An unmanned aerial vehicle-based wind power fan blade defect automatic positioning and identification system

Abstract

The application discloses a kind of wind power fan blade defect automatic positioning and identification system based on unmanned aerial vehicle, belong to wind power generation equipment detection technical field.It includes wireless communication module, the wireless communication module connects intelligent positioning and defect identification module;The wireless communication module is used to receive the target GPS coordinate or route GPS coordinate data D3 of remote control end and transmission is given to intelligent positioning and defect identification module, and final identification result and task state obtained after intelligent positioning and defect identification module are handled are transmitted to remote control end by wireless communication;The intelligent positioning and defect identification module includes data acquisition module, positioning module, navigation control module, wind power fan blade segmentation and segmentation module and defect identification module.The present application is through wind power fan blade segmentation, segmentation fan blade automatic cutting, defect identification etc., compared with existing defect identification method, not only can the accurate identification of wind power fan blade defect be realized automatically, but also can accurately locate wind power fan blade defect area.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power generation equipment detection, and in particular to an automatic wind turbine blade defect positioning and identification system based on an unmanned aerial vehicle (UAV). Background Art

[0002] As a key component of clean energy, wind power generation is becoming increasingly important in the global energy mix. Wind turbine blades, critical components of wind turbines, have a direct impact on power generation efficiency and safety. Due to their long-term exposure to harsh outdoor environments, subjected to strong winds, sudden temperature fluctuations, sandstorms, and lightning strikes, wind turbine blades are susceptible to various defects such as cracks, wear, and corrosion. Promptly identifying these defects is crucial to ensuring the stable operation of wind turbine equipment.

[0003] At present, there are mainly the following methods for identifying defects in wind turbine blades: (1) Manual inspection method: professional inspectors observe from a distance through a telescope or climb to the blade for close inspection. The former is limited by the observation distance and accuracy, making it difficult to detect small defects; although the latter can be viewed at close range, the inspectors face the risk of working at height, and manual inspection is inefficient and highly subjective, and the inspection results of different inspectors may vary greatly. (2) Traditional non-destructive testing technology: such as ultrasonic testing, magnetic particle testing, infrared thermal imaging testing, etc. Ultrasonic testing can detect internal defects, but it requires contact with the blade surface and is complex to operate; magnetic particle testing is only applicable to ferromagnetic materials and has a limited scope of application; infrared thermal imaging testing is easily affected by ambient temperature and has unstable detection accuracy. These methods have high requirements for the inspection environment and equipment, and are difficult to be effectively applied in large-scale wind farms. (3) Computer vision inspection method based on fixed camera position: a camera is installed on the ground or on the tower to obtain the blade image, and defect identification is performed using traditional image processing algorithms. However, the environment in which wind turbine blades are located is complex, and factors such as lighting changes and background interference affect image quality. Traditional image processing algorithms find it difficult to accurately extract blade features and defect information, resulting in low detection accuracy and reliability.

[0004] Problems with existing technologies: (1) Low detection efficiency: Manual inspection and traditional non-destructive testing methods are cumbersome and slow. For the numerous blades in large-scale wind farms, the inspection cycle is long, making it difficult to detect potential defects in a timely manner, affecting equipment maintenance and operation. (2) Insufficient detection accuracy: Manual inspection is greatly affected by subjective factors. Traditional computer vision inspection methods have limited ability to extract features of small defects and complex defects in complex environments, and are prone to missed detection and false detection. (3) Poor adaptability: Traditional inspection methods are usually designed for specific blade types and environments. For blades of different models, installation angles and environmental conditions, the inspection parameters need to be readjusted or the equipment needs to be replaced, resulting in poor versatility and flexibility. (4) High safety risks: When manually climbing fan blades for inspection, inspectors face safety risks such as falling from heights, which threatens their lives. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a wind turbine blade defect automatic positioning and identification system based on a drone.

[0006] The objective of the present invention is achieved through the following technical solutions: a wind turbine blade defect automatic positioning and identification system based on a drone, comprising a wireless communication module, the wireless communication module being connected to an intelligent positioning and defect identification module; the wireless communication module being used to receive target GPS coordinates or route GPS coordinate data D3 from a remote control terminal and transmit the data to the intelligent positioning and defect identification module, and transmitting the final identification result and task status obtained after processing by the intelligent positioning and defect identification module to the remote control terminal via wireless communication; the intelligent positioning and defect identification module comprising a data acquisition module, a positioning module, a navigation control module, a wind turbine blade segmentation and slicing module, and a defect identification module;

[0007] The data acquisition module is used to obtain the target GPS coordinates or route GPS coordinate data D3, the current edge intelligent drone GPS positioning data D1 and the best shooting point image data D2, and transmit them to the positioning module and the navigation control module;

[0008] The positioning module is used to identify the target based on the target GPS coordinates or route GPS coordinate data D3, and use the best shooting point image data D2 to perform visual positioning when approaching the identified target, and convert the visual positioning result into GPS coordinates to obtain visual positioning data which is input into the navigation control module;

[0009] The navigation control module is used to control the edge intelligent drone to fly to the best shooting point to shoot wind turbine blade image data I according to the data input by the data acquisition module and the visual positioning data;

[0010] The wind turbine blade segmentation and slicing module is used to perform image segmentation on the wind turbine blade image data I, then align the segmented blades to the wind turbine standard blades using the alignment matrix T, and finally cut the wind turbine blade segmentation area image into image blocks of the same size according to a preset slicing template;

[0011] The defect recognition module is used to input the image blocks into the defect recognition model based on the Transformer architecture, and the defect recognition model classifies the defects of each image block; then, the four vertex coordinates of the defective image block are mapped to the original image using the inverse alignment formula to complete the positioning of the defective area of ​​the wind turbine fan blade and calculate the defect area; finally, the defect block number, defect type and defect block vertex coordinates are output to the wireless communication module as the final recognition result.

[0012] Preferably, the positioning module includes the following steps when performing positioning:

[0013] It is determined that the distance between the current edge intelligent UAV GPS positioning and the target wind turbine equipment GPS positioning is greater than the distance threshold. If it is greater than the distance threshold, the target wind turbine equipment GPS positioning is directly input into the navigation control module to complete the global rough navigation from the edge intelligent UAV to the target wind turbine equipment; if it is less than or equal to the distance threshold, the edge intelligent UAV is visually positioned relative to the target wind turbine blade through the best shooting point image data D2, and the visual positioning result is converted into GPS coordinates to obtain visual positioning data, which is input into the navigation control module to realize local precise navigation from the edge intelligent UAV to the best shooting point of the target wind turbine blade.

[0014] Preferably, the visual positioning of the edge intelligent drone relative to the target wind turbine blade is performed using the best shooting point image data D2, specifically including the following steps:

[0015] The 2D key points of the target wind turbine blade are detected based on the image data D2 of the best shooting point; then the posture is solved, and the 2D key points of the blade are matched with the 3D coordinates of the wind turbine standard blade to obtain the 2D key points and 3D coordinate matching pairs; finally, the cv::SolvePnP() interface of Opencv is used to solve the visual positioning result.

[0016] Preferably, the wind turbine blade segmentation and slicing module includes the following steps when performing image segmentation and slicing:

[0017] Using the segmentation model to segment the wind turbine blade image data I to obtain segmented blades;

[0018] Then, the endpoint coordinates of the target wind turbine blade and the endpoint coordinates of the wind turbine standard blade are obtained, and the segmented blade is aligned to the wind turbine standard blade through the alignment matrix T to obtain the wind turbine blade segmentation area image;

[0019] Finally, the wind turbine blade segmentation area image is cut into image blocks of the same size according to the preset segmentation template.

[0020] Preferably, the defect identification module includes the following steps when performing defect identification:

[0021] Input the image patch into the defect recognition model;

[0022] Encode the image block into an output vector of length Length through the MLP embedding layer;

[0023] Use the relative position coding algorithm to position code the number of each image block to obtain a position coding vector with a coding length of LengthPoE;

[0024] Concatenate the output vector and the position encoding vector to obtain a concatenated vector of length Length+LengthPoE;

[0025] The concatenated vector is then input into N Encoder layers consisting of a self-attention mechanism and MLP feedforward;

[0026] Then, the vector output by the encoder layer is fed into the prediction layer consisting of a multi-layer fully connected network (MLP) and a softmax layer, and finally Num recognition results are obtained. Each recognition result includes the defect type and confidence level.

[0027] Then, the inverse alignment formula is used to map the four vertex coordinates of the defective image block to the original image, completing the location of the defective area of ​​the wind turbine fan blade and counting the defect area; finally, the defect block number, defect type and vertex coordinates of the defect block are output to the wireless communication module as the recognition result.

[0028] Preferably, the segmentation model is a DeepLab V3+ model, a UNet model or a Mask2Former model; and the alignment matrix T is calculated by matrix inversion or SVD decomposition.

[0029] Preferably, the defect types include normal blade, broken blade, cracked blade, corroded blade and other defects; the Length is 64, the LengthPoE is 16, and the N is 10.

[0030] Preferably, the task status is a global variable, and the task status includes navigating, positioning wind turbine blades, identifying defects, and identification completed.

[0031] The beneficial effects of the present invention are:

[0032] 1) This invention offers the advantages of requiring no human intervention, being safe, and highly efficient. Regarding automatic wind turbine blade positioning, it utilizes an edge-based intelligent drone to automatically navigate and position the blades, capturing images from optimal locations. This solves the problem of wind turbine blade data collection relying on manual or fixed-position photography. Furthermore, using images captured from optimal locations significantly improves the accuracy of defect identification. Furthermore, compared to existing defect identification methods, this method not only automatically and accurately identifies wind turbine blade defects, but also accurately locates defective areas through the steps of blade segmentation, automatic blade dicing, and defect identification.

[0033] 2) Through the positioning module based on GPS and visual positioning, the drone can automatically navigate and obtain images of the optimal shooting point of the wind turbine fan blade, effectively solving the problem of data acquisition relying on manual or fixed-position shooting. At the same time, the images taken at the optimal point can greatly improve the accuracy of defect identification.

[0034] 3) Through the defect recognition module based on the Transformer model, the segmentation, cutting, and recognition of the wind power fan blade can be automatically realized, which not only improves the high accuracy, but also accurately locates the defect area of the wind power fan blade. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 It is a system principle block diagram of the application;

[0036] Figure 2 It is an automatic positioning and navigation control flow chart;

[0037] Figure 3 It is a wind power fan blade slicing and defect recognition flow chart;

[0038] Figure 4 It is a wind power fan blade segmentation graph and standard fan blade graph alignment schematic diagram;

[0039] Figure 5 It is a defect recognition model schematic diagram based on the Transformer architecture. DETAILED DESCRIPTION

[0040] The technical solutions of the application will be described below in conjunction with the embodiments, apparently, the described embodiments are only a part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0041] Reference Figures 1-5 , the application provides a kind of technical scheme: a kind of wind power fan blade defect automatic positioning and identification system based on unmanned aerial vehicle, including wireless communication module, the wireless communication module connects intelligent positioning and defect identification module;The wireless communication module is used to receive the target GPS coordinate or route GPS coordinate data D3 of remote control end and transmit to intelligent positioning and defect identification module, and the final identification result and task state obtained after intelligent positioning and defect identification module processing are transmitted to remote control end by wireless communication;The intelligent positioning and defect identification module includes data acquisition module, positioning module, navigation control module, wind power fan blade segmentation and cutting module and defect identification module;

[0042] The data acquisition module is used to obtain target GPS coordinate or route GPS coordinate data D3, current edge intelligent unmanned aerial vehicle GPS positioning data D1 and best shooting point image data D2, and transmit to positioning module and navigation control module;

[0043] The positioning module is used to identify the target based on the target GPS coordinates or route GPS coordinate data D3, and use the best shooting point image data D2 to perform visual positioning when approaching the identified target, and convert the visual positioning result into GPS coordinates to obtain visual positioning data which is input into the navigation control module;

[0044] The navigation control module is used to control the edge intelligent drone to fly to the best shooting point to shoot wind turbine blade image data I according to the data input by the data acquisition module and the visual positioning data;

[0045] The wind turbine blade segmentation and slicing module is used to perform image segmentation on the wind turbine blade image data I, then align the segmented blades to the wind turbine standard blades using the alignment matrix T, and finally cut the wind turbine blade segmentation area image into image blocks of the same size according to a preset slicing template;

[0046] The defect recognition module is used to input the image blocks into the defect recognition model based on the Transformer architecture, and the defect recognition model classifies the defects of each image block; then, the four vertex coordinates of the defective image block are mapped to the original image using the inverse alignment formula to complete the positioning of the defective area of ​​the wind turbine fan blade and calculate the defect area; finally, the defect block number, defect type and defect block vertex coordinates are output to the wireless communication module as the final recognition result.

[0047] In this embodiment, the positioning module realizes automatic positioning by combining GPS and visual information, and inputs the positioning results into the navigation control module, thereby controlling the drone to automatically fly to the optimal shooting point of the wind turbine blade and complete the shooting of the wind turbine blade image data I.

[0048] like Figure 2As shown, D1 is used as the current edge intelligent drone's location, Pos_cur. The GPS location is compared with the target point, Pos_target, determined by D3. If the distance between the two is greater than the distance threshold C_threshold (set to 1m in the implementation), it indicates that the target point has not been approached. At this time, the GPS positioning result is directly input into the navigation control module, completing the global coarse navigation of the edge intelligent drone to the target wind turbine equipment. If the distance is less than or equal to the distance threshold C_threshold, the drone is visually positioned relative to the wind turbine blade using GPS and visual information, and the positioning result is input into the navigation control module to achieve local precise navigation of the edge intelligent drone to the optimal shooting point of the target wind turbine blade. The optimal shooting point refers to the GPS position when the edge intelligent drone's camera is facing the wind turbine blade and meets the optimal point condition. The target GPS coordinates refer to the target GPS coordinates sent by the remote control terminal to the wireless communication module or the preset target GPS coordinates in the edge intelligent drone via wireless communication. The route GPS coordinates refer to the route GPS coordinates sent by the remote control terminal to the wireless communication module or the preset route GPS coordinates in the edge intelligent drone via wireless communication.

[0049] The navigation control module uses a flight control system with autonomous navigation and waypoint flight capabilities, such as PX4Autopilot.

[0050] In some embodiments, the positioning module includes the following steps when performing positioning:

[0051] It is determined that the distance between the current edge intelligent UAV GPS positioning and the target wind turbine equipment GPS positioning is greater than the distance threshold. If it is greater than the distance threshold, the target wind turbine equipment GPS positioning is directly input into the navigation control module to complete the global rough navigation from the edge intelligent UAV to the target wind turbine equipment; if it is less than or equal to the distance threshold, the edge intelligent UAV is visually positioned relative to the target wind turbine blade through the best shooting point image data D2, and the visual positioning result is converted into GPS coordinates to obtain visual positioning data, which is input into the navigation control module to realize local precise navigation from the edge intelligent UAV to the best shooting point of the target wind turbine blade.

[0052] In some embodiments, visual positioning of an edge intelligent drone relative to a target wind turbine blade is performed using the best shooting point image data D2, specifically including the following steps:

[0053] The 2D key points of the target wind turbine blade are detected based on the image data D2 of the best shooting point; then the posture is solved, and the 2D key points of the blade are matched with the 3D coordinates of the wind turbine standard blade to obtain the 2D key points and 3D coordinate matching pairs; finally, the cv::SolvePnP() interface of Opencv is used to solve the visual positioning result.

[0054] In this embodiment, if Figure 4 As shown in the figure, the fan blade 2D key points are A1, B1, C1 and D; the pose solution uses the four detected fan blade 2D key points and the four fan blade 3D coordinates corresponding to the wind turbine standard fan blade (A0, B0, C0 and D0, which can be obtained from the wind turbine fan blade structure design drawings), thereby obtaining four pairs of 2D key points and 3D coordinate matching pairs, and using Opencv's cv::SolvePnP() interface to solve the visual positioning result pos.

[0055] Convert the visual positioning result pos to the GPS coordinate system. Assuming the transformation matrix T_gps_cam between the edge intelligent drone's GPS module and camera is known, the visual positioning result pos in the GPS coordinate system is pos_gps = T_gps_cam * pos. Using pos_gps and the coordinates of the wind turbine blade center, we can determine whether the drone has reached the optimal shooting point.

[0056] In some embodiments, the wind turbine blade segmentation and slicing module includes the following steps when performing image segmentation and slicing:

[0057] Using the segmentation model to segment the wind turbine blade image data I to obtain segmented blades;

[0058] Then, the endpoint coordinates of the target wind turbine blade and the endpoint coordinates of the wind turbine standard blade are obtained, and the segmented blade is aligned to the wind turbine standard blade through the alignment matrix T to obtain the wind turbine blade segmentation area image;

[0059] Finally, the wind turbine blade segmentation area image is cut into image blocks of the same size according to the preset segmentation template.

[0060] In this embodiment, if Figure 4 As shown, the endpoint coordinates of the target wind turbine blade include A1, B1 and C1, and the endpoint coordinates of the standard wind turbine blade include A0, B0 and C0. inv() is a matrix inversion operation; Num image blocks with a side length of L are used as the preset segmentation template. In the actual embodiment, Num is 26 and L is 100 pixels.

[0061] In some embodiments, the defect recognition module includes the following steps when performing defect recognition:

[0062] Input the image patch into the defect recognition model;

[0063] Encode the image block into an output vector of length Length through the MLP embedding layer;

[0064] Use the relative position coding algorithm to position code the number of each image block to obtain a position coding vector with a coding length of LengthPoE;

[0065] Concatenate the output vector and the position encoding vector to obtain a concatenated vector of length Length+LengthPoE;

[0066] The concatenated vector is then input into N Encoder layers consisting of a self-attention mechanism and MLP feedforward;

[0067] Then, the vector output by the encoder layer is fed into the prediction layer consisting of a multi-layer fully connected network (MLP) and a softmax layer, and finally Num recognition results are obtained. Each recognition result includes the defect type and confidence level.

[0068] Then, the inverse alignment formula is used to map the four vertex coordinates of the defective image block to the original image, completing the location of the defective area of ​​the wind turbine fan blade and counting the defect area; finally, the defect block number, defect type and vertex coordinates of the defect block are output to the wireless communication module as the recognition result.

[0069] In this embodiment, if Figure 3 As shown in the figure, the captured image is segmented into wind turbine blades, then aligned to the standard blades, and then the aligned wind turbine blade image is sliced; then the sliced ​​image blocks are subjected to defect recognition, and then reversely aligned to the original image, and finally the recognition results are assembled, and the recognition results and task status are output.

[0070] In some embodiments, the segmentation model is a DeepLab V3+ model, a UNet model, or a Mask2Former model; and the alignment matrix T is calculated by matrix inversion or SVD decomposition.

[0071] In some embodiments, the defect types include normal blade, broken blade, cracked blade, corroded blade, and other defects; the Length is 64, the LengthPoE is 16, and N is 10.

[0072] In some embodiments, the task status is a global variable, and the task status includes navigating, positioning wind turbine blades, identifying defects, and identification completed.

[0073] The foregoing description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein and should not be construed as excluding other embodiments. Rather, the present invention can be used in various other combinations, modifications, and environments and can be modified within the scope of the concept described herein through the above teachings or techniques or knowledge in the relevant field. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention are intended to be protected by the appended claims.

Claims

1. A wind turbine blade defect automatic location and identification system based on drones, characterized by: It includes a wireless communication module, which is connected to the intelligent positioning and defect recognition module; the wireless communication module is used to receive the target GPS coordinates or route GPS coordinate data D3 of the remote control terminal and transmit it to the intelligent positioning and defect recognition module, and transmit the final recognition result and task status obtained after processing by the intelligent positioning and defect recognition module to the remote control terminal via wireless communication; the intelligent positioning and defect recognition module includes a data acquisition module, a positioning module, a navigation control module, a wind turbine blade segmentation and cutting module, and a defect recognition module; The data acquisition module is used to obtain the target GPS coordinates or route GPS coordinate data D3, the current edge intelligent drone GPS positioning data D1 and the best shooting point image data D2, and transmit them to the positioning module and the navigation control module; The positioning module is used to identify the target based on the target GPS coordinates or route GPS coordinate data D3, and use the best shooting point image data D2 to perform visual positioning when approaching the identified target, and convert the visual positioning result into GPS coordinates to obtain visual positioning data which is input into the navigation control module; The navigation control module is used to control the edge intelligent drone to fly to the optimal shooting point to shoot wind turbine fan blade image data I according to the data input by the data acquisition module and the visual positioning data; the optimal shooting point refers to the GPS position when the edge intelligent drone camera is facing the wind turbine fan blade and meets the optimal point condition; The wind turbine blade segmentation and slicing module is used to perform image segmentation on the wind turbine blade image data I, then align the segmented blades to the wind turbine standard blades using the alignment matrix T, and finally cut the wind turbine blade segmented area image into image blocks of the same size according to a preset slicing template; The defect recognition module is used to input the image blocks into the defect recognition model based on the Transformer architecture. The defect recognition model classifies each image block into defects. Then, the inverse alignment formula is used to map the four vertex coordinates of the defective image block to the original image, completing the location of the defective area of ​​the wind turbine blade and calculating the defect area. Finally, the defect block number, defect type and vertex coordinates of the defect block are output as the final recognition result to the wireless communication module. The positioning module includes the following steps when performing positioning: It is determined that the distance between the current edge intelligent UAV GPS positioning and the target wind turbine equipment GPS positioning is greater than the distance threshold. If it is greater than the distance threshold, the target wind turbine equipment GPS positioning is directly input into the navigation control module to complete the global rough navigation from the edge intelligent UAV to the target wind turbine equipment; if it is less than or equal to the distance threshold, the edge intelligent UAV is visually positioned relative to the target wind turbine blade through the best shooting point image data D2, and the visual positioning result is converted into GPS coordinates to obtain visual positioning data, which is input into the navigation control module to realize local precise navigation from the edge intelligent UAV to the best shooting point of the target wind turbine blade.

2. The UAV-based wind turbine blade defect automatic positioning and identification system according to claim 1 is characterized by: The visual positioning of the edge intelligent drone relative to the target wind turbine blade is performed using the best shooting point image data D2, which specifically includes the following steps: The 2D key points of the target wind turbine blade are detected based on the image data D2 of the best shooting point; then the posture is solved, and the 2D key points of the blade are matched with the 3D coordinates of the wind turbine standard blade to obtain the 2D key points and 3D coordinate matching pairs; finally, the cv::SolvePnP() interface of Opencv is used to solve the visual positioning result.

3. The UAV-based wind turbine blade defect automatic location and identification system according to claim 1 is characterized by: The wind turbine blade segmentation and slicing module includes the following steps when performing image segmentation and slicing: Using the segmentation model to segment the wind turbine blade image data I to obtain segmented blades; Then, the endpoint coordinates of the target wind turbine blade and the endpoint coordinates of the wind turbine standard blade are obtained, and the segmented blade is aligned to the wind turbine standard blade through the alignment matrix T to obtain the wind turbine blade segmentation area image; Finally, the wind turbine blade segmentation area image is cut into image blocks of the same size according to the preset segmentation template.

4. The UAV-based wind turbine blade defect automatic location and identification system according to claim 1 is characterized by: The defect recognition module includes the following steps when performing defect recognition: Input the image patch into the defect recognition model; Encode the image block into an output vector of length Length through the MLP embedding layer; Use the relative position coding algorithm to position code the number of each image block to obtain a position coding vector with a coding length of LengthPoE; Concatenate the output vector and the position encoding vector to obtain a concatenated vector of length Length+LengthPoE; The concatenated vector is then input into N Encoder layers consisting of a self-attention mechanism and MLP feedforward; Then, the vector output by the encoder layer is fed into the prediction layer consisting of a multi-layer fully connected network (MLP) and a softmax layer, and finally Num recognition results are obtained. Each recognition result includes the defect type and confidence level. Then, the inverse alignment formula is used to map the four vertex coordinates of the defective image block to the original image, completing the location of the defective area of ​​the wind turbine fan blade and counting the defect area; finally, the defect block number, defect type and vertex coordinates of the defect block are output to the wireless communication module as the recognition result.

5. The UAV-based wind turbine blade defect automatic positioning and identification system according to claim 3 is characterized by: The segmentation model is a DeepLab V3+ model, a UNet model, or a Mask2Former model; the alignment matrix T is calculated by matrix inversion or SVD decomposition.

6. The UAV-based wind turbine blade defect automatic location and identification system according to claim 4 is characterized by: The defect type includes normal blade, broken blade, cracked blade, corroded blade and other defects; the Length is 64, the LengthPoE is 16, and the N is 10.

7. The UAV-based wind turbine blade defect automatic location and identification system according to any one of claims 1 to 6, characterized in that: The task status is a global variable, and the task status includes navigating, positioning wind turbine blades, identifying defects, and identification completed.

Images