel belongs to the temperature-gradient edge pixel type; when the probability When the i-th pixel is a non-edge pixel, the i-th pixel is considered a non-edge pixel.

[0015] This invention employs a dual threshold mechanism to classify edge types into three levels, and this hierarchical strategy reduces the false alarm rate of fault identification.

[0016] Preferably, obtaining the significance coefficients corresponding to all temperature-abrupt edge pixels and temperature-gradient edge pixels includes: ;in, It is the significance coefficient corresponding to the i-th pixel; It represents the degree to which the i-th pixel belongs to a temperature-abrupt edge; Is the i-th pixel at the i-th position? The normalized value of the temperature corresponding to the second; It uses the least squares method to calculate the i-th pixel at the i-th position. The normalized value of the temperature value within a second is fitted to obtain the fitting slope; It is the first weight of the i-th pixel. It is the second weight of the i-th pixel. It is the third weight of the i-th pixel. ,and ; It is the hyperbolic tangent function. It is a natural exponential function.

[0017] This invention obtains a significance coefficient by integrating morphological features, temperature, and variation trends, thereby strengthening the dominance of morphological features; wherein, Amplify the contribution of high-temperature regions. Suppressing transient overload interference improves the detection rate of actual fault points.

[0018] Preferably, the degree of abnormal temperature rise in the obtained tower area image includes: ;in, It indicates the degree of abnormal temperature rise in the tower area image; It is the area of ​​all regions with abrupt temperature changes. It is the area of ​​all regions with gradual temperature changes. It represents the image area of ​​the entire tower region; It is the average of the significance coefficients of all edge pixels with temperature abrupt changes. It is the average of the saliency coefficients of all temperature-gradient edge pixels.

[0019] This invention classifies and evaluates the degree of abnormal temperature rise in tower area images by setting an abnormal temperature rise threshold, enabling accurate judgment of the tower's health status. When the abnormal temperature rise of the tower is detected to reach or exceed the preset threshold, a serious abnormal state can be identified in a timely manner and an emergency maintenance mechanism can be triggered. This effectively avoids the risk of accidents caused by failure to detect key fault points in a timely manner, significantly improves the safety and reliability of power distribution line operation, and provides a scientific basis for preventive maintenance of power facilities.

[0020] Preferably, the process of obtaining the area of ​​all temperature-abrupt regions, the area of ​​all temperature-gradient regions, and the area of ​​the entire tower region image includes: taking the total number of pixels contained in the entire tower region image as the area of ​​the entire tower region image, denoted as... Connectivity analysis was used to obtain several independent temperature-altering regions composed of edge pixels with abrupt temperature changes, and several independent temperature-gradiently changing regions composed of edge pixels with gradual temperature changes. The total number of pixels in all temperature-altering regions was used as the area of ​​all temperature-altering regions. The total number of pixels within all temperature-gradient regions is used as the area of ​​all temperature-gradient regions. .

[0021] This invention combines connected component analysis with weighted area ratios: the proportion of aberrant regions. Reflects the spatial distribution of urgent defects; proportion of slowly changing areas. Reflects the extent of long-term risk diffusion; significance mean , Quantify the intensity of anomalies to avoid the problem of missed detection by the traditional temperature threshold method.

[0022] Preferably, adjusting the fixed maintenance cycle of the power distribution lines based on the abnormal temperature rise in images of all tower areas includes: ;in, It is the correction cycle interval for the next maintenance of the power distribution line; It is the arithmetic mean of the abnormal temperature rise in the images of all tower areas in the power distribution line; It is the scheduled interval for the maintenance of power distribution lines; It is an adjustment factor, which requires... ; It is the overall anomaly threshold, which requires... .

[0023] Preferably, the step of adjusting the fixed maintenance cycle of the power distribution lines based on the abnormal temperature rise levels in all tower area images further includes: setting an abnormal temperature rise threshold. The degree of abnormal temperature rise in the tower area image Greater than or equal to the abnormal temperature rise threshold Emergency repairs were still carried out at that time.

[0024] This invention establishes a dynamic correction model for maintenance cycles, which significantly shortens the cycle for severe anomalies, slightly shortens the cycle for moderate anomalies, and maintains the original cycle for low anomalies, thereby optimizing the efficiency of operation and maintenance resource allocation.

[0025] The beneficial effects of this invention are as follows: This invention accurately segments the tower region in infrared images using the Otsu threshold algorithm, reducing computational load. Considering that the enhancement requirements for edges in different temperature change regions within the tower region image vary (e.g., temperature abrupt edges and temperature gradual change edges), this invention analyzes the local fluctuation of gradient amplitude and the stability of gradient direction aggregation of pixels in the tower region image at different scales. This allows for accurate calculation and identification of edge pixels belonging to temperature abrupt and temperature gradual change types. Based on temperature magnitude, temperature change, and the degree to which a pixel belongs to a temperature abrupt edge, the invention obtains the significance coefficients corresponding to all temperature abrupt and temperature gradual change edge pixels. These significance coefficients quantify the importance of pixels as abnormal temperature rise edges, further revealing the degree of abnormal temperature rise in power distribution lines, effectively improving the accuracy of intelligent detection of power distribution lines. Attached Figure Description

[0026] Figure 1 This is a schematic flowchart illustrating an intelligent detection method for power system distribution lines according to the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0029] This invention discloses an intelligent detection method for power system distribution lines, referring to... Figure 1 This includes steps S1 to S4:

[0030] S1. Perform Otsu threshold segmentation on the infrared image of the tower in the power distribution line to obtain the tower area image; perform Gaussian pyramid decomposition on the tower area image to obtain image layers of different scales; based on the gradient magnitude and gradient direction of the pixels in the image layers of different scales, obtain the gradient magnitude and gradient direction of each pixel in the tower area image at different scales.

[0031] It should be noted that when acquiring infrared images of towers within power distribution lines, random thermal noise inside the acquisition equipment can cause noise interference to the infrared images. This noise is not only a useless area in the infrared images that does not need to be enhanced, but it can also affect the subsequent diagnostic results.

[0032] Based on this, this embodiment first converts the infrared image collected by the UAV into a grayscale image, and then uses a Gaussian filtering algorithm to denoise the infrared image, while preserving the details of the image. The step of removing noise from the infrared image using a Gaussian filtering algorithm can be obtained by existing technology and will not be described in detail here.

[0033] Specifically, the Otsu threshold segmentation is performed on the infrared image of the towers in the power distribution line to obtain the tower area image. The specific method is as follows: construct the grayscale histogram of the denoised infrared image, use the maximum inter-class variance method to determine the threshold for binarization segmentation, fill the holes with morphology, and use connected component analysis to mark the tower area to obtain a binary image, including foreground pixels and background pixels; all pixels in the infrared image that have the same position as the foreground pixels in the binary image are combined to form the tower area image. The specific steps of the Otsu threshold segmentation algorithm can be implemented by existing technology and will not be elaborated here.

[0034] Furthermore, Gaussian pyramid decomposition is performed on the pole region image to obtain image layers of different scales. The specific method for Gaussian pyramid decomposition on the pole region image is as follows: First, Gaussian smoothing filtering and downsampling operations are performed iteratively to construct an image pyramid structure with multi-scale characteristics. During the generation of this pyramid, the size of adjacent layers strictly follows the rule of halving layer by layer, that is, the resolution of each layer is reduced by 50% compared to the previous layer, thus obtaining image layers of different scales. The specific algorithm implementation of Gaussian pyramid decomposition belongs to the existing technology in the field of image processing, and its standard process can be followed by existing technical solutions. Detailed derivation will not be elaborated here.

[0035] Specifically, the gradient magnitude and gradient direction of each pixel at different scales of the image layer are obtained. The specific method is as follows: using the standard Sobel operator, gradient features are extracted sequentially for each pixel in the obtained image layers at different scales; the size of the tower area image is denoted as... Then the first The size of the layer image is For the tower area image with coordinates as The pixel in the first position The coordinates of the mapped pixels in the layer image are , It is the floor function, used to obtain the coordinates in the tower area image. The pixels are mapped to the pixels in the image layer at different scales; then the gradient features of the mapped pixels of each pixel in the tower area image at different scales are used as the gradient features of each pixel in the image layer at different scales. The gradient features include the gradient magnitude and gradient direction.

[0036] S2. Based on the degree of local fluctuation of the gradient magnitude of the pixel at different scales and the degree of stability of the gradient direction, the degree to which the pixel belongs to the temperature change type edge is obtained, which is used to classify the pixel into temperature change type edge pixels, temperature gradual change type edge pixels and non-edge pixels.

[0037] It should be noted that the degree of local fluctuation calculated based on gradient magnitude entropy can reflect the degree of disorder of gradient magnitudes in the neighborhood to some extent: when the magnitude difference in the neighborhood is large, the entropy value is high, corresponding to abrupt edges, such as the sharp boundary of a faulty fuse; while when the magnitude distribution is uniform, the entropy value is low, corresponding to gradually changing edges, such as the diffusion boundary of abnormal temperature rise. In addition, the directional aggregation stability calculated based on gradient direction variance can reflect the degree of dispersion of gradient directions in the neighborhood to some extent: when the directions are disordered, the variance is large, corresponding to abrupt edges with convergence of gradients in multiple directions; while when the directions are consistent, the variance is small, corresponding to gradually changing edges with convergent gradient directions.

[0038] Based on this, for pixels in the tower area image, the embodiments of the present invention can obtain the degree to which a pixel belongs to a temperature change type edge by the degree of local fluctuation of the gradient amplitude between different scales of the pixel and the degree of aggregation stability of the gradient direction.

[0039] Specifically, based on the local fluctuation of gradient magnitude and the stability of gradient direction aggregation of pixels at different scales in the tower area image, the method for determining the degree to which a pixel in the tower area image belongs to a temperature abrupt change edge is as follows: Obtain a 3×3 neighborhood centered on each pixel; calculate the degree of temperature abrupt change edge for all pixels within the neighborhood at different scales. The entropy value of the gradient magnitude at all scales is calculated and normalized; the average entropy value of the gradient magnitude at all scales is calculated as the local fluctuation of the gradient magnitude of each pixel at different scales in the tower region image; the entropy value of the gradient magnitude of all pixels in the neighborhood at all scales is calculated. The variance of the gradient direction is calculated and normalized; the average variance of the gradient direction at all scales is calculated as the degree of stability of the gradient direction aggregation of each pixel at different scales in the tower area image.

[0040] Furthermore, the degree to which the i-th pixel belongs to a temperature-abrupt edge is calculated, as detailed in the following formula:

[0041] ;

[0042] in, It represents the degree to which the i-th pixel belongs to a temperature-abrupt edge, with the range limited to... between; Is the i-th pixel at the i-th position? At each scale, the entropy value of the gradient magnitude of all pixels in a 3×3 neighborhood centered on the i-th pixel. Is the i-th pixel at the i-th position? At each scale, the variance of the gradient direction of all pixels in a 3×3 neighborhood centered on the i-th pixel. The total number of scales; Indicates the degree of local fluctuation in gradient magnitude; Indicates the degree of clustering stability along the gradient direction; This indicates normalization processing, which can be achieved using maximum and minimum value normalization.

[0043] Specifically, based on the comparison between the degree to which a pixel belongs to a temperature-sudden change edge and the edge probability threshold, pixels are divided into temperature-sudden change edge pixels, temperature-gradient change edge pixels, and non-edge pixels. The specific method is as follows: Set two edge probability thresholds. When probability When the probability indicates that the gradient magnitude of the neighborhood fluctuates drastically and the fluctuation direction is poorly clustered, which meets the characteristics of a sudden change edge, then the i-th pixel is considered to be a temperature sudden change edge pixel; when the probability When the probability indicates that the gradient magnitude fluctuation in the neighborhood is gentle and the fluctuation direction is strongly clustered, which conforms to the characteristics of a gradually changing edge, then the i-th pixel is considered to belong to a pixel of a gradually changing temperature edge; when the probability When the gradient amplitude fluctuation in the neighborhood is very smooth and the fluctuation direction is very strongly clustered, which is consistent with the characteristics of non-edge, the i-th pixel is considered to be a non-edge pixel. In this way, all temperature-abrupt edge pixels, temperature-gradient edge pixels, and non-edge pixels are obtained.

[0044] Among them, the edge probability threshold The specific settings can be configured according to actual needs, and the following conditions must be met: The embodiments of the present invention are not limited in many ways, for example .

[0045] S3. Based on the temperature magnitude, temperature change, and the degree to which a pixel belongs to a temperature abrupt edge, obtain the significance coefficients corresponding to all temperature abrupt edge pixels and temperature gradually changing edge pixels.

[0046] It should be noted that the degree to which each pixel belongs to a temperature-abrupt edge obtained in step S2 is determined from the perspective of image gradient morphology to determine the edge type of a pixel. In actual fault detection, edge pixels with high-temperature characteristics are more likely to have faults, while edges that are edges but do not have high-temperature characteristics are less likely to have faults, such as the outline of the metal parts of a tower. Therefore, it is necessary to further combine temperature characteristics to obtain the significance coefficient of each pixel. The significance coefficient of each pixel is used to reflect that each pixel has both the morphological characteristics of an edge and the physical properties of high temperature, so that the corresponding pixel is the target that needs to be focused on in the future.

[0047] Based on this, embodiments of the present invention can obtain the significance coefficient of each pixel by using a nonlinear weighting method based on the similarity of the temperature of the pixels in the infrared image and the degree to which the pixels belong to the edge of a temperature change.

[0048] Specifically, based on the infrared images collected by the drone every second, the temperature value of each pixel in the tower area image at each second is obtained. Then, a maximum-minimum value algorithm is used to normalize the temperature value of each pixel at each second, and the temperature value of the i-th pixel at the i-th second is normalized. The normalized value of the temperature corresponding to the second is denoted as .

[0049] Furthermore, regarding the first Within a given second, for all infrared images of the tower region, the normalized temperature values ​​of the corresponding i-th pixel are fitted using the least squares method, and the resulting fitting slope is denoted as... This fitted slope is used to reflect the i-th pixel in the tower area image at the th... The trend of temperature change per second.

[0050] in, This represents the time increment, and its value range is a positive integer greater than or equal to 1. The specific value can be set according to actual needs; for example, when... When, it means that for the i-th pixel in the tower area image, at the time... All normalized temperature values ​​within a given second are fitted using the least squares method to reflect the changing trend of the temperature value corresponding to the i-th pixel in the tower area image within the time range corresponding to 3 seconds after the current moment.

[0051] Among them, when the fitting slope When, it reflects the i-th pixel in the tower area image at the th... The temperature shows an increasing trend over seconds, when the fitting slope When, it reflects the i-th pixel in the tower area image at the th... The temperature decreases over time, and the fitted slope... When, it reflects the i-th pixel in the tower area image at the th... The temperature remained stable over the seconds.

[0052] Specifically, based on the temperature magnitude, temperature variation, and the degree to which a pixel belongs to a temperature-abrupt edge, the significance coefficients corresponding to all temperature-abrupt edge pixels and temperature-gradient edge pixels are obtained, as shown in the following formula:

[0053] ;

[0054] In the relation, It is the significance coefficient corresponding to the i-th pixel; It represents the degree to which the i-th pixel belongs to a temperature-abrupt edge, with the range limited to... between; Is the i-th pixel at the i-th position? The normalized value of the temperature corresponding to the second reflects the temperature of the i-th pixel. Is the i-th pixel at the i-th position? All normalized temperature values ​​within a second are fitted using the least squares method, and the resulting fitting slope reflects the temperature change of the i-th pixel. It is the first weight of the i-th pixel. It is the second weight of the i-th pixel. It is the third weight of the i-th pixel; Its function is to The scope of the item is limited to between; It is a hyperbolic tangent function, and its function is to limit the rate of temperature change to a certain value. To avoid the formula being affected by extreme values, It is a natural exponential function.

[0055] in, It is a natural exponential function, which amplifies the temperature contribution of high-temperature regions, consistent with the physical phenomenon that "high temperatures are more likely to cause failures": when the i-th pixel is at the th... When the overall temperature trend continues to rise within a second, the fitting slope , The term will be greater than 1, thus amplifying the current temperature. The influence of this causes a sharp increase in the saliency of a point that is heating up; when the i-th pixel is at the th... When the overall temperature trend stabilizes within a second, the fitting slope , The term will be approximately equal to 1, at which point the dynamic factor will be approximately equal to the current temperature value; when the i-th pixel is at the... When the overall temperature trend continues to decrease within a second, the fitting slope , The term will be less than 1, thus suppressing the current temperature. This effectively eliminates temperature interference at points that are recovering from momentary overloads.

[0056] Specifically, due to the first weight of the i-th pixel Second weight and the third weight The physical meanings described are different; therefore, the first weight of the i-th pixel is... Second weight and the third weight There is a size relationship: the first weight of the i-th pixel. The physical meaning described is the importance of image gradient morphological features; second weight The physical meaning described is the importance of the absolute value of temperature; third weight. The physical meaning described is the importance of temperature change trends; based on experience with image gradient morphological features as the primary factor, temperature as a secondary factor, and temperature change trends as a reference, the weights are... It can be configured according to actual needs, and must meet the following requirements: ,and The embodiments of the present invention are not limited in many ways, for example: .

[0057] When the i-th pixel belongs to the temperature change type edge The larger the value, the more likely the i-th pixel is to be at the i-th position. Normalized value of temperature corresponding to seconds The larger the value, and the i-th pixel is at the th The corresponding fitting slope within seconds The larger the value, the greater the significance coefficient of the i-th pixel, reflecting an edge with a high temperature and a continuing upward trend in temperature. Conversely, the smaller the value, the greater the significance coefficient of the i-th pixel. The larger the significance coefficient of the i-th pixel, the greater the importance of the i-th pixel as an edge of abnormal temperature rise, and vice versa.

[0058] Based on the above operations, the significance coefficients corresponding to all edge pixels with abrupt temperature changes and edge pixels with gradual temperature changes are obtained.

[0059] S4. Based on the significance coefficients of all edge pixels with abrupt temperature changes and edge pixels with gradual temperature changes, the area ratios of the temperature-abrupt and temperature-gradient regions are weighted to obtain the degree of abnormal temperature rise in the tower area image, and mapped to the overall abnormal temperature rise of the power distribution line to guide maintenance decisions.

[0060] It should be noted that by calculating the area ratio of temperature-abrupt and temperature-gradient regions, and combining this with the significance coefficient of each edge pixel, the spatial distribution breadth and intensity of abnormal temperature rise regions can be quantified. The significance coefficient of each edge pixel can comprehensively reflect the importance of morphological features and temperature attributes. Regions with a high area ratio of abrupt temperature rise and a large significance coefficient, such as areas with poor fuse contact, usually correspond to urgent defects and require immediate repair. On the other hand, regions with a low area ratio of gradual temperature rise and a small significance coefficient, such as areas with insulation aging, indicate potential long-term risks. Finally, the results of local anomaly detection are mapped to the overall status of the tower, enabling accurate assessment and graded early warning of the health status of power distribution lines.

[0061] Specifically, the total number of pixels contained in the entire tower area image is counted and used as the area of ​​the entire tower area image, denoted as . Connected component analysis is used to obtain several independent temperature-abrupt regions composed of edge pixels with abrupt temperature changes, and several independent temperature-gradient regions composed of edge pixels with gradual temperature changes. The specific operation of connected component analysis can be implemented by existing technology and will not be elaborated here.

[0062] Furthermore, the total number of pixels in all temperature-change regions, including those at the edges of the temperature-change regions, is counted and used as the area of ​​all temperature-change regions, denoted as . The total number of pixels in all temperature-gradient regions, including those at the edges, is counted and used as the area of ​​all temperature-gradient regions, denoted as . .

[0063] Specifically, the area ratios of temperature-abrupt and temperature-gradient regions are weighted based on the significance coefficients of all edge pixels corresponding to temperature-abrupt and temperature-gradient edge pixels to obtain the degree of abnormal temperature rise in the tower area image. See the following formula for details:

[0064] ;

[0065] in, This represents the degree of abnormal heating in the tower area image, with a value range of [value range missing]. The greater the degree of abnormal heating in the tower area image, the more abnormal the tower area image is; It is the area of ​​all regions with abrupt temperature changes. It is the area of ​​all regions with gradual temperature changes. It represents the image area of ​​the entire tower region; , These are the average significance coefficients of all edge pixels with abrupt temperature changes and edge pixels with gradual temperature changes within their respective temperature change regions.

[0066] Specifically, It is the ratio of the area of ​​the temperature-abrupt region to the total image area of ​​the tower area; It is the ratio of the area of ​​the temperature-gradient region to the total image area of ​​the tower area; Divide the middle denominator by the total number of pixels in the tower area This is to eliminate the influence of image size; when the anomalous intensity of the abrupt change region is greater, and the anomalous intensity of the gradually changing region is greater, the anomalous temperature rise of the tower region image is greater, indicating that the tower region image is more abnormal, and vice versa.

[0067] Furthermore, based on the comparison between the degree of abnormal temperature rise in the tower area image and the abnormal temperature rise threshold, the tower condition assessment result is obtained. The specific method is as follows: Set the abnormal temperature rise threshold. It can be configured according to actual needs, and must meet certain requirements. The embodiments of the present invention are not limited in many ways, but are exemplary. The degree of abnormal temperature rise in the tower area image Greater than or equal to the abnormal temperature rise threshold When the temperature rises significantly in the tower area image, it indicates that the tower is in a severely abnormal state. For towers with significant abnormal temperature rise, emergency repairs are still carried out even though it is not yet time for scheduled maintenance.

[0068] Based on the above steps, the degree of abnormal temperature rise in each tower area image of the acquired power distribution line can be calculated. The arithmetic mean of the abnormal temperature rise in all tower area images of the power distribution line is calculated as the overall abnormal temperature rise of the power distribution line. The original maintenance cycle interval of the power distribution line is then corrected based on the overall abnormal temperature rise of the power distribution line to obtain the corrected maintenance cycle interval for the next maintenance of the power distribution line. See the following formula for details:

[0069] ;

[0070] in, It is the correction cycle interval for the next maintenance of the power distribution line; It is the arithmetic mean of the abnormal temperature rise in the images of all tower areas in the power distribution line; It is the scheduled interval for the maintenance of power distribution lines; This is an adjustment factor, which can be set according to actual needs and must meet certain requirements. The embodiments of the present invention are not limited in many ways, but are exemplary. ; This is the overall anomaly threshold, which can be set according to actual needs and must meet certain requirements. The embodiments of the present invention are not limited in many ways, but are exemplary. .

[0071] Specifically, when When this occurs, it indicates that the power distribution line is in a historically severe abnormal state, and the revised maintenance interval should be shortened to 0.5 times the original interval. For example, if the original maintenance interval was 30 days, it should be revised to 15 days. When this occurs, it indicates that the power distribution line is in a historically moderately abnormal state, and the revised maintenance interval should be shortened to 0.8 times the original interval. For example, if the original maintenance interval was 30 days, it should be revised to 24 days. When this occurs, it indicates that the power distribution line is in a historically low-level abnormal state, and the correction interval for the next maintenance is consistent with the original interval.

[0072] This embodiment provides an intelligent detection method for power system distribution lines. By analyzing multi-scale gradient and temperature time-series data, it distinguishes pixels with abrupt changes and gradual changes, dynamically calculates the significance coefficient of pixels, and combines weighted evaluation to accurately locate faults such as poor fuse contact, significantly improving detection accuracy and power supply reliability.

Claims

1. An intelligent detection method for power system distribution lines, characterized in that, include: Infrared images of power distribution line towers are segmented using the Otsu threshold method to obtain images of the tower area. Gaussian pyramid decomposition is performed on the tower area image to obtain image layers at different scales. Based on the gradient magnitude and gradient direction of the pixels in the image layers at different scales, the gradient magnitude and gradient direction of each pixel in the tower area image at different scales are obtained. Based on the degree of local fluctuation of gradient magnitude and the degree of stability of gradient direction between different scales, the degree to which a pixel belongs to a temperature-abrupt edge is obtained, and the pixels are divided into temperature-abrupt edge pixels, temperature-gradient edge pixels and non-edge pixels. Based on the temperature size, temperature change, and degree of belonging to the temperature abrupt edge of each pixel, the significance coefficients corresponding to all temperature abrupt edge pixels and temperature gradual edge pixels are obtained. These coefficients are used to weight the area ratio of the temperature abrupt region and the temperature gradual region composed of temperature abrupt edge pixels and temperature gradual edge pixels, so as to obtain the degree of abnormal temperature rise in the tower area image. Based on the abnormal temperature rise in images of all tower areas, adjust the fixed cycle for power distribution line maintenance.

2. The intelligent detection method for power system distribution lines according to claim 1, characterized in that, The process of obtaining the tower area image includes: A grayscale histogram of the denoised infrared image is constructed. The threshold is determined by maximizing the inter-class variance method for binarization segmentation. After morphological filling of holes, connected component analysis is used to mark the tower region to obtain a binary image, including foreground and background pixels. All pixels in the infrared image that are in the same position as the foreground pixels in the binary image are combined to form the tower region image.

3. The intelligent detection method for power system distribution lines according to claim 1, characterized in that, The degree to which a pixel belongs to a temperature-abrupt edge includes: In the In an image at various scales, a 3×3 neighborhood is constructed centered on the i-th pixel, serving as the neighborhood of the i-th pixel at the i-th pixel in the i-th image. The neighborhood at each scale of the image is quantified by the following formula: ； in, It represents the degree to which the i-th pixel belongs to a temperature-abrupt edge; Is the i-th pixel at the i-th position? Within a neighborhood at a given scale, all pixels at the th... The entropy value corresponding to the gradient magnitude at each scale; Is the i-th pixel at the i-th position? Within a neighborhood at a given scale, all pixels at the th... The variance of the corresponding gradient direction at each scale; It is the total number of scales. It is a maximum and minimum value normalization.

4. The intelligent detection method for power system distribution lines according to claim 1, characterized in that, The process of obtaining the gradient magnitude and gradient direction of each pixel in the tower area image at different scales includes: The size of the tower area image is denoted as . Then the first The size of the layer image is For the tower area image with coordinates as The pixel in the first position The coordinates of the mapped pixels in the layer image are , It is the floor function, used to obtain the coordinates in the tower area image. The pixels are mapped to the pixels in the image layer at different scales; then the gradient features of the mapped pixels of each pixel in the tower area image at different scales are used as the gradient features of each pixel in the image layer at different scales. The gradient features include the gradient magnitude and gradient direction.

5. The intelligent detection method for power system distribution lines according to claim 1, characterized in that, The process of dividing pixels into temperature-changing edge pixels, temperature-gradient edge pixels, and non-edge pixels includes: Set two edge probability thresholds ,Require When probability When the i-th pixel belongs to the edge pixel of temperature change type; when the probability When the i-th pixel belongs to the temperature-gradient edge pixel type; when the probability When the i-th pixel is a non-edge pixel, the i-th pixel is considered a non-edge pixel.

6. The intelligent detection method for power system distribution lines according to claim 1, characterized in that, The process of obtaining the significance coefficients corresponding to all temperature-abrupt edge pixels and temperature-gradient edge pixels includes: ； in, It is the significance coefficient corresponding to the i-th pixel; It represents the degree to which the i-th pixel belongs to a temperature-abrupt edge; Is the i-th pixel at the i-th position? The normalized value of the temperature corresponding to the second; It uses the least squares method to calculate the i-th pixel at the i-th position. The normalized value of the temperature value within a second is fitted to obtain the fitting slope; It is the first weight of the i-th pixel. It is the second weight of the i-th pixel. It is the third weight of the i-th pixel. ,and ; It is the hyperbolic tangent function. It is a natural exponential function.

7. The intelligent detection method for power system distribution lines according to claim 1, characterized in that, The degree of abnormal temperature rise in the obtained tower area image includes: ;in, It indicates the degree of abnormal temperature rise in the tower area image; It is the area of ​​all regions with abrupt temperature changes. It is the area of ​​all regions with gradual temperature changes. It represents the image area of ​​the entire tower region; It is the average of the significance coefficients of all edge pixels with temperature abrupt changes. It is the average of the saliency coefficients of all temperature-gradient edge pixels.

8. The intelligent detection method for power system distribution lines according to claim 7, characterized in that, The process of obtaining the area of ​​all temperature-abrupt regions, the area of ​​all temperature-gradient regions, and the area of ​​the entire tower region image includes: The total number of pixels contained in the entire tower area image is taken as the area of ​​the entire tower area image, denoted as . Connectivity analysis was used to obtain several independent temperature-altering regions composed of edge pixels with abrupt temperature changes, and several independent temperature-gradiently changing regions composed of edge pixels with gradual temperature changes. The total number of pixels in all temperature-altering regions was used as the area of ​​all temperature-altering regions. The total number of pixels within all temperature-gradient regions is used as the area of ​​all temperature-gradient regions. .

9. The intelligent detection method for power system distribution lines according to claim 1, characterized in that, The fixed cycle for adjusting the maintenance of power distribution lines based on the abnormal temperature rise in images of all tower areas includes: ； in, It is the correction cycle interval for the next maintenance of the power distribution line; It is the arithmetic mean of the abnormal temperature rise in the images of all tower areas in the power distribution line; It is the scheduled interval for the maintenance of power distribution lines; It is an adjustment factor, which requires... ; It is the overall anomaly threshold, which requires... .

10. The intelligent detection method for power system distribution lines according to claim 7, characterized in that, The method of adjusting the maintenance cycle of power distribution lines based on the abnormal temperature rise in images of all tower areas also includes: Set abnormal temperature rise threshold The degree of abnormal temperature rise in the tower area image Greater than or equal to the abnormal temperature rise threshold Emergency repairs will be carried out in such cases.

Images