A method and system for detecting wear of coating on fan blade surface

Abstract

The present application relates to the field of image processing technology, and specifically to a method and system for detecting wear of a coating on a wind turbine blade surface. The method comprises: obtaining a color wind turbine blade image, performing noise reduction processing, and converting the noise-reduced image into a grayscale image; obtaining edge pixel points and connected domains of each grayscale image; respectively obtaining the first eigenvalue, second eigenvalue, and third eigenvalue of each connected domain; obtaining the wear feature significance value of each connected domain; clustering all connected domains; obtaining the wear significance coefficient of each connected domain; obtaining the neighborhood radius for performing LBP feature extraction on the pixels of each connected domain to calculate the LBP eigenvalue of each pixel; and identifying the wear area and wear category in the wind turbine blade image. The present application improves the detection accuracy of the model for wear of a coating on a wind turbine blade surface by adaptively adjusting the neighborhood radius when calculating the LBP eigenvalue.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method and system for detecting wear of coating on the surface of a wind turbine blade. Background Art

[0002] With the booming wind power industry, wind turbine blades, as key components of wind turbines, face a crucial impact on their performance and service life. Wind turbine blades are constantly exposed to complex natural environments, such as strong winds, dust, ultraviolet rays, and drastic changes in temperature and humidity. This makes the blade surface coating extremely susceptible to wear and tear. Once the blade surface coating wears, it not only reduces the blade's aerodynamic performance, resulting in a decrease in wind energy conversion efficiency and increased power generation costs, but also directly exposes the blade substrate to harsh environments, accelerating corrosion and aging, shortening the blade's service life, and even potentially causing serious safety incidents such as blade breakage, posing a significant threat to the safe and stable operation of wind farms.

[0003] Due to the special shape of fan blades, wear often occurs at the blade edges, and the direction of wear extension is uncertain, resulting in varying sizes of wear areas. When using the LBP (Local Binary Pattern) feature extraction algorithm for feature extraction, feature extraction is performed based on a pre-set local window. This local window may not contain the entire wear area, or may contain a large number of non-wear areas. It also ignores the important influence of the surface features of fan blade wear on wear feature extraction, making it impossible for the extracted features to accurately and comprehensively reflect the wear condition of the fan blades. This results in large deviations in feature extraction and a low accuracy rate in fan blade surface wear detection. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a method and system for detecting wear of the coating on the surface of a wind turbine blade. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a method for detecting wear of a coating on a surface of a wind turbine blade, the method comprising the following steps:

[0006] Obtain a color image of a wind turbine blade, perform noise reduction processing on it, and convert the noise-reduced image into a grayscale image;

[0007] Obtain edge pixels and connected domains of each grayscale image; obtain a first eigenvalue of each connected domain based on the length and width of the minimum circumscribed rectangle of each connected domain; obtain a second eigenvalue of each connected domain based on the number of pixels in each connected domain and the number of pixels in the corresponding minimum circumscribed rectangle; obtain a third eigenvalue of each connected domain based on the number of pixels that overlap with the border of the corresponding minimum circumscribed rectangle and the number of all edge pixels of the corresponding connected domain; obtain a wear feature significance value corresponding to each connected domain based on the first eigenvalue, second eigenvalue, and third eigenvalue of each connected domain;

[0008] All connected domains are clustered according to their first, second, and third eigenvalues; the wear significance coefficient of each connected domain is obtained based on the wear characteristic significance value of each connected domain and the difference between the wear characteristic significance value of each connected domain and the wear characteristic significance value of all connected domains in its cluster;

[0009] The wear significance coefficient of each connected domain is used to correct the initial neighborhood radius when extracting LBP features from the pixels of each connected domain, so as to calculate the LBP eigenvalue of each pixel. The wear area and wear category in the wind blade image to be detected are identified based on the LBP eigenvalues of all pixels in each grayscale image and all denoised color wind blade images.

[0010] Preferably, the first characteristic value of each connected domain is the aspect ratio of the minimum circumscribed rectangle of each connected domain.

[0011] Preferably, the second characteristic value of each connected domain is a ratio of the number of pixels in each connected domain to the number of pixels in the corresponding minimum circumscribed rectangle.

[0012] Preferably, the method for determining the third eigenvalue of each connected domain is: obtaining the number of pixel points that overlap with the corresponding minimum circumscribed rectangular border of each connected domain, and recording the ratio of the number to the number of all edge pixels in the connected domain as the third eigenvalue of each connected domain.

[0013] Preferably, the calculation formula for the wear characteristic significance value of each connected domain is: i =a i ×b i ×c i Where x i represents the significant value of the wear feature of the i-th connected domain; a i 、b i and c i represents the first eigenvalue, second eigenvalue, and third eigenvalue of the i-th connected component respectively.

[0014] Preferably, the calculation formula for the wear significance coefficient of each connected domain is: Where, Li represents the wear significance coefficient of the i-th connected domain; x i represents the significant value of the wear feature of the i-th connected domain; x v n represents the wear feature significance value of the vth connected domain in the cluster where the i-th connected domain is located; i represents the number of all connected domains in the cluster where the i-th connected domain is located, and ω is a first preset constant.

[0015] Preferably, the calculation formula for the neighborhood radius after the initial neighborhood radius is corrected when performing LBP feature extraction on the pixels of each connected domain is: i =r0+round(δ×l i );where r i It represents the neighborhood radius after correction when LBP feature extraction is performed on each pixel point in the i-th connected domain, r0 represents the preset initial neighborhood radius, l i represents the normalized result of the wear significance coefficient of the i-th connected domain, δ is the second preset constant, and round() is used to round up the input data.

[0016] Preferably, the specific process of identifying the wear area and wear category in the wind blade image to be detected based on the LBP feature values of all pixels in each grayscale image and all denoised color wind blade images is:

[0017] Obtain the LBP feature map of each grayscale image according to the LBP feature values of all pixels in each grayscale image;

[0018] Obtain all spliced fan blade images based on the LBP feature map of each grayscale image and each denoised color fan blade image;

[0019] All the spliced wind blade images are used as input for training of the neural network model to obtain a wear detection model, which is then used to identify the wear areas and wear categories in the wind blade images to be inspected.

[0020] Preferably, the acquisition process of all the spliced fan blade images is: splicing the single-channel LBP feature map of each grayscale image with the corresponding denoised three-channel color fan blade image in the channel dimension to obtain all the spliced fan blade images.

[0021] In the second aspect, an embodiment of the present application also provides a system for detecting wear of the surface coating of a wind blade, comprising a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned methods for detecting wear of the surface coating of a wind blade are implemented.

[0022] This application has at least the following beneficial effects:

[0023] This application firstly reduces the influence of environmental interference on the image feature analysis of the wind turbine blade surface through image acquisition and image noise reduction processing; in the subsequent processing process, by deeply studying the inward extension of the wear edge and the range difference of the coating wear, it can accurately analyze the coating wear characteristics of different local areas; on this basis, the relevant parameters during feature extraction are optimized and adjusted, and then the characteristic values that can accurately characterize the wear characteristics of the wind turbine blades are obtained. Compared with the traditional local area feature analysis method, the method proposed in this application fully considers the uncertainty of the wear extension direction, and can adaptively adjust the neighborhood radius of each pixel point when extracting the wear characteristics of the wind turbine blade surface, so that the extracted features more accurately reflect the wear characteristics of the wind turbine surface; the extracted wear features are fused with the wind turbine blade image, so that the trained wear detection model can more accurately detect the wear area on the wind turbine blade surface, improve the detection accuracy of the model, and provide a strong guarantee for the safe and stable operation of wind power generation equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0025] Figure 1 A flowchart of a method for detecting wear of a coating on a surface of a wind turbine blade provided in one embodiment of the present application;

[0026] Figure 2 A flowchart for obtaining the corrected neighborhood radius of pixels of each connected domain when performing LBP feature extraction is provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] To further illustrate the technical means and effectiveness of this application to achieve the intended invention objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a method and system for detecting wear of a coating on a wind turbine blade surface, including its specific implementation, structure, features, and effectiveness. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0028] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0029] The following describes in detail a method and system for detecting surface coating wear of wind turbine blades provided by the present application with reference to the accompanying drawings.

[0030] See also Figure 1 , which shows a flowchart of a method for detecting wear of a surface coating of a wind turbine blade provided by one embodiment of the present application, the method comprising the following steps:

[0031] Step 1: Obtain a color wind turbine blade image, perform noise reduction processing, and convert the noise-reduced image into a grayscale image.

[0032] DJI's Matrice 300RTK drone, equipped with the Zenmuse H20N camera, boasts a maximum flight time of 55 minutes and is adaptable to varying weather conditions, capable of stable flight in force 6 winds. The Zenmuse H20N integrates a 20-megapixel wide-angle camera, a 12-megapixel zoom camera, and a thermal imaging camera, enabling it to capture high-quality images of wind turbine blades in a variety of lighting conditions.

[0033] With the wind turbines shut down, the drone's flight path was planned based on the turbine's location and layout. The drone's flight altitude was set at 10-15 meters and its speed at 23 meters per second to ensure the camera could clearly capture every part of the blades. The captured images of the wind turbine blades were color RGB images. During the acquisition process, multi-angle photography was used, capturing each blade from different angles, including the front, side, and top, to obtain more comprehensive wear information.

[0034] Considering that the fan blade images collected under the external environment are subject to the influence of large interference factors, and the wear and tear of the fan blades are usually at the edges of the fan blades, it is necessary to combine the spatial proximity of the image and the similarity of pixel values to better retain the edge information while smoothing the image; and for the noise generated by uneven illumination in the fan blade image, it is necessary to maintain the texture characteristics of the blades at the same time when reducing the noise. Therefore, in this embodiment, a bilateral filtering algorithm is used to reduce the noise of the fan blade image to obtain a fan blade image after noise reduction. The bilateral filtering algorithm is a well-known technology, and the specific process will not be repeated here. Furthermore, the denoised image is grayscaled to convert the color image into a grayscale image to reduce the amount of data.

[0035] Step 2: Obtain the edge pixel points and connected domains of each grayscale image; obtain the first eigenvalue of each connected domain based on the length and width of the minimum circumscribed rectangle of each connected domain; obtain the second eigenvalue of each connected domain based on the number of pixels in each connected domain and the number of pixels in the corresponding minimum circumscribed rectangle; obtain the third eigenvalue of each connected domain based on the number of pixels that overlap with the border of the corresponding minimum circumscribed rectangle and the number of all edge pixels of the corresponding connected domain; obtain the wear feature significance value corresponding to each connected domain based on the first eigenvalue, second eigenvalue and third eigenvalue of each connected domain.

[0036] Traditional wind blade surface image analysis relies heavily on localized feature analysis. However, due to the unique shape of wind blades, wear often occurs at the blade edges, and the direction of wear is uncertain, resulting in varying wear areas. In this scenario, if traditional localized feature analysis methods are used, the wear area will likely be divided into different local regions for feature extraction. This will result in significant deviations in the acquired wind blade surface features, severely impacting the accuracy of wind blade surface wear detection.

[0037] Currently, commonly used image feature extraction methods include multi-scale fusion feature extraction, local binary patterns, and gradient histograms. However, these methods are based on pre-set local window expansion analysis and fail to fully consider the important influence of wind turbine blade wear surface characteristics on wear feature analysis.

[0038] Considering that the wear of wind turbine blades is mainly manifested as leading edge erosion, surface coating shedding and sand erosion, and usually extends from the edge of the blade to the inside. In this embodiment, the grayscale image obtained after preprocessing is used as input data, and the edge detection algorithm is first used to obtain the edge pixel points of the wind turbine blade image. The edge detection algorithm includes but is not limited to the Canny edge detection algorithm and the Sobel edge detection algorithm, which are not limited in this embodiment. Then, the connected domain analysis algorithm is used to obtain the connected domain in the grayscale image. The connected domain analysis algorithm classifies the interconnected pixels into the same connected domain based on the connectivity of the pixels, thereby separating different objects or areas in the image. The connected domain analysis algorithm is a well-known technology, and the specific process will not be repeated.

[0039] Since the wear of wind turbine blades has the characteristic of extending from the edge to the inside, for each connected domain, it is necessary to obtain its minimum circumscribed rectangle. Calculate the aspect ratio of the minimum circumscribed rectangle of each connected domain, and use the aspect ratio as the first eigenvalue of the corresponding connected domain. The larger the first eigenvalue, the more significant the extension feature of the range where the connected domain is located. At the same time, calculate the ratio of the number of pixels in each connected domain to the number of pixels in the corresponding minimum circumscribed rectangle, and use the ratio as the second eigenvalue of the corresponding connected domain. The smaller the second eigenvalue, the smaller the extension range of the range where the connected domain is located. By analyzing these features, the wear condition of the wind turbine blades can be judged more accurately.

[0040] In order to further reduce the interference of the shape of the wind turbine blades on the wear feature analysis, it is considered that the wear of the wind turbine blades usually extends from the edge to the inside, that is, one side of the edge where the wear area is located is relatively flat, while the shape of the pixels on the other side is quite different. Based on this, when processing the image, the following operations are performed on the connected domain: the pixels in the connected domain that coincide with the minimum circumscribed rectangular border are marked, the number of these marked pixels is counted, and the ratio of the number to the number of all edge pixels in the connected domain is calculated, and the ratio is used as the third eigenvalue of the corresponding connected domain. In a physical sense, the larger the third eigenvalue, the more likely the area where the connected domain is located is to be an area with more significant wear features.

[0041] Furthermore, as a preferred embodiment, the wear feature significance value corresponding to each connected domain is obtained according to the first eigenvalue, the second eigenvalue and the third eigenvalue of each connected domain, so as to characterize the significance of the wear feature in each connected domain.

[0042] In this embodiment, the wear feature significance value of the i-th connected domain is recorded as x i , its specific expression is: i =a i ×b i ×c i Where x i represents the significant value of the wear feature of the i-th connected domain; a i 、b i and c i The larger the wear feature significance value of the connected domain, the more significant the regional characteristics caused by wear in the area where the current connected domain is located.

[0043] Step 3: Cluster all connected domains according to the first eigenvalue, second eigenvalue, and third eigenvalue of each connected domain; obtain the wear significance coefficient of each connected domain based on the wear feature significance value of each connected domain and the difference between the wear feature significance value of each connected domain and the wear feature significance value of all connected domains in its cluster.

[0044] The first eigenvalue, second eigenvalue, and third eigenvalue of each connected domain calculated in the above steps together constitute the feature list of each connected domain. In order to analyze the wear characteristics more deeply, it is necessary to compare the features of all connected domains. At this time, the feature lists corresponding to all connected domains are used as input data, and the connected domains corresponding to all feature lists are divided into multiple clusters using the agglomerative hierarchical clustering algorithm. During the clustering process, the Manhattan distance is calculated to determine the similarity between the feature lists, and then the connected domains with similar wear characteristics are clustered into one category, so that the wear of the wind turbine blades can be analyzed and determined more efficiently in the future. The agglomerative hierarchical clustering algorithm is a well-known technology, and the specific process will not be repeated here.

[0045] Furthermore, based on the similarity of the feature vectors reflecting the overall wear characteristics in the current cluster, the wear characteristic significance of the local area of the wind turbine blade surface is further accurately reflected. Specifically, as a preferred embodiment, the wear characteristic significance value of each connected domain is obtained based on the wear characteristic significance value of each connected domain and the difference between the wear characteristic significance value of each connected domain and the wear characteristic significance value of all connected domains in the cluster where it is located. In this embodiment, the wear significance coefficient of the i-th connected domain is denoted as L i , its specific expression is: Where, L i represents the wear significance coefficient of the i-th connected domain; x i represents the significant value of the wear feature of the i-th connected domain; x v n represents the wear feature significance value of the vth connected domain in the cluster where the i-th connected domain is located; i represents the number of all connected domains in the cluster where the i-th connected domain is located, ω is the first preset constant used to prevent the denominator from being 0, and can be any positive number less than 0.01. In this embodiment, it is 0.001. The calculated wear significant coefficient L i The larger the value, the more significant the wear characteristics in the connected domain. Further analysis of the differences in wear characteristic significance values between connected domains after feature vector clustering aims to accurately determine the difference in wear significance between the local area where the connected domain is located and other similar connected domains in the cluster based on the overall difference, thereby accurately analyzing the significance of wear characteristics of different connected domains.

[0046] Step 4: Use the wear significance coefficient of each connected domain to correct the initial neighborhood radius when performing LBP feature extraction on the pixel points of each connected domain to calculate the LBP eigenvalue of each pixel point; based on the LBP eigenvalues of all pixels in each grayscale image and all denoised color wind blade images, identify the wear area and wear category in the wind blade image to be detected.

[0047] Furthermore, based on the calculated wear significance coefficient of each connected domain, the neighborhood radius of the LBP (Local Binary Pattern) feature extraction for the pixels in each connected domain is adjusted. Specifically, if the wear characteristics in the local area where the connected domain is located are more significant, a larger neighborhood radius for LBP feature extraction should be set in the corresponding connected domain to extract more features. In this embodiment, the neighborhood radius after the LBP feature extraction for each pixel in the i-th connected domain is corrected is denoted as r i , the specific calculation formula is: i =r0+round(δ×l i );where r i represents the neighborhood radius after correction when LBP feature extraction is performed on each pixel point in the i-th connected domain, r0 represents the preset initial neighborhood radius, the size is 1-3, and is set to 1 in this embodiment; l i represents the wear significance coefficient L of the i-th connected domain i The normalized result of δ is the second preset constant. To avoid the area radius being too large, δ is set to 10 in this embodiment. Round() is used to round up the input data. The flowchart for obtaining the corrected neighborhood radius of each connected area pixel point during LBP feature extraction is as follows: Figure 2 As shown. Further, the wear features in the grayscale image are extracted based on the calculated neighborhood radius to obtain the LBP feature value of each pixel in the grayscale image. The LBP feature extraction algorithm is a well-known technology and the specific process will not be repeated here.

[0048] Furthermore, based on the LBP eigenvalue of each pixel in the grayscale image, an LBP feature map of the same size as the grayscale image is obtained, and the feature map accurately reflects the local features of each pixel in the grayscale image. Check whether the data types of the LBP feature map and the color fan blade image are consistent. If not, they need to be converted to the same data type. Normally, the data type of the image is an unsigned 8-bit integer (uint8). If the data type of the LBP feature map is different from that of the color fan blade image, the data type of the LBP feature map is converted to the same data type as the color fan blade image to ensure compatibility with subsequent stitching operations.

[0049] Since the captured wind blade image is a color RGB image with three channels, the single-channel LBP feature map and the denoised three-channel color wind blade image are spliced in the channel dimension to obtain a spliced wind blade image. The number of channels of the spliced image is 4. The spliced wind blade image contains both the color information of the wind blade image and the local feature information of the LBP feature map.

[0050] In this example, the ResNet model in the convolutional neural network is selected for training. The ResNet model solves the gradient vanishing problem in deep neural networks by introducing residual blocks. It can train deeper networks and thus learn more complex wear characteristics of wind turbine blade surfaces. The specific process is as follows:

[0051] (1) Data division: The spliced wind turbine blade images are used as the dataset and divided into training set, validation set, and test set in a ratio of 7:2:1 or 8:1:1.

[0052] (2) Model training: The ResNet model is trained using the training set, with appropriate learning rates (usually set between 0.0001 and 0.1, with an initial learning rate of 0.0001 in this embodiment), batch sizes (usually set between 16 and 128, with a batch size of 128 in this embodiment), and number of training rounds (usually set between 50 and 300, with a batch size of 150 in this embodiment) set to obtain a trained wear detection model. During training, the cross-entropy loss function is used as the loss function, and the Adam optimizer is used as the optimizer. The model is evaluated using the validation set, and the model parameters are updated based on the evaluation results.

[0053] (3) Model testing: Use the test set to test the trained wear detection model, calculate the model's accuracy, recall rate, F1 value and other indicators, and evaluate the model's performance.

[0054] (4) Recognition result output: The model outputs the wear area and wear category (including but not limited to leading edge wear and coating loss) in the image of the wind turbine blade to be inspected. At the same time, the recognition results are displayed in a visual manner, marking the wear area on the original image and providing an assessment of the severity of the wear.

[0055] Based on the same inventive concept as the above-mentioned method, an embodiment of the present application also provides a system for detecting wear of the surface coating of a wind blade, comprising a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned methods for detecting wear of the surface coating of a wind blade are implemented.

[0056] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0057] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0058] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for detecting wear of a fan blade surface coating, characterized in that: The method comprises the following steps: Obtain a color image of a wind turbine blade, perform noise reduction processing on it, and convert the noise-reduced image into a grayscale image; Obtain edge pixels and connected domains of each grayscale image; obtain a first eigenvalue of each connected domain based on the length and width of the minimum circumscribed rectangle of each connected domain; obtain a second eigenvalue of each connected domain based on the number of pixels in each connected domain and the number of pixels in the corresponding minimum circumscribed rectangle; obtain a third eigenvalue of each connected domain based on the number of pixels that overlap with the border of the corresponding minimum circumscribed rectangle and the number of all edge pixels of the corresponding connected domain; obtain a wear feature significance value corresponding to each connected domain based on the first eigenvalue, second eigenvalue, and third eigenvalue of each connected domain; All connected domains are clustered according to the first eigenvalue, second eigenvalue, and third eigenvalue of each connected domain; the average of the differences between the wear characteristic significance value of each connected domain and each wear characteristic significance value of all connected domains in the cluster to which it belongs is calculated, recorded as the first mean, and the wear significance coefficient of each connected domain is obtained according to the wear characteristic significance value of each connected domain and the first mean; The initial neighborhood radius of the pixels of each connected domain when performing LBP feature extraction is corrected using the wear significance coefficient of each connected domain to calculate the LBP feature value of each pixel. The correction process is as follows: a rounded-up result of the product of the normalized wear significance coefficient of each connected domain and a second preset constant is calculated, and the sum of the rounded-up result and the preset initial neighborhood radius is used as the corrected neighborhood radius corresponding to each connected domain; Obtain the LBP feature map of each grayscale image according to the LBP feature values of all pixels in each grayscale image; The single-channel LBP feature map of each grayscale image is spliced with the corresponding denoised three-channel color fan blade image in the channel dimension to obtain the fan blade image after splicing of each grayscale image; The wind blade image obtained by splicing all grayscale images is used as the input of the neural network model for training to obtain a wear detection model, which is then used to identify the wear areas and wear categories in the wind blade image to be inspected.

2. A method for detecting wear of a wind turbine blade surface coating according to claim 1, characterized in that: The first eigenvalue of each connected region is the aspect ratio of the minimum circumscribed rectangle of each connected region.

3. A method for detecting wear of a fan blade surface coating according to claim 2, characterized in that: The second characteristic value of each connected domain is the ratio of the number of pixels in each connected domain to the number of pixels in the corresponding minimum circumscribed rectangle.

4. A method for detecting wear of a wind turbine blade surface coating according to claim 3, characterized in that: The method for determining the third eigenvalue of each connected domain is: obtaining the number of pixel points that overlap with the corresponding minimum circumscribed rectangular frame of each connected domain, and recording the ratio of the number to the number of all edge pixels in the connected domain as the third eigenvalue of each connected domain.

5. A method for detecting wear of a wind turbine blade surface coating according to claim 4, characterized in that: The calculation formula of the wear characteristic significance value of each connected domain is: i =a i ×b i ×c i Where x i represents the significant value of the wear feature of the i-th connected domain; a i 、b i and c i They represent the first eigenvalue, second eigenvalue, and third eigenvalue of the i-th connected component respectively.

6. A method for detecting wear of a wind turbine blade surface coating according to claim 5, characterized in that: The calculation formula of the wear significant coefficient of each connected domain is: Where, L i represents the wear significance coefficient of the i-th connected domain; x i represents the significant value of the wear feature of the i-th connected domain; x v represents the wear feature significance value of the vth connected domain in the cluster where the i-th connected domain is located; n i represents the number of all connected domains in the cluster where the i-th connected domain is located, and ω is a first preset constant.

7. A wind turbine blade surface coating wear detection system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method for detecting wear of the surface coating of a wind turbine blade as described in any one of claims 1 to 6 are implemented.

Images