Corrosion detection method for steel part of power transmission line

Abstract

The invention relates to a transmission line steel part corrosion detection method, which comprises the following steps of: acquiring a corrosion image of a high-voltage transmission line, removing a background by using a Depth Anything V2 model, and extracting a close-shot image; the method comprises the following steps: manually marking a close-shot image according to the corrosion degree to obtain marked data, and dividing a completely-corroded pixel set into different types of subsets through a K-means clustering algorithm, thereby calculating a central point of a corroded pixel and defining a color space triangle; the method comprises the following steps: acquiring a to-be-detected power transmission line image, performing background removal and component segmentation to obtain a region image of each component, and on the basis, performing depth calculation to screen out a foreground image; performing color value projection on the screened part area image, calculating a corrosion degree quantitative value, and counting evaluation parameters including an average corrosion degree, a maximum corrosion degree and a corrosion degree proportion; and according to the evaluation parameters, a corresponding corrosion grade is formulated, and finally scientific corrosion evaluation and monitoring are provided for the power transmission line.

Description

Technical Field

[0001] The invention relates to the technical field of power transmission line detection, and mainly to a method for detecting corrosion of steel components of a power transmission line. Background Art

[0002] Because transmission lines are exposed to the outdoors for a long time, they are susceptible to moisture, salt and other corrosive substances in the atmosphere, especially in humid or rainy areas. These factors will accelerate the oxidation reaction of metal components, leading to corrosion. Rust will not only reduce the mechanical strength of the transmission line and its components, but may also cause poor electrical contact, resulting in reduced power transmission efficiency, and even cause safety accidents such as disconnection and short circuit. Therefore, regular corrosion detection of transmission lines is essential to ensure the safe and stable operation of the power system.

[0003] For example, CN110726725A "A method and device for detecting corrosion of hardware of power transmission line" discloses "A method and device for detecting corrosion of hardware of power transmission line, the method first acquires the transmission line image, processes it to obtain the target hardware image; then segments the obtained target hardware image to obtain the foreground image and the background image; performs gamma transformation on the foreground image, uses the super red algorithm to perform grayscale processing on the gamma transformed foreground image to obtain the grayscale image of the foreground image; finally, determines the corrosion condition of the hardware according to the grayscale image of the foreground image. The present invention focuses on performing gamma transformation on the foreground image to perform grayscale processing on the hardware in the target hardware image. Strengthening the processing can adjust the impact of too strong or too weak light on the hardware in the target hardware image, so as to more accurately judge the rust degree and area of ​​the hardware, provide comprehensive rust defect information for operation and maintenance personnel, and provide effective technical means for realizing intelligent power operation and maintenance. However, the invention needs to set a brightness threshold to adjust the γ value of the gamma transform. This threshold-dependent method may be unstable under different environmental conditions. If the threshold is set improperly, it may affect the accuracy of corrosion detection. Therefore, a method for detecting corrosion of steel components of a power transmission line that maintains detection accuracy under different environmental conditions is urgently needed. Summary of the invention

[0004] In order to solve the above problems existing in the prior art, the present application provides a method for detecting corrosion of steel components of a power transmission line.

[0005] The technical solution of this application is as follows:

[0006] A method for detecting corrosion of steel components of a power transmission line, the method comprising:

[0007] Obtain a corrosion image of a high-voltage transmission line, remove the background and distant view in the corrosion image of the high-voltage transmission line based on the Depth Anything V2 model, and obtain the corresponding close-up image;

[0008] Manually annotate the rust in the close-up image according to the degree of rust to obtain annotated data; obtain a non-corroded pixel set and a completely corroded pixel set according to the annotated data, divide the completely corroded pixel set into two non-intersecting completely corroded pixel subsets of different types using a Kmean clustering algorithm, calculate and obtain the center points of the non-corroded pixel set and the completely corroded pixel subset, and define a color space triangle according to the center points;

[0009] Acquire an image of a high-voltage transmission line to be detected, remove the background and distant view in the image of the high-voltage transmission line to be detected based on the Depth Anything V2 model, and obtain a close-up image to be detected; process the close-up image to be detected using a skeleton generation technique to obtain a target structure, obtain prompt points of the Segment Anything2 model from the target structure, perform component segmentation on the image of the high-voltage transmission line to be detected according to the prompt points, and obtain regional images corresponding to various components in the high-voltage transmission line to be detected; perform depth calculation on the segmented regional images, and filter out the component regional images located in the foreground;

[0010] The color values ​​of the pixels in the screened component area image are projected onto the color space triangle to obtain the corresponding projection point; the quantified value of the degree of rust at the corresponding projection point is calculated, and the evaluation parameters in the component area image are counted, the evaluation parameters including the average degree of rust, the maximum degree of rust and the proportion of rust of all pixels, and the rust grade of the component corresponding to the component area image is evaluated according to the evaluation parameters.

[0011] As a preferred embodiment of the present invention, the background and distant view in the high-voltage transmission line corrosion image are removed based on the Depth Anything V2 model to obtain a close-up image. Specifically, the depth d corresponding to each pixel p in the high-voltage transmission line image is calculated. p , sort the depths corresponding to all pixels of the high-voltage transmission line image from large to small, regard the pixels in the last 50% of the sorting as the background, and obtain the close-up image after removing the background and the corresponding rust pixel set P.

[0012] As a preferred embodiment of the present invention, the rust in the close-up image is manually annotated according to the degree of rust to obtain annotated data, wherein:

[0013] The rust degree ranges from [0,1], and the rust degree w is marked for the pixels in the area without rust in the close-up image. pThe corresponding value is 0; the most severely corroded area in the close-up image contains pixels with the degree of corrosion w p The corresponding value is 1; obtain the annotation data corresponding to the rust pixel set P, and the annotation data includes the rust degree w corresponding to the pixel p and color value coordinate c p =(R p ,G p ,B p ).

[0014] As a preferred implementation of the present invention, the center points of the non-corroded pixel set and the completely corroded pixel subset are calculated, and a color space triangle is defined according to the center points, specifically:

[0015] According to the corroded pixel set P, the non-corroded pixel set S and the completely corroded pixel set T are obtained, which can be expressed as follows:

[0016] S={p∈Pw p =0};

[0017] T={p∈Pw p =1};

[0018] Calculate the red component R corresponding to any pixel in the completely corroded pixel set T p The mean The formula is:

[0019]

[0020] Where |T| represents the number of pixels in the completely corroded pixel set T;

[0021] The completely corroded pixel set T is divided into two groups according to the red component R p Divide into subset T1 0 and The formula is:

[0022]

[0023] Calculate the subset T1 0 and The center point and The formula is:

[0024]

[0025] Re-divide the subsets according to the center point, and iterate the division process until convergence, which can be expressed as:

[0026]

[0027] Where i is the final number of iterations, 0≤i≤N;

[0028] Get subset T1 N and The corresponding center points are and

[0029] For the uncorroded pixel set S and the completely corroded pixel subset T1 N and The reduction is performed by repeatedly calculating the center point of the set and deleting the k|S| points farthest from the center point in the set, where k is the reduction ratio; the set reduction process is repeated until the center point of the set does not change, and the final uncorroded pixel set S' and its corresponding center point μ are obtained. S , the completely corroded pixel subsets T1' and T2' and their corresponding center points and

[0030] The center point μ S , and As the three vertices of the triangle, get the color space triangle in color space Among them, μ S , and The color values ​​of the three vertices are respectively denoted as c S =(R S ,G S ,B S ), and

[0031] As a preferred embodiment of the present invention, the prompt points of the Segment Anything 2 model are obtained from the target structure, and the image of the high-voltage transmission line to be detected is segmented according to the prompt points to obtain the regional images corresponding to the components in the high-voltage transmission line to be detected, specifically:

[0032] Select the node with the highest degree from the target structure. If the target structure has no node, select any point on the longest edge of the target structure as a node.

[0033] Set the node's ID to 1, set the IDs of the background points around the node and the key points of the node's corresponding connecting edges to 0, use the IDs as the prompt points of the current component, and call the Segment Anything 2 model to generate the mask of the current component;

[0034] Extract the corresponding part of the image of the high-voltage transmission line to be detected according to the mask and save it as the regional image of the current component;

[0035] The masks generated by each component are traversed in a loop to obtain the regional images corresponding to each component in the high-voltage transmission line to be detected.

[0036] As a preferred implementation of the present invention, depth calculation is performed on the segmented regional image to filter out the component regional image located in the foreground, specifically:

[0037] Calculate the depth of the regional image corresponding to each component, expressed as:

[0038]

[0039] In the formula, For component F i The corresponding regional image depth;

[0040] According to the depth of the regional image corresponding to each component, the maximum depth d of all components is obtained max , filter out the component area image located in the foreground, wherein the component area image located in the foreground corresponds to the depth

[0041] As a preferred implementation of the present invention, the color values ​​of the pixels in the screened component region image are projected onto the color space triangle to obtain the corresponding projection points, specifically:

[0042] Along c S Towards and Draw a perpendicular line that intersects point c d , is the rust change direction corresponding to the component area image pixel in the color space triangle, which is expressed as:

[0043]

[0044] Compute color space triangle The normal vector The formula is:

[0045]

[0046] According to the normal vector Calculate the color value c of any pixel in the component area image p In color space triangle Projection point of the plane The formula is:

[0047]

[0048] According to the projection point and the direction of rust change Get the corresponding projection point cq , expressed as:

[0049]

[0050] As a preferred embodiment of the present invention, the quantitative value w of the corrosion degree of the corresponding projection point is calculated. p The formula is:

[0051]

[0052] As a preferred embodiment of the present invention, the evaluation parameters in the component region image are counted, and the evaluation parameters include the average corrosion degree, the maximum corrosion degree and the corrosion degree ratio of all pixels, which are expressed as follows:

[0053]

[0054] In the formula, is the average degree of corrosion, w max is the maximum corrosion degree, w(v) is the corrosion degree ratio, N is the component F i The number of pixels, v is the degree of corrosion.

[0055] As a preferred embodiment of the present invention, the rust level of the component corresponding to the component region image is evaluated according to the evaluation parameters as follows:

[0056] The preset corrosion grade and the range of the corresponding corrosion degree quantitative value are as follows, wherein the corrosion grade includes light corrosion, moderate corrosion and severe corrosion, wherein the light corrosion is divided into grade 1 corrosion and grade 2 corrosion, the moderate corrosion is divided into grade 3 corrosion and grade 4 corrosion, and the severe corrosion is divided into grade 5 corrosion and grade 6 corrosion, wherein:

[0057] When 0≤maximum corrosion degree quantitative value<0.3, 0<average corrosion degree quantitative value<0.1, the component is in level 1 corrosion; when 0≤maximum corrosion degree quantitative value<0.3, 0.1≤average corrosion degree quantitative value<0.2, the component is in level 2 corrosion;

[0058] When the maximum rust degree quantitative value is ≥0.3 and the rust degree percentage quantitative value is <3%, the component is at level 3 rust;

[0059] When the maximum rust degree quantitative value is ≥ 0.3, and the rust degree percentage quantitative value is 3%≤<10%, the component is at level 4 rust;

[0060] When the maximum rust degree quantitative value is ≥0.3 and the rust degree percentage quantitative value is 10%≤<33%, the component is at level 5 rust;

[0061] When the maximum rust degree quantitative value is ≥ 0.3 and 33% ≤ the rust degree percentage quantitative value, the component is at level 6 rust.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] 1. The present invention provides a method for detecting corrosion of steel components of power transmission lines. The Depth Anything V2 model is used to extract the image foreground, so that the subsequent corrosion detection is more targeted and the influence of the image background on the detection results is reduced. The prompt points of the Segment Anything 2 model are generated based on the skeleton generation technology, which improves the accuracy of component segmentation.

[0064] 2. The present invention provides a method for detecting corrosion of steel parts of power transmission lines, which classifies the completely corroded pixel set through the K-means clustering algorithm, accurately defines the color space and performs further analysis, so that the corrosion detection is more detailed and reliable;

[0065] 3. The present invention provides a method for detecting rust of steel components of power transmission lines. The method performs depth calculation on the segmented regional images, screens out the regional images of foreground components, further projects the pixel color values ​​of these regional images into the color space triangle, quantitatively evaluates the degree of rust, and determines the rust grade based on the evaluation parameters. The method not only quantifies the degree of rust, but also can perform a comprehensive evaluation based on multiple dimensions such as the proportion of rust, average rust and maximum rust, making the rust detection more comprehensive and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 is a method flow chart of an embodiment of the present invention;

[0067] Figure 2 Schematic diagram of a color space triangle according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0069] The present invention provides the following technical solution: a method for detecting corrosion of steel components of a power transmission line.

[0070] Embodiment 1:

[0071] S1. Obtain a corrosion image of a high-voltage transmission line, remove the background and distant view in the corrosion image of the high-voltage transmission line based on the Depth Anything V2 model, and obtain a corresponding close-up image;

[0072] Preferably, the depth d corresponding to each pixel p of the high-voltage transmission line image is calculated p , sort the depths corresponding to all pixels of the high-voltage transmission line image from large to small, regard the pixels in the last 50% of the sorting as the background, and obtain the close-up image after removing the background and the corresponding rust pixel set P;

[0073] S2, manually annotating the rust in the close-up image according to the degree of rust to obtain annotated data; obtaining a non-corroded pixel set and a completely corroded pixel set according to the annotated data, dividing the completely corroded pixel set into two non-intersecting completely corroded pixel subsets of different types using a Kmean clustering algorithm, calculating and obtaining the center points of the non-corroded pixel set and the completely corroded pixel subset, and defining a color space triangle according to the center points;

[0074] S21. Manual labeling:

[0075] The rust degree ranges from [0,1], and the rust degree w is marked for the pixels in the area without rust in the close-up image. p The corresponding value is 0; the most severely corroded area in the close-up image contains pixels with the degree of corrosion w p The corresponding value is 1; obtain the annotation data corresponding to the rust pixel set P, and the annotation data includes the rust degree w corresponding to the pixel p and color value coordinate c p =(R p ,G p ,B p );

[0076] Preferably, when manually marking, the area selection can be any closed shape such as polygons and circles, and there is no need to mark all the rusted areas without corrosion, just select a part of them for marking; similar to the areas without corrosion, select the areas with the most serious corrosion in the image for marking;

[0077] S22, non-corroded pixel set and completely corroded pixel set:

[0078] The rust on the high-voltage line is mainly composed of ferrous oxide (black), ferric oxide (reddish brown), ferric oxide (black) and ferric hydroxide (reddish brown). Therefore, the severely corroded area usually appears black or reddish brown. In this embodiment, in order to increase the detection accuracy, the K-Means algorithm is used to distinguish black from reddish brown, and K=2 is set. T is clustered, and the uncorroded pixel set S and the completely corroded pixel set T are obtained according to the corroded pixel set P, which is expressed as follows:

[0079] S={p∈Pw p =0};

[0080] T={p∈Pw p =1};

[0081] Calculate the red component R corresponding to any pixel in the completely corroded pixel set T p The mean The formula is:

[0082]

[0083] Where |T| represents the number of pixels in the completely corroded pixel set T;

[0084] The completely corroded pixel set T is divided into two groups according to the red component R p Divide into subset T1 0 and The formula is:

[0085]

[0086] Calculate the subset T1 0 and The center point and The formula is:

[0087]

[0088] Re-divide the subsets according to the center point, and iterate the division process until convergence, which can be expressed as:

[0089]

[0090] Where i is the final number of iterations, 0≤i≤N;

[0091] Get subset T1 N and The corresponding center points are and

[0092] For the uncorroded pixel set S and the completely corroded pixel subset T1 Nand The reduction is performed by repeatedly calculating the center point of the set and deleting the kS| points in the set that are farthest from the center point, where k is the reduction ratio. In this embodiment, k∈(0,0.5); the set reduction process is repeated until the center point of the set does not change, and the final uncorroded pixel set S' and its corresponding center point μ are obtained. S , the completely corroded pixel subsets T1' and T2' and their corresponding center points and

[0093] S23, color space triangle:

[0094] The center point μ S , and As the three vertices of the triangle, get the color space triangle in color space Among them, μ S , and The color values ​​of the three vertices are respectively denoted as c S =(R S ,G S ,B S ), and

[0095] S3, obtaining an image of a high-voltage transmission line to be detected, removing the background and distant view in the image of the high-voltage transmission line to be detected based on the Depth Anything V2 model, and obtaining a close-up image to be detected; processing the close-up image to be detected using the skeleton generation technology to obtain the target structure, obtaining the prompt points of the Segment Anything 2 model from the target structure, performing component segmentation on the image of the high-voltage transmission line to be detected according to the prompt points, and obtaining the regional images corresponding to the various components in the high-voltage transmission line to be detected; performing depth calculation on the segmented regional images, and screening out the component regional images located in the foreground;

[0096] S31, the step of obtaining the near-view image to be detected is the same as the step in step S1, and will not be repeated here;

[0097] S32, component segmentation:

[0098] Select the node with the highest degree from the target structure. If the target structure has no node, select any point in the longest edge of the target structure as the node. Preferably, in this embodiment, if there are multiple nodes with the same highest degree, a node can be randomly selected.

[0099] The node identifier is set to 1, the identifiers of the background points around the node and the key points of the node's corresponding connection edges are set to 0, the identifier is used as the prompt point of the current component, and the Segment Anything 2 model is called to generate the mask of the current component; preferably, in this implementation, the node is surrounded by a neighborhood with a certain radius preset for the node;

[0100] Extract the corresponding part of the image of the high-voltage transmission line to be detected according to the mask and save it as the regional image of the current component;

[0101] The masks generated by each component are looped through to obtain the regional images corresponding to each component in the high-voltage transmission line to be detected. If the generated mask overlaps with the masks of other components, the overlapping parts are assigned to the corresponding components according to the principle of proximity.

[0102] S33, component area image located in the foreground:

[0103] Calculate the depth of the regional image corresponding to each component, expressed as:

[0104]

[0105] In the formula, For component F i The corresponding regional image depth;

[0106] According to the depth of the regional image corresponding to each component, the maximum depth d of all components is obtained max , filter out the component area image located in the foreground, wherein the component area image located in the foreground corresponds to the depth

[0107] S4, projecting the color values ​​of the pixels in the screened component area image to the color space triangle to obtain the corresponding projection point; calculating the quantized value of the degree of corrosion of the corresponding projection point, and counting the evaluation parameters in the component area image, the evaluation parameters including the average degree of corrosion, the maximum degree of corrosion and the proportion of corrosion of all pixels, and evaluating the corrosion grade of the component corresponding to the component area image according to the evaluation parameters;

[0108] S41, such as Figure 2 As shown, along c S Towards and Draw a perpendicular line that intersects point c d , is the rust change direction corresponding to the component area image pixel in the color space triangle, which is expressed as:

[0109]

[0110] Compute color space triangle The normal vector That is, calculate the cross product of the two sides of the triangle formed by the color value, expressed as:

[0111]

[0112] According to the normal vector Calculate the color value c of any pixel in the component area image p In color space triangle Projection point of the plane The formula is:

[0113]

[0114] According to the projection point and the direction of rust change Get the corresponding projection point c q , expressed as:

[0115]

[0116] If the corresponding projection point c q If the value is close to 0, it means that it is closer to no corrosion; if the value is close to 1, it means that the degree of corrosion is relatively serious;

[0117] S42. Calculate the quantitative value w of the degree of corrosion at the corresponding projection point p The formula is:

[0118]

[0119] S43, the evaluation parameters include the average corrosion degree, the maximum corrosion degree and the corrosion degree ratio of all pixels, which are expressed as follows:

[0120]

[0121] In the formula, is the average degree of corrosion, w max is the maximum corrosion degree, w(v) is the corrosion degree ratio, N is the component F i The number of pixels, v is the degree of corrosion;

[0122] S44. Based on the ASTM D610 international standard, preset corrosion grades and their corresponding ranges of quantitative values ​​of corrosion degrees, wherein the corrosion grades include mild corrosion, moderate corrosion and severe corrosion, wherein mild corrosion is divided into grade 1 corrosion and grade 2 corrosion, moderate corrosion is divided into grade 3 corrosion and grade 4 corrosion, and severe corrosion is divided into grade 5 corrosion and grade 6 corrosion, wherein:

[0123] When 0≤maximum corrosion degree quantitative value<0.3, 0<average corrosion degree quantitative value<0.1, the component is in level 1 corrosion; when 0≤maximum corrosion degree quantitative value<0.3, 0.1≤average corrosion degree quantitative value<0.2, the component is in level 2 corrosion;

[0124] When the maximum rust degree quantitative value is ≥0.3 and the rust degree percentage quantitative value is <3%, the component is at level 3 rust;

[0125] When the maximum rust degree quantitative value is ≥ 0.3, and the rust degree percentage quantitative value is 3%≤<10%, the component is at level 4 rust;

[0126] When the maximum rust degree quantitative value is ≥0.3 and the rust degree percentage quantitative value is 10%≤<33%, the component is at level 5 rust;

[0127] When the maximum rust degree quantitative value is ≥ 0.3 and 33% ≤ the rust degree percentage quantitative value, the component is at level 6 rust.

[0128] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for detecting corrosion of steel components of a power transmission line, characterized in that: The method comprises: Obtain a corrosion image of a high-voltage transmission line, remove the background and distant view in the corrosion image of the high-voltage transmission line based on the Depth Anything V2 model, and obtain the corresponding close-up image; Manually annotate the rust in the close-up image according to the degree of rust to obtain annotated data; obtain a non-corroded pixel set and a completely corroded pixel set according to the annotated data, divide the completely corroded pixel set into two non-intersecting completely corroded pixel subsets of different types using a Kmean clustering algorithm, calculate and obtain the center points of the non-corroded pixel set and the completely corroded pixel subset, and define a color space triangle according to the center points; Acquire an image of a high-voltage transmission line to be detected, remove the background and distant view in the image of the high-voltage transmission line to be detected based on the Depth Anything V2 model, and obtain a close-up image to be detected; process the close-up image to be detected using a skeleton generation technique to obtain a target structure, obtain prompt points of the Segment Anything2 model from the target structure, perform component segmentation on the image of the high-voltage transmission line to be detected according to the prompt points, and obtain regional images corresponding to various components in the high-voltage transmission line to be detected; perform depth calculation on the segmented regional images, and filter out the component regional images located in the foreground; The color values ​​of the pixels in the screened component area image are projected onto the color space triangle to obtain the corresponding projection point; the quantified value of the degree of rust at the corresponding projection point is calculated, and the evaluation parameters in the component area image are counted, the evaluation parameters including the average degree of rust, the maximum degree of rust and the proportion of rust of all pixels, and the rust grade of the component corresponding to the component area image is evaluated according to the evaluation parameters.

2. A method for detecting corrosion of steel components of a power transmission line according to claim 1, characterized in that: Based on the Depth Anything V2 model, the background and foreground in the image of high-voltage transmission line corrosion are removed to obtain a close-up image. Specifically, the depth d corresponding to each pixel p in the image of the high-voltage transmission line is calculated. p , sort the depths corresponding to all pixels of the high-voltage transmission line image from large to small, regard the pixels in the last 50% of the sorting as the background, and obtain the close-up image after removing the background and the corresponding rust pixel set P.

3. A method for detecting corrosion of steel components of a power transmission line according to claim 2, characterized in that: The rust in the close-up image is manually annotated according to the degree of rust to obtain annotated data, wherein: The rust degree ranges from [0,1], and the rust degree w is marked for the pixels in the area without rust in the close-up image. p The corresponding value is 0; the most severely corroded area in the close-up image contains pixels with the degree of corrosion w p The corresponding value is 1; obtain the annotation data corresponding to the rust pixel set P, and the annotation data includes the rust degree w corresponding to the pixel p and color value coordinate c p =(R p ,G p ,B p ).

4. A method for detecting corrosion of steel components of a power transmission line according to claim 3, characterized in that: The center points of the uncorroded pixel set and the completely corroded pixel subset are calculated, and a color space triangle is defined according to the center points, specifically: According to the corroded pixel set P, the non-corroded pixel set S and the completely corroded pixel set T are obtained, which can be expressed as follows: S={p∈Pw p =0}; T={p∈Pw p =1}; Calculate the red component R corresponding to any pixel in the completely corroded pixel set T p The mean The formula is: Where |T| represents the number of pixels in the completely corroded pixel set T; The completely corroded pixel set T is divided into two groups according to the red component R p Divide into subsets and The formula is: Calculate the subset and The center point and The formula is: Re-divide the subsets according to the center point, and iterate the division process until convergence, which can be expressed as: Where i is the final number of iterations, 0≤i≤N; Get a subset and The corresponding center points are and For the uncorroded pixel set S and the completely corroded pixel subset and The reduction is performed by repeatedly calculating the center point of the set and deleting the kS| points in the set that are farthest from the center point, where k is the reduction ratio; the set reduction process is repeated until the center point of the set does not change, and the final uncorroded pixel set S' and its corresponding center point μ are obtained. S , the completely corroded pixel subsets T1' and T2' and their corresponding center points and The center point μ S , and As the three vertices of the triangle, get the color space triangle in color space Among them, μ S , and The color values ​​of the three vertices are respectively denoted as c S =(R S ,G S ,B S ), and 5. A method for detecting corrosion of steel components of a power transmission line according to claim 4, characterized in that: Obtain the prompt points of the Segment Anything 2 model from the target structure, perform component segmentation on the image of the high-voltage transmission line to be detected according to the prompt points, and obtain the regional images corresponding to the components in the high-voltage transmission line to be detected, specifically: Select the node with the highest degree from the target structure. If the target structure has no node, select any point on the longest edge of the target structure as a node. Set the node's ID to 1, set the IDs of the background points around the node and the key points of the node's corresponding connecting edges to 0, use the IDs as the prompt points of the current component, and call the Segment Anything 2 model to generate the mask of the current component; Extract the corresponding part of the image of the high-voltage transmission line to be detected according to the mask and save it as the regional image of the current component; The masks generated by each component are traversed in a loop to obtain the regional images corresponding to each component in the high-voltage transmission line to be detected.

6. A method for detecting corrosion of steel components of a power transmission line according to claim 5, characterized in that: The depth of the segmented regional image is calculated to filter out the component regional image in the foreground, specifically: Calculate the depth of the regional image corresponding to each component, expressed as: In the formula, For component F i The corresponding regional image depth; According to the depth of the regional image corresponding to each component, the maximum depth d of all components is obtained max , filter out the component area image located in the foreground, wherein the component area image located in the foreground corresponds to the depth 7. A method for detecting corrosion of steel components of a power transmission line according to claim 6, characterized in that: Project the color values ​​of the pixels in the filtered component area image to the color space triangle to obtain the corresponding projection points, specifically: Along c S Towards and Draw a perpendicular line that intersects point c d , is the rust change direction corresponding to the component area image pixel in the color space triangle, which is expressed as: Compute color space triangle The normal vector The formula is: According to the normal vector Calculate the color value c of any pixel in the component area image p In color space triangle Projection point of the plane The formula is: According to the projection point and the direction of rust change Get the corresponding projection point c q , expressed as:

8. A method for detecting corrosion of steel components of a power transmission line according to claim 7, characterized in that: Calculate the quantitative value w of the corrosion degree corresponding to the projection point p The formula is:

9. A method for detecting corrosion of steel components of a power transmission line according to claim 8, characterized in that: The evaluation parameters in the component area image are counted, and the evaluation parameters include the average corrosion degree, the maximum corrosion degree and the corrosion degree ratio of all pixels, which are expressed as follows: In the formula, is the average degree of corrosion, w max is the maximum corrosion degree, w(v) is the corrosion degree ratio, N is the component F i The number of pixels, v is the degree of corrosion.

10. A method for detecting corrosion of steel components of a power transmission line according to claim 9, characterized in that: The corrosion level of the component corresponding to the component region image is evaluated according to the evaluation parameters as follows: The preset corrosion grade and the range of the corresponding corrosion degree quantitative value are as follows, wherein the corrosion grade includes light corrosion, moderate corrosion and severe corrosion, wherein the light corrosion is divided into grade 1 corrosion and grade 2 corrosion, the moderate corrosion is divided into grade 3 corrosion and grade 4 corrosion, and the severe corrosion is divided into grade 5 corrosion and grade 6 corrosion, wherein: When 0≤maximum corrosion degree quantitative value<0.3, 0<average corrosion degree quantitative value<0.1, the component is in level 1 corrosion; when 0≤maximum corrosion degree quantitative value<0.3, 0.1≤average corrosion degree quantitative value<0.2, the component is in level 2 corrosion; When the maximum rust degree quantitative value is ≥0.3 and the rust degree percentage quantitative value is <3%, the component is at level 3 rust; When the maximum rust degree quantitative value is ≥ 0.3, and the rust degree percentage quantitative value is 3%≤<10%, the component is at level 4 rust; When the maximum rust degree quantitative value is ≥0.3 and the rust degree percentage quantitative value is 10%≤<33%, the component is at level 5 rust; When the maximum rust degree quantitative value is ≥ 0.3 and 33% ≤ the rust degree percentage quantitative value, the component is at level 6 rust.

Images