A UAV-assisted device and method for monitoring surface defects in wind turbine blades

Abstract

This application relates to the field of defect detection technology, specifically to a device and method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicles (UAVs). The method includes: acquiring visible light and infrared images; acquiring each suspected defect region in all visible light images; obtaining a defect significance index for a single suspected defect region based on its edge features, grayscale features, gradient features, and texture features in the visible light image; obtaining an infrared difference coefficient for a single suspected defect region based on its temperature features in the infrared image; obtaining a comprehensive defect significance value for a single suspected defect region; and obtaining a defect significance coefficient for a single suspected defect region based on its dynamic changes during the monitoring period, thereby determining whether a defect exists on the surface of the wind turbine blade corresponding to the suspected defect region. This improves the accuracy of wind turbine blade surface defect detection.

Description

Technical Field

[0001] This application relates to the field of defect detection technology, specifically to a device and method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicles (UAVs). Background Technology

[0002] Wind turbine blades are key components for capturing wind energy. When the blades are subjected to wind force, the resulting drag and lift convert the wind's kinetic energy into mechanical energy, which is then output as electrical energy through the transmission system. Because wind farms are typically located in geographically complex and climatically harsh environments such as mountains and oceans, wind turbine blades are susceptible to hurricane erosion and damage from lightning, rain, and snow. The protective coating on the blade surface may peel off or develop cracks, affecting the normal operation of the wind turbine. Therefore, it is necessary to monitor defects on wind turbine blades.

[0003] By combining drone technology with intelligent sensors and visual inspection technology, automated inspection of wind turbine blade defects in wind farms can be achieved using drones, offering lower costs and higher efficiency compared to traditional methods. The invention patent CN113406107B, a wind turbine blade defect detection system, sends image data collected by the drone to a cloud server. It then uses a traditional CNN model for defect training and detection. However, due to the complexity of defect types and susceptibility to environmental interference, the accuracy of conventional neural network models in identifying wind turbine blade defects still needs improvement. Existing methods, when detecting surface defects on blades, face challenges due to the complexity and varying sizes of defects, susceptibility to environmental interference, and the dynamic changes in defect areas caused by the continuous rotation of wind turbine blades under aerodynamic and centrifugal forces. This makes it difficult to accurately extract corresponding defect features, resulting in low accuracy for some defect identifications. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide a device and method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicles (UAVs). The specific technical solution adopted is as follows:

[0005] In a first aspect, embodiments of this application provide a method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicles (UAVs), the method comprising the following steps:

[0006] Acquire visible light and infrared images of the wind turbine blades;

[0007] Acquire all suspected defect regions and background regions in all visible light images within the monitoring period; obtain the difference change coefficient of a single suspected defect region based on the gray level difference between a single suspected defect region and the background region in the image, as well as the gradient of the edge pixels in the suspected defect region; and obtain the defect significance index of a single suspected defect region by combining the curvature of each edge pixel in a single suspected defect region and the gray level co-occurrence matrix of all pixels.

[0008] Based on the grayscale difference between the infrared region and the background region in the minimum bounding rectangle of the infrared region corresponding to a single suspected defect region in the infrared image, and the dispersion of grayscale values ​​of all pixels in the infrared region, the infrared difference coefficient of a single suspected defect region is obtained; the comprehensive defect significance value of a single suspected defect region is obtained based on the defect significance index and the infrared difference coefficient of a single suspected defect region.

[0009] Based on the combined defect significance value of a single suspected defect area in the first visible light image and suspected defect areas representing the same blade area in other visible light images within the monitoring period, the defect significance coefficient of a single suspected defect area in the first visible light image is obtained to determine whether there is a defect on the surface of the wind turbine blade corresponding to the suspected defect area.

[0010] Preferably, the specific process for acquiring each suspected defect region and background region of all visible light images within the monitoring period is as follows:

[0011] The grayscale values ​​of all pixels in the first visible light image within the monitoring period are compared with a first preset threshold. Pixels with grayscale values ​​less than the first preset threshold are recorded as suspected defect points, and all pixels with grayscale values ​​greater than or equal to the first preset threshold are recorded as background points. Based on the distribution of suspected defect points and background points, each suspected defect region and background region in the first visible light image is obtained.

[0012] The regions representing the same blade region as the suspected defect regions in the first visible light image are obtained from all other visible light images within the monitoring period, thus obtaining the suspected defect regions and background regions in all other visible light images.

[0013] Preferably, the method for determining the difference variation coefficient of a single suspected defect region is as follows: calculate the difference between the average gray value of all pixels in the single suspected defect region and the average gray value of all pixels in the background region of the image, and calculate the gradient of all edge pixels in the visible light image of the suspected defect region. The product of the average magnitude of all the gradients in the single suspected defect region and the difference is used as the difference variation coefficient of the corresponding single suspected defect region.

[0014] Preferably, the method for determining the defect significance index of a single suspected defect region is as follows:

[0015] The mean curvature of all edge pixels in a single suspected defect region is used as the edge disorder coefficient of the single suspected defect region.

[0016] Calculate the gray-level co-occurrence matrix of all pixels within a single suspected defect region, and use the entropy of the gray-level co-occurrence matrix as the texture complexity coefficient of the corresponding single suspected defect region;

[0017] The positive fusion result of the edge disorder coefficient, difference variation coefficient and texture complexity coefficient of a single suspected defect region is used as the defect significance index of the corresponding single suspected defect region.

[0018] Preferably, the process of obtaining the infrared difference coefficient of a single suspected defect region is as follows: obtain the minimum bounding rectangle of the infrared region corresponding to the single suspected defect region in the infrared image, calculate the difference between the mean gray value of all pixels in the infrared region and the mean gray value of all pixels in the background region in the minimum bounding rectangle, and use the product of the absolute value of the difference in the single suspected defect region and the standard deviation of the gray values ​​of all pixels in the infrared region as the infrared difference coefficient of the single suspected defect region.

[0019] Preferably, the comprehensive defect significance value of a single suspected defect region is the product of the defect significance index and the infrared difference coefficient of the single suspected defect region.

[0020] Preferably, the process for obtaining the defect significance coefficient of a single suspected defect region in the first visible light image is as follows:

[0021] Based on the comprehensive defect significance value of a single suspected defect area in the first visible light image within the monitoring period and the corresponding suspected defect areas in all other visible light images of the same blade area, the comprehensive defect significance sequence of the single suspected defect area in the first visible light image during the first and second rotations of the wind turbine blade is obtained.

[0022] The formula for calculating the significance coefficient of a single suspected defect region in the first visible light image is: In the formula, L i P represents the defect significance coefficient of the i-th suspected defect region in the first visible light image. i Q is the mean of all data in the two combined defect saliency sequences of the i-th suspected defect region in the first visible light image. i R is the standard deviation of all data in the two combined defect saliency sequences of the i-th suspected defect region in the first visible light image. i DTW distance is the distance between two composite defect saliency sequences in the i-th suspected defect region of the first visible light image.

[0023] Preferably, the process of obtaining the comprehensive defect significance sequence of a single suspected defect region in the first visible light image during the first and second rotations is as follows: the comprehensive defect significance values ​​of the single suspected defect region in the first visible light image and the suspected defect regions of the same leaf region in all other visible light images within the monitoring period are arranged in the order of the image frames, so that the comprehensive defect significance sequence of the single suspected defect region in the first visible light image during the first and second rotations can be obtained respectively.

[0024] Preferably, the specific process for determining whether there is a defect on the surface of the wind turbine blade corresponding to the suspected defect area is as follows:

[0025] Normalize the defect significance coefficients of all suspected defect areas in the first visible light image within the monitoring period;

[0026] If the normalized result of the defect significance coefficient of a single suspected defect area is greater than or equal to the second preset threshold, then there is a defect on the surface of the wind turbine blade corresponding to the suspected defect area; otherwise, there is no defect on the surface of the wind turbine blade corresponding to the suspected defect area.

[0027] Secondly, embodiments of this application also provide a device for monitoring surface defects of wind turbine blades based on unmanned aerial vehicles (UAVs), including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described methods for monitoring surface defects of wind turbine blades based on unmanned aerial vehicles (UAVs).

[0028] This application has at least the following beneficial effects:

[0029] This application constructs a defect significance index by deeply analyzing the edge features, grayscale features, gradient features, and texture features of the defective parts of the wind turbine blade in visible light images. Combined with the temperature features of the defective parts of the wind turbine blade in infrared images, a comprehensive defect significance value is obtained for each suspected defect area. Its advantage is that compared with defect identification methods based on a single type of image, it can more comprehensively reflect the surface defect features of the blade, which helps to improve the detection accuracy of wind turbine blade surface defects. Furthermore, it considers the dynamic change characteristics of the defect position of the wind turbine blade due to the influence of aerodynamic forces and centrifugal forces generated during rotation, and constructs a defect significance coefficient for each suspected defect area. This is used to monitor wind turbine blade surface defects, thereby improving the accuracy of wind turbine blade defect detection. Attached Figure Description

[0030] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 A flowchart illustrating the steps of a method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicle (UAV) assistance, provided in one embodiment of this application.

[0032] Figure 2 This application provides a process for obtaining the defect significance coefficient of a single suspected defect region in the first visible light image within a monitoring period, as provided in one embodiment of the application. Detailed Implementation

[0033] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the unmanned aerial vehicle (UAV)-assisted wind turbine blade surface defect monitoring device and method proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0034] Unless otherwise defined, 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 pertains.

[0035] The following description, in conjunction with the accompanying drawings, details the specific scheme of the unmanned aerial vehicle-assisted wind turbine blade surface defect monitoring device and method provided in this application.

[0036] Please see Figure 1 The diagram illustrates a flowchart of a method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicle (UAV) assistance, according to an embodiment of this application. The method includes the following steps:

[0037] Step 1: Acquire visible light and infrared images of the wind turbine blades.

[0038] With the development of drone technology, its widespread application in fields such as visual inspection has brought new methods to blade defect detection. Using drone-assisted monitoring greatly ensures personnel safety and significantly improves inspection efficiency. There are various methods for blade defect detection, among which using computer vision detection algorithms to detect surface defects on wind turbine blades best meets practical application needs, achieving the best detection results at the lowest cost. This embodiment uses a drone equipped with intelligent sensors, such as high-definition cameras and infrared thermal imagers, to monitor surface defects on wind turbine blades.

[0039] Due to the significant weight of wind turbine blades, they are subjected to aerodynamic and centrifugal forces during rotation. The thermal conductivity of defective areas on the blades differs more pronounced from normal areas due to material or structural discontinuities, and these defective areas are prone to shape changes and temperature variations. Therefore, this application employs a dynamic detection method to continuously acquire surface image data of the wind turbine blades. A gimbal camera is mounted on a drone to improve the stability of the acquired images. During acquisition, the drone remains stationary, ensuring the camera's view covers all blades of a single wind turbine. Visible light images (RGB images) of the wind turbine blade surface are acquired using a high-definition camera, while infrared images (single-channel grayscale images) are acquired using an infrared thermal imager. Each frame of visible light image corresponds to one frame of infrared image, with the acquisition frequency set at 25 frames per second. The data acquired by the intelligent sensors is transmitted wirelessly to an edge computing server for processing and analysis.

[0040] Step 2: Obtain the suspected defect regions and background regions of all visible light images within the monitoring period; based on the gray level difference between a single suspected defect region and the background region in the image, as well as the gradient of the edge pixels in the suspected defect region, obtain the difference change coefficient of the single suspected defect region; and combine the curvature of each edge pixel in the single suspected defect region with the gray level co-occurrence matrix of all pixels to obtain the defect significance index of the single suspected defect region.

[0041] After being subjected to hurricane erosion and thunderstorm damage, the protective coating on wind turbine blades may suffer from coating peeling, wear, and cracking defects. Visible light images captured by high-definition cameras can clearly show the surface condition of the blades. These defective areas are characterized by irregular edge shapes, significant color differences, rapid changes, and complex internal textures. Infrared images can clearly reflect the thermal characteristics of the blades. By detecting temperature differences on the blade surface, some surface defects that are difficult to detect visually can be found. Compared to normal areas, defective areas in infrared images of wind turbine blades are characterized by irregular shapes, significant color differences, but relatively slow changes. As can be seen, visible light images and infrared images provide different types of information. This application combines the two during the inspection process to gain a more comprehensive understanding of the health condition of the wind turbine blades, which helps to improve the accuracy and reliability of defect detection.

[0042] First, the features of the defect area in each frame of the visible light image and infrared image are analyzed. This embodiment uses two rotations of the wind turbine blades as a monitoring cycle, and takes any one of these cycles as an example for the following analysis. First, all visible light images within the monitoring cycle are converted from RGB images to grayscale images. Then, the resulting grayscale images are subjected to median filtering to reduce noise, resulting in a preprocessed visible light image. Since there is a significant color difference between the defect area and the normal wind turbine blade surface, to segment out the possible defective parts, this embodiment uses the grayscale values ​​of all pixels in the first visible light image within the monitoring cycle as input to the iterative threshold segmentation method. The maximum threshold difference is set to 5 to obtain the segmentation threshold, which is recorded as the first preset threshold. The iterative threshold segmentation method is a known technique, and the specific process will not be elaborated further. Pixels with grayscale values ​​less than the first preset threshold are recorded as suspected defect points, and all pixels with grayscale values ​​greater than or equal to the first preset threshold are recorded as background points. Based on the distribution of suspected defect points, a connected component analysis algorithm is used to divide the suspected defect points into multiple connected components. Each connected component is denoted as a suspected defect region in the first visible light image, and the region in the first visible light image excluding all suspected defect regions is denoted as the background region of the first visible light image. Further, the first and second visible light images within the monitoring period are used as inputs to the Farneback dense optical flow algorithm to obtain corresponding regions in the two images. These corresponding regions represent the same blade area. The Farneback dense optical flow algorithm is a well-known technique, and its specific process will not be elaborated further. The regions in the second visible light image corresponding to each suspected defect region in the first visible light image are denoted as the suspected defect regions in the second visible light image. Similarly, the suspected defect regions of all visible light images within the monitoring period can be obtained.

[0043] Taking the i-th suspected defect region in the first visible light image as an example, all its corresponding edge pixels are divided into multiple segments, each segment consisting of N adjacent edge pixels, where N ranges from [5,8], and in this embodiment, N is 6. Polynomial fitting technology is used to obtain the fitting curve for each segment of edge pixels, and then the curvature value at each pixel in the fitting curve is calculated. The mean curvature of all edge pixels in the i-th suspected defect region is used as the edge disorder coefficient of the suspected defect region, reflecting the degree of irregularity of the corresponding region's edges. Then, the difference between the mean grayscale value of all pixels in the suspected defect region and the mean grayscale value of all pixels in the background region is calculated, and the Laplacian operator is used to calculate the gradient of all edge pixels in the suspected defect region in the visible light image. The product of the mean of the magnitudes of all gradients in the i-th suspected defect region and the difference is used as the difference change coefficient of the i-th suspected defect region, reflecting the magnitude of the color difference between the corresponding region and the background, as well as the speed of color change. Furthermore, to obtain the internal texture features, the gray-level co-occurrence matrix of all pixels within the i-th suspected defect region is calculated. The entropy of the gray-level co-occurrence matrix is ​​used as the texture complexity coefficient of the i-th suspected defect region, which reflects the complexity of the texture within the corresponding region. The calculation of the gray-level co-occurrence matrix is ​​a well-known technique, and the specific process will not be elaborated further.

[0044] The defective area of ​​the wind turbine blade usually contains at least one or more of the above features. Therefore, as a preferred embodiment, the positive fusion result of the edge confusion coefficient, difference variation coefficient and texture complexity coefficient corresponding to a single suspected defective area is used as the defect significance index of the corresponding single suspected defective area.

[0045] In this embodiment, the sum of the edge disorder coefficient, difference variation coefficient, and texture complexity coefficient corresponding to the i-th suspected defect region is used as the defect significance index of the i-th suspected defect region. This value reflects the probability that the corresponding suspected defect region belongs to the defect part in the visible light image. The larger the value of the defect significance index, the greater the probability that the corresponding suspected defect region belongs to the defect part.

[0046] Step 3: Based on the grayscale difference between the infrared region and the background region in the minimum bounding rectangle of the infrared region corresponding to a single suspected defect region in the infrared image, and the dispersion of grayscale values ​​of all pixels in the infrared region, obtain the infrared difference coefficient of a single suspected defect region; obtain the comprehensive defect significance value of a single suspected defect region based on the defect significance index and the infrared difference coefficient of a single suspected defect region.

[0047] Furthermore, in infrared images, defective areas on wind turbine blades exhibit different characteristics compared to normal wind turbine areas in visible light images. Specifically, for each suspected defective area in the visible light image, the temperature distribution of the corresponding suspected defective area can be obtained in the same area of ​​the infrared image. Unlike visible light images, infrared images of normal wind turbine blades may contain different temperature regions, and defective parts of the wind turbine blades appear as locally larger grayscale value differences in the infrared image. Therefore, taking the infrared region corresponding to the i-th suspected defective area as an example, to obtain the local grayscale difference characteristics, the minimum bounding rectangle of the infrared region is obtained, and the difference between the mean grayscale value of all pixels within the infrared region and the mean grayscale value of all pixels within the background region is calculated. The product of the absolute value of the difference in the i-th suspected defective area and the standard deviation of the grayscale values ​​of all pixels in the infrared region is used as the infrared difference coefficient for the i-th suspected defective area. This value reflects the magnitude of the local color difference of the suspected defective area in the infrared image. The larger the value of the infrared difference coefficient, the more likely a defect exists on the surface of the wind turbine blade corresponding to the suspected defective area.

[0048] Furthermore, the saliency of defect features in the same suspected defect area may differ between visible light and infrared images. For example, some defects may be clearly identifiable in visible light images but not in infrared images; conversely, some defects may be clearly identifiable in infrared images but not in visible light images. Therefore, to improve monitoring reliability, a comprehensive defect saliency value is constructed based on the characteristics of the same suspected defect area in both visible and infrared images to characterize the defect saliency of the corresponding suspected defect area. In this embodiment, the product of the defect saliency index and the infrared difference coefficient of the i-th suspected defect area is used as the comprehensive defect saliency value of the i-th suspected defect area. The obtained comprehensive defect saliency value reflects the defect saliency of the corresponding suspected defect area. The larger the comprehensive defect saliency value, the greater the defect saliency of the corresponding suspected defect area.

[0049] Step 4: Based on the combined defect significance value of a single suspected defect area in the first visible light image and suspected defect areas representing the same blade area in other visible light images within the monitoring period, obtain the defect significance coefficient of a single suspected defect area in the first visible light image to determine whether there is a defect on the surface of the wind turbine blade corresponding to the suspected defect area.

[0050] Furthermore, contaminants may be present on the surface of wind turbine blades due to environmental disturbances. These contaminants may exhibit certain defect characteristics in images, but the internal structure of the blade coating at the location of these contaminants may not be damaged, making it easy to misjudge the contaminants as blade surface defects. True blade surface defects are easily affected by aerodynamic and centrifugal forces during wind turbine rotation. The defect characteristics in different frames will show changes in shape and temperature, exhibiting corresponding periodic characteristics with the periodic rotation of the blade. Therefore, if a suspected defect area exhibits the above characteristics, it can be further confirmed that the suspected defect area belongs to the blade surface defect.

[0051] Furthermore, taking the i-th suspected defect region in the first visible light image within a monitoring cycle as an example, in the entire monitoring cycle, there are suspected defect regions in other visible light images besides the first visible light image that represent the same blade region as the i-th suspected defect region. By arranging the comprehensive defect significance values ​​of the i-th suspected defect region and all the corresponding suspected defect regions in the same blade region according to the order of the image frames, the comprehensive defect significance sequence of the i-th suspected defect region during the first and second cycles of rotation can be obtained respectively.

[0052] Based on the average level, dispersion, and degree of difference of the two comprehensive defect significance sequences of each suspected defect region, a defect significance coefficient is constructed for each suspected defect region to characterize the probability that each suspected defect region belongs to a defect region.

[0053] In this embodiment, the defect significance coefficient of the i-th suspected defect region in the first visible light image is denoted as L. i Its specific expression is: In the formula, L i P represents the defect significance coefficient of the i-th suspected defect region in the first visible light image. i Q is the mean of all data in the two combined defect saliency sequences of the i-th suspected defect region in the first visible light image. i R is the standard deviation of all data in the two combined defect saliency sequences of the i-th suspected defect region in the first visible light image. i Let be the DTW distance between two composite defect saliency sequences in the i-th suspected defect region of the first visible light image. The calculation of the DTW distance is a well-known technique, and the specific process will not be described in detail here.

[0054] P i The larger the value of Q, the higher the probability that the suspected defect area is a defect in the visible light image; i The larger the value, the more pronounced the dynamic changes in the suspected defect area during wind turbine blade rotation; R iThe smaller the value, the more similar the two comprehensive defect salient sequences of the i-th suspected defect region are, indicating that the variation characteristics corresponding to this suspected defect region also exhibit corresponding periodic characteristics due to the periodic rotation of the wind turbine blades. The obtained L... i The larger the value, the more significant the defect characteristics of the suspected defect area are when considered in both static images and the dynamic changes during wind turbine rotation, making it more likely to be identified as a defect. The process for obtaining the defect significance coefficient of a single suspected defect area in the first visible light image within the monitoring period is as follows: Figure 2 As shown.

[0055] Finally, the defect significance coefficients of all suspected defect areas in the first visible light image within a monitoring period are calculated. Based on the obtained defect significance coefficients of each suspected defect area, surface defects of the wind turbine blades are monitored. All defect significance coefficients are normalized using the sigmoid function and compared with a second preset threshold. The second preset threshold ranges from [0.4, 0.8], and in this embodiment, it is set to 0.7. If the normalized result of the defect significance coefficient of a single suspected defect area is greater than or equal to the second preset threshold, then the surface of the wind turbine blade corresponding to that suspected defect area has a defect; otherwise, the surface of the wind turbine blade corresponding to that suspected defect area does not have a defect.

[0056] Similarly, the presence of defects on the surface of the wind turbine blades corresponding to all other suspected defect areas in the first visible light image can be evaluated to obtain the monitoring results of the surface defects of the entire wind turbine blades during the monitoring period.

[0057] The above methods help improve the accuracy of visual inspection-based identification of surface defects in wind turbine blades.

[0058] Based on the same inventive concept as the above methods, this application also provides a device for monitoring surface defects of wind turbine blades based on unmanned aerial vehicles (UAVs), including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described methods for monitoring surface defects of wind turbine blades based on unmanned aerial vehicles (UAVs).

[0059] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0060] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0061] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicle (UAV) assistance, characterized in that, The method includes the following steps: Acquire visible light and infrared images of the wind turbine blades; Acquire all suspected defect regions and background regions in all visible light images within the monitoring period; obtain the difference change coefficient of a single suspected defect region based on the gray level difference between a single suspected defect region and the background region in the image, as well as the gradient of the edge pixels in the suspected defect region; and obtain the defect significance index of a single suspected defect region by combining the curvature of each edge pixel in a single suspected defect region and the gray level co-occurrence matrix of all pixels. Based on the grayscale difference between the infrared region and the background region in the minimum bounding rectangle of the infrared region corresponding to a single suspected defect region in the infrared image, and the dispersion of grayscale values ​​of all pixels in the infrared region, the infrared difference coefficient of a single suspected defect region is obtained; the comprehensive defect significance value of a single suspected defect region is obtained based on the defect significance index and the infrared difference coefficient of a single suspected defect region. Based on the combined defect significance value of a single suspected defect area in the first visible light image and suspected defect areas representing the same blade area in other visible light images within the monitoring period, the defect significance coefficient of a single suspected defect area in the first visible light image is obtained to determine whether there is a defect on the surface of the wind turbine blade corresponding to the suspected defect area. The method for determining the defect significance index of a single suspected defect region is as follows: The mean curvature of all edge pixels in a single suspected defect region is used as the edge disorder coefficient of the single suspected defect region. Calculate the gray-level co-occurrence matrix of all pixels within a single suspected defect region, and use the entropy of the gray-level co-occurrence matrix as the texture complexity coefficient of the corresponding single suspected defect region; The positive fusion result of the edge disorder coefficient, difference variation coefficient and texture complexity coefficient of a single suspected defect region is used as the defect significance index of the corresponding single suspected defect region; The process of obtaining the infrared difference coefficient of a single suspected defect region is as follows: obtain the minimum bounding rectangle of the infrared region corresponding to the single suspected defect region in the infrared image, calculate the difference between the gray mean of all pixels in the infrared region and the gray mean of all pixels in the background region in the minimum bounding rectangle, and take the product of the absolute value of the difference in the single suspected defect region and the standard deviation of the gray value of all pixels in the infrared region as the infrared difference coefficient of the single suspected defect region. The comprehensive defect significance value of a single suspected defect region is the product of the defect significance index and the infrared difference coefficient of the single suspected defect region; The process of obtaining the defect significance coefficient of a single suspected defect region in the first visible light image is as follows: Based on the comprehensive defect significance value of a single suspected defect area in the first visible light image within the monitoring period and the corresponding suspected defect areas in all other visible light images of the same blade area, the comprehensive defect significance sequence of the single suspected defect area in the first visible light image during the first and second rotations of the wind turbine blade is obtained. The formula for calculating the significance coefficient of a single suspected defect region in the first visible light image is: In the formula, Let be the defect significance coefficient of the i-th suspected defect region in the first visible light image. Let be the mean of all data in the two combined defect saliency sequences of the i-th suspected defect region in the first visible light image. Let be the standard deviation of all data in the two combined defect saliency sequences of the i-th suspected defect region in the first visible light image. DTW distance is the distance between two composite defect saliency sequences in the i-th suspected defect region of the first visible light image.

2. The method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicle (UAV) assistance as described in claim 1, characterized in that, The specific process for acquiring each suspected defect region and background region in all visible light images within the monitoring period is as follows: The grayscale values ​​of all pixels in the first visible light image within the monitoring period are compared with a first preset threshold. Pixels with grayscale values ​​less than the first preset threshold are recorded as suspected defect points, and all pixels with grayscale values ​​greater than or equal to the first preset threshold are recorded as background points. Based on the distribution of suspected defect points and background points, each suspected defect region and background region in the first visible light image is obtained. The regions representing the same blade region as the suspected defect regions in the first visible light image are obtained from all other visible light images within the monitoring period, thus obtaining the suspected defect regions and background regions in all other visible light images.

3. The method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicle (UAV) assistance as described in claim 1, characterized in that, The method for determining the difference variation coefficient of a single suspected defect region is as follows: calculate the difference between the average gray value of all pixels in the single suspected defect region and the average gray value of all pixels in the background region of the image, and calculate the gradient of all edge pixels in the visible light image of the suspected defect region. The product of the average magnitude of all the gradients in the single suspected defect region and the difference is used as the difference variation coefficient of the corresponding single suspected defect region.

4. The method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicle (UAV) assistance as described in claim 1, characterized in that, The process of obtaining the comprehensive defect significance sequence of a single suspected defect region in the first visible light image during the first and second weeks of rotation is as follows: the comprehensive defect significance values ​​of the single suspected defect region in the first visible light image and the suspected defect regions of the same leaf region in all other visible light images within the monitoring period are arranged in the order of the image frames, and the comprehensive defect significance sequence of the single suspected defect region in the first visible light image during the first and second weeks of rotation can be obtained respectively.

5. The method for monitoring surface defects of wind turbine blades based on unmanned aerial vehicle (UAV) assistance as described in claim 1, characterized in that, The specific process for determining whether there is a defect on the surface of the wind turbine blade corresponding to the suspected defect area is as follows: Normalize the defect significance coefficients of all suspected defect areas in the first visible light image within the monitoring period; If the normalized result of the defect significance coefficient of a single suspected defect area is greater than or equal to the second preset threshold, then there is a defect on the surface of the wind turbine blade corresponding to the suspected defect area; otherwise, there is no defect on the surface of the wind turbine blade corresponding to the suspected defect area.

6. A UAV-assisted wind turbine blade surface defect monitoring device, 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, it implements the steps of the unmanned aerial vehicle-assisted wind turbine blade surface defect monitoring method as described in any one of claims 1-5.

Images