Wind driven generator surface damage detection method, system and equipment and storage medium
By using the YOLOv5_EMA model to detect the surface image of the wind turbine, the problems of low manual judgment efficiency and high misjudgment rate in the prior art are solved, and the rapid and accurate detection of surface damage of the wind turbine is achieved.
Patent Information
- Application Number
- CN202411780118.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-05
- Publication Date
- 2025-03-18
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
After the existing methods of judging damage through drones, the main method of judging damage is to make manual judgments based on experience, resulting in low judgment efficiency and high misjudgment rate under the large scale of inspection, which affects the surface damage inspection work of wind turbines.
The YOLOv5_EMA model is used to detect the surface image to be detected by the wind turbine. The detection feature data is obtained through the trained model, and the probability that the image is judged as surface damage is calculated, and then whether there is surface damage in the wind turbine is determined.
It realizes rapid and accurate detection of surface damage of wind turbines, improves detection efficiency, reduces the misjudgment rate, and avoids the problems of low manual judgment efficiency and high misjudgment rate.
Smart Images
Figure CN119648670A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of wind power generation fault inspection and repair, and in particular relates to a method, system, equipment and storage medium for detecting surface damage of a wind power generator. Background Art
[0002] With the rapid development of the economy and the transformation of the energy structure, wind power generation plays an increasingly important role in the construction of the national energy system due to its advantages of being renewable, environmentally friendly, low-cost, short infrastructure cycle, flexible installed capacity and strong adaptability. Wind turbines, or windmills, are the core equipment for wind power generation. They capture wind energy through blades and transfer the kinetic energy of wind to the generator through blades, speed increasers and other devices, and the generator then converts mechanical energy into electrical energy.
[0003] Since wind turbines are generally installed in the wild or on the coastline, they are extremely susceptible to surface damage caused by lightning, hail, sandstorms or other weather conditions during the long-term power generation process. If the surface damage of the wind turbine is not detected and repaired in time, the power generation efficiency of the wind turbine will be greatly affected. As the surface damage accumulates, it may even cause the wind turbine to break, bringing great safety hazards and economic losses.
[0004] Traditional methods for inspecting surface damage of wind turbines include: manual climbing inspection, which is inefficient and high-risk. As the scale of wind power generation expands, this method has great limitations; visual inspection, directly observing the surface of the wind turbine through a telescope, but this method requires extremely high experience and technical level of the observer, so it has high limitations; drones collect images of the surface of the wind turbine and judge the images. This method has low risk and high efficiency and is suitable for large-scale wind turbine surface damage inspections.
[0005] However, after the existing drone is used to collect images of the surface of the wind turbine, the method of judging damage is mainly through manual judgment based on experience. When the inspection scale is large, it will lead to low judgment efficiency and high misjudgment rate, which will have an adverse impact on the surface damage inspection work of the wind turbine. Summary of the invention
[0006] In order to solve the technical problem that surface damage judgment of collected wind turbine images is mainly performed manually based on experience in the background technology, which leads to low judgment efficiency and high misjudgment rate when the inspection scale is large, and has an adverse effect on the surface damage inspection work of wind turbines, the present invention provides a wind turbine surface damage detection method, system, equipment and storage medium.
[0007] In order to achieve the above object, the present invention adopts the following technical solution:
[0008] In a first aspect, the present invention provides a method for detecting surface damage of a wind turbine generator, the method comprising:
[0009] S1: Acquire the surface image of the wind turbine to be inspected;
[0010] S2: Inputting the surface image of the wind turbine to be inspected into the trained YOLOv5_EMA model for inspection to obtain inspection feature data; wherein the trained YOLOv5_EMA model is generated based on the YOLOv5 model and the EMA attention module, and the trained YOLOv5_EMA model is formed by training with a training set, and the training set includes multiple images of surface damage of wind turbines;
[0011] S3: Determine the probability of judging the to-be-detected surface image of the wind turbine as a surface damage image according to the detection feature data, and judge whether the corresponding wind turbine has surface damage based on the probability.
[0012] Optionally, the step of obtaining the trained YOLOv5_EMA model includes:
[0013] Establish a database of wind turbine surface damage images;
[0014] Generate a training set based on a database of wind turbine surface damage images;
[0015] The untrained YOLOv5_EMA model is constructed based on the YOLOv5 model and the EMA attention module;
[0016] The untrained YOLOv5_EMA model is iteratively trained using the training set to form a trained YOLOv5_EMA model.
[0017] Optionally, the step of obtaining the trained YOLOv5_EMA model further includes:
[0018] Generate a validation set based on a database of wind turbine surface damage images;
[0019] In the process of training the untrained YOLOv5_EMA model, verification is performed using the verification set. When the verification result tends to converge, the training is stopped to obtain the trained YOLOv5_EMA model.
[0020] Optionally, based on the YOLOv5 model and the EMA attention module, the specific steps to construct an untrained YOLOv5_EMA model are:
[0021] A YOLOv5 benchmark model is constructed, and the EMA attention module is inserted into the first C3 module of the YOLOv5 benchmark model to form a new module C3_EMA module, and an untrained YOLOv5_EMA model is obtained; among them, the C3_EMA module includes a parallel main branch and EMA branch.
[0022] Optionally, in step S2, the step of detecting the surface image to be detected by using the trained YOLOv5_EMA model is as follows:
[0023] S2.1: Input the surface image to be detected into the Focus layer in the YOLOv5_EMA model for slicing to obtain the first feature map;
[0024] S2.2: The first feature map is fed into the C3_EMA module in the YOLOv5_EMA model, features are extracted through the main branch and the EMA branch respectively, and features are enhanced using the attention mechanism in the EMA branch to obtain a second feature map;
[0025] S2.3: extract features from the second feature map through the backbone network in the YOLOv5_EMA model, and fuse features at different scales through upsampling and feature concatenation to obtain a third feature map;
[0026] S2.4: Input the third feature map into the Detect module in the YOLOv5_EMA model, and obtain detection feature data through the Detect module.
[0027] Optionally, the training set includes a variety of surface damage images of wind turbines, and each type of surface damage image includes multiple images.
[0028] Optionally, the step S3 includes determining, based on the detection feature data, a probability of identifying the surface image of the wind turbine to be detected as a certain surface damage image;
[0029] If the probability that the surface image to be inspected of the wind turbine is identified as a certain surface damage image is greater than 70%, then the judgment result is that the corresponding wind turbine has this type of surface damage;
[0030] If the probability of identifying the to-be-detected surface image of the wind turbine as any type of surface damage image is less than or equal to 70%, the judgment result is that the corresponding wind turbine has no surface damage.
[0031] In a second aspect, the present invention further provides a wind turbine surface damage detection system, comprising:
[0032] An image acquisition unit, used for acquiring an image of a surface to be inspected of a wind turbine;
[0033] A processing unit is used to input the surface image to be detected of the wind turbine into the trained YOLOv5_EMA model for detection to obtain detection feature data; and according to the detection feature data, determine the probability of judging the surface image to be detected of the wind turbine as a surface damage image, and based on the probability, judge whether the wind turbine has surface damage;
[0034] The output unit outputs a judgment result on whether the wind turbine has surface damage.
[0035] In a third aspect, the present invention also provides a wind turbine surface damage detection device, the wind turbine surface damage detection device comprising a processing component and a storage component, the storage component storing instructions, the instructions being generated according to any one of the wind turbine surface damage detection methods described above, and the processing component calling the instructions.
[0036] In a fourth aspect, the present invention further provides a storage medium, wherein the storage medium stores instructions, wherein the instructions are generated according to any one of the wind turbine surface damage detection methods described above.
[0037] The beneficial effects of the present invention are:
[0038] (1) The present invention provides a method for detecting surface damage of a wind turbine generator. The surface image of the wind turbine generator to be detected is detected by a trained YOLOv5_EMA model, thereby obtaining detection feature data. According to the detection feature data, the probability of judging the surface image of the wind turbine generator to be detected as a surface damage image is determined. Based on the probability, it is judged whether the corresponding wind turbine generator has surface damage. The present invention utilizes the YOLOv5_EMA model to achieve rapid and accurate detection and judgment of the surface image of the wind turbine generator to be detected, thereby improving the detection efficiency and avoiding the technical problem in the prior art that the surface image of the wind turbine generator is judged manually based on experience, which has low judgment efficiency, high misjudgment rate, and has an adverse effect on the surface damage inspection work of the wind turbine generator.
[0039] (2) At the same time, the present invention trains the YOLOv5_EMA model through surface damage images of various wind turbines. During use, the type of surface damage of the wind turbine can be judged to improve the inspection efficiency of the surface damage of the wind turbine.
[0040] (3) The present invention also provides a wind turbine surface damage detection system, which can be used to detect the surface image of the wind turbine to obtain a judgment result on whether the wind turbine has surface damage, and output the judgment result to improve the inspection efficiency of the wind turbine surface damage. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 It is a schematic flow chart of the wind turbine surface damage detection method in the present invention;
[0042] Figure 2 It is a flow chart of the steps of detecting the surface image to be detected by using the trained YOLOv5_EMA model in the present invention;
[0043] Figure 3 is a schematic diagram of image enhancement of a wind turbine surface damage image in a training set in the present invention;
[0044] Figure 4 This is a schematic diagram of the effect of the wind turbine surface damage detection method of the present invention;
[0045] Figure 5 It is another schematic diagram of the effect of the wind turbine surface damage detection method of the present invention. DETAILED DESCRIPTION
[0046] The present invention provides a method, system and equipment for detecting surface damage of a wind turbine generator, which are used for detecting surface damage of a wind turbine generator.
[0047] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.
[0048] Example 1
[0049] First, see Figures 1 to 5 , showing a method for detecting surface damage of a wind turbine described in the present application, comprising the steps of:
[0050] S1: Acquire the surface image of the wind turbine to be inspected;
[0051] S2: Inputting the surface image of the wind turbine to be inspected into the trained YOLOv5_EMA model for inspection to obtain inspection feature data; wherein the trained YOLOv5_EMA model is generated based on the YOLOv5 model and the EMA attention module, and the trained YOLOv5_EMA model is formed by training with a training set, and the training set includes multiple images of surface damage of wind turbines;
[0052] S3: Determine the probability of judging the to-be-detected surface image of the wind turbine as a surface damage image according to the detection feature data, and judge whether the corresponding wind turbine has surface damage based on the probability.
[0053] In this embodiment, a method for detecting surface damage of a wind turbine is provided. The acquired surface image of the wind turbine to be detected is detected by a trained YOLOv5_EMA model, thereby obtaining detection feature data. According to the detection feature data, the probability of judging the surface image of the wind turbine to be detected as a surface damage image is determined. Based on the probability, it is judged whether the corresponding wind turbine has surface damage. The present invention utilizes the YOLOv5_EMA model to achieve rapid and accurate detection and judgment of the surface image of the wind turbine to be detected, thereby improving the detection efficiency and avoiding the technical problem in the prior art that the surface image of the wind turbine is judged manually based on experience, the judgment efficiency is low, the misjudgment rate is high, and the surface damage inspection work of the wind turbine is adversely affected.
[0054] Optionally, the step of obtaining the trained YOLOv5_EMA model in the present invention includes:
[0055] Establish a database of wind turbine surface damage images;
[0056] Generate a training set based on a database of wind turbine surface damage images;
[0057] The untrained YOLOv5_EMA model is constructed based on the YOLOv5 model and the EMA attention module;
[0058] The untrained YOLOv5_EMA model is iteratively trained using the training set to form a trained YOLOv5_EMA model.
[0059] Furthermore, the database of wind turbine surface damage images established in this embodiment should include a plurality of collected wind turbine surface images, from which some high-quality wind turbine surface damage images are screened out to generate a training set.
[0060] Specifically, in this embodiment, the high-quality wind turbine surface damage image refers to an image that is clear and has surface damage, and the training set is generated using the high-quality wind turbine surface damage image.
[0061] Optionally, the step of obtaining the trained YOLOv5_EMA model in the present invention further includes:
[0062] Generate a validation set based on a database of wind turbine surface damage images;
[0063] In the process of iterative training of the untrained YOLOv5_EMA model, verification is performed through the verification set. When the verification result tends to converge, the training is stopped to obtain the trained YOLOv5_EMA model.
[0064] Specifically, in this embodiment, some high-quality wind turbine surface damage images are screened out from the database of wind turbine surface damage images, and some high-quality wind turbine surface damage images are screened out from the database to generate a training set and a validation set. Further, a test set can also be generated to test the trained YOLOv5_EMA model. Specifically, the ratio of high-quality wind turbine surface damage images included in the training set, the test set, and the validation set can be 7:2:1.
[0065] Further, see Figure 3 In the present invention, when the untrained YOLOv5_EMA model is trained by the training set, the surface damage image of the wind turbine in the training set can be enhanced, wherein the image enhancement specifically includes:
[0066] The functions in the YOLOv5_EMA model are used to perform color enhancement, rotation, translation, flipping and other operations on the surface damage images of wind turbines, and then multiple images are spliced using the Mosaic data enhancement method, and the spliced images are used for training the YOLOv5_EMA model. By performing image enhancement on the surface damage images of wind turbines, the training data is expanded to avoid overfitting and improve the training efficiency of the YOLOv5_EMA model.
[0067] Optionally, the specific steps of constructing an untrained YOLOv5_EMA model in the present invention are:
[0068] A YOLOv5 benchmark model is constructed, and the EMA attention module is inserted into the first C3 module of the YOLOv5 benchmark model to form a new module C3_EMA module, and an untrained YOLOv5_EMA model is obtained; among them, the C3_EMA module includes a parallel main branch and EMA branch.
[0069] Specifically, in this embodiment, the constructed YOLOv5 benchmark model can be specifically selected to construct a YOLOv5s network model; the step of inserting the EMA attention module includes: writing the EMA module code into the common.py configuration file of YOLOv5, adding the class name to the registry in the yolo.py file of YOLOv5, and creating a model configuration file yolov5s_C3_EMA1.yaml.
[0070] Optionally, the steps of detecting the surface image to be detected by using the trained YOLOv5_EMA model in the present invention are:
[0071] S2.1: Input the surface image to be detected into the Focus layer in the YOLOv5_EMA model for slicing to obtain the first feature map;
[0072] S2.2: The first feature map is fed into the C3_EMA module in the YOLOv5_EMA model, features are extracted through the main branch and the EMA branch respectively, and features are enhanced using the attention mechanism in the EMA branch to obtain a second feature map;
[0073] S2.3: extract features from the second feature map through the backbone network in the YOLOv5_EMA model, and fuse features at different scales through upsampling and feature concatenation to obtain a third feature map;
[0074] S2.4: Input the third feature map into the Detect module in the YOLOv5_EMA model, and obtain detection feature data through the Detect module.
[0075] In this embodiment, the steps of detecting the surface image to be detected are specifically as follows:
[0076] S2.1: The surface image to be detected is used as input information and input into the Focus layer of the YOLOv5_EMA model for slicing to obtain the first feature map. In this step, features are extracted in each 2x2 area through slicing, and these features are spliced together, thereby reducing the resolution of the image and increasing the number of channels to obtain the first feature map;
[0077] S2.2: The first feature map is fed into the C3_EMA module in the YOLOv5_EMA model, and features are extracted through the main branch and the EMA branch respectively, and the features are enhanced using the attention mechanism in the EMA branch to obtain the second feature map. In this step, the C3_EMA module has two parallel branches: the main branch and the EMA branch. The main branch uses standard convolution operations to extract features, while the EMA branch performs weighted enhancement on the features through the EMA attention mechanism. In the C3_EMA module, the outputs of the main branch and the EMA branch are concatenated in the channel dimension to obtain the second feature map;
[0078] S2.3: Extract features from the second feature map through the backbone network in the YOLOv5_EMA model, and fuse features at different scales through upsampling and feature concatenation to obtain a third feature map. In this step, the second feature map is further processed through a 1x1 convolutional layer to achieve deep fusion of features; the processed second feature map enters the backbone network in the YOLOv5_EMA model to extract features, and fuse features at different scales through upsampling and feature concatenation to obtain a third feature map.
[0079] S2.4: Input the third feature map into the Detect module in the YOLOv5_EMA model, and obtain detection feature data through the Detect module.
[0080] Optionally, in the present invention, the training set includes a plurality of surface damage images of wind turbines, and each type of surface damage image is multiple.
[0081] Optionally, step S3 of the present invention includes, according to the detection feature data, determining the probability of identifying the surface image to be detected of the wind turbine as a certain surface damage image;
[0082] If the probability that the surface image to be inspected of the wind turbine is identified as a certain surface damage image is greater than 70%, then the judgment result is that the corresponding wind turbine has this type of surface damage;
[0083] If the probability of identifying the to-be-detected surface image of the wind turbine as any type of surface damage image is less than or equal to 70%, the judgment result is that the corresponding wind turbine has no surface damage.
[0084] In this embodiment, the training set includes a variety of surface damage images of wind turbines, such as rust, corrosion, paint peeling, decomposition cracking or lightning strike. During use, the specific type of wind turbine surface damage can be judged to improve the inspection efficiency of wind turbine surface damage.
[0085] It should be noted that, for the probability standard of identifying the surface image to be detected of the wind turbine as a certain surface damage image, those skilled in the art can set it according to actual usage requirements, and the present invention only provides a more preferred example.
[0086] Furthermore, the wind turbine surface damage detection method of the present invention can also divide the damage risk levels according to different types of surface damage during use, so that maintenance personnel can perform corresponding maintenance operations according to different risk levels, achieve the effect of graded maintenance, and improve the inspection and maintenance efficiency of wind turbines.
[0087] Example 2
[0088] In a second aspect, the present invention further provides a wind turbine surface damage detection system, comprising:
[0089] An image acquisition unit, used for acquiring an image of a surface to be inspected of a wind turbine;
[0090] A processing unit is used to input the surface image to be detected of the wind turbine into the trained YOLOv5_EMA model for detection to obtain detection feature data; and according to the detection feature data, determine the probability of judging the surface image to be detected of the wind turbine as a surface damage image, and based on the probability, judge whether the wind turbine has surface damage;
[0091] The output unit outputs a judgment result on whether the wind turbine has surface damage.
[0092] The wind turbine surface damage detection system provided in this embodiment is used to detect the surface image of the wind turbine to obtain a judgment result on whether the wind turbine has surface damage, and output the judgment result to improve the inspection efficiency of the wind turbine surface damage.
[0093] It should be noted that the wind turbine surface damage detection system provided in this embodiment is completely corresponding to the wind turbine surface damage detection method in Example 1. The specific implementation content can be found in the content of Example 1 and will not be repeated here.
[0094] Furthermore, in this embodiment, the output unit also outputs the image of the wind turbine surface damage after annotating it, so that the user can intuitively obtain the specific conditions of the wind turbine surface damage and perform corresponding maintenance operations.
[0095] Example 3
[0096] In a third aspect, the present invention also provides a wind turbine surface damage detection device, the wind turbine surface damage detection device comprising a processing component and a storage component, the storage component storing instructions, the instructions being generated according to any one of the wind turbine surface damage detection methods in Example 1, and the processing component calling the instructions.
[0097] It should be noted that the wind turbine surface damage detection device provided in this embodiment is completely corresponding to the wind turbine surface damage detection method in Example 1. The specific implementation content can be found in the content of Example 1 and will not be repeated here.
[0098] For example, the wind turbine surface damage detection equipment in this embodiment can be mounted on a drone device. While the drone is used to capture images of the wind turbine surface, surface damage detection can be performed directly on the drone, and the detection result image can be fed back to the user through data transmission, thereby improving the inspection efficiency of wind turbine surface damage.
[0099] Example 4
[0100] In a fourth aspect, the present invention further provides a storage medium, wherein the storage medium stores instructions, wherein the instructions are generated according to any one of the wind turbine surface damage detection methods in Example 1.
[0101] Specifically, the storage component described in the present invention may include a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk or a CD-ROM. It should be noted that those skilled in the art may select the form and type of the storage component according to actual production and use requirements, and this is not further limited in this embodiment.
[0102] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.
Claims
1. A method for detecting surface damage of a wind turbine, characterized in that: The wind turbine surface damage detection method comprises: S1: Acquire the surface image of the wind turbine to be inspected; S2: Inputting the surface image of the wind turbine to be inspected into the trained YOLOv5_EMA model for inspection to obtain inspection feature data; wherein the trained YOLOv5_EMA model is generated based on the YOLOv5 model and the EMA attention module, and the trained YOLOv5_EMA model is formed by training with a training set, and the training set includes multiple images of surface damage of wind turbines; S3: Determine the probability of judging the to-be-detected surface image of the wind turbine as a surface damage image according to the detection feature data, and judge whether the corresponding wind turbine has surface damage based on the probability.
2. The wind turbine surface damage detection method according to claim 1, characterized in that: The steps of obtaining the trained YOLOv5_EMA model include: Establish a database of wind turbine surface damage images; Generate a training set based on a database of wind turbine surface damage images; The untrained YOLOv5_EMA model is constructed based on the YOLOv5 model and the EMA attention module; The untrained YOLOv5_EMA model is iteratively trained using the training set to form a trained YOLOv5_EMA model.
3. The method for detecting surface damage of a wind turbine according to claim 2, characterized in that: The step of obtaining the trained YOLOv5_EMA model also includes: Generate a validation set based on a database of wind turbine surface damage images; In the process of iterative training of the untrained YOLOv5_EMA model, verification is performed through the verification set. When the verification result tends to converge, the training is stopped to obtain the trained YOLOv5_EMA model.
4. The method for detecting surface damage of a wind turbine according to claim 2, characterized in that: According to the YOLOv5 model and EMA attention module, the specific steps to construct the untrained YOLOv5_EMA model are: A YOLOv5 benchmark model is constructed, and the EMA attention module is inserted into the first C3 module of the YOLOv5 benchmark model to form a new module C3_EMA module, and an untrained YOLOv5_EMA model is obtained; among them, the C3_EMA module includes a parallel main branch and EMA branch.
5. The method for detecting surface damage of a wind turbine according to claim 4, characterized in that: In step S2, the steps of detecting the surface image to be detected by using the trained YOLOv5_EMA model are as follows: S2.1: Input the surface image to be detected into the Focus layer in the YOLOv5_EMA model for slicing to obtain the first feature map; S2.2: The first feature map is fed into the C3_EMA module in the YOLOv5_EMA model, features are extracted through the main branch and the EMA branch respectively, and features are enhanced using the attention mechanism in the EMA branch to obtain a second feature map; S2.3: extract features from the second feature map through the backbone network in the YOLOv5_EMA model, and fuse features at different scales through upsampling and feature concatenation to obtain a third feature map; S2.4: Input the third feature map into the Detect module in the YOLOv5_EMA model, and obtain detection feature data through the Detect module.
6. The method for detecting surface damage of a wind turbine according to claim 1, characterized in that: The training set includes a variety of surface damage images of wind turbines, and each type of surface damage image has a plurality of images.
7. The method for detecting surface damage of a wind turbine according to claim 6, characterized in that: The step S3 includes determining, based on the detection feature data, the probability of identifying the surface image to be detected of the wind turbine as a certain surface damage image; If the probability that the surface image to be inspected of the wind turbine is identified as a certain surface damage image is greater than 70%, then the judgment result is that the corresponding wind turbine has this type of surface damage; If the probability of identifying the to-be-detected surface image of the wind turbine as any type of surface damage image is less than or equal to 70%, the judgment result is that the corresponding wind turbine has no surface damage.
8. A wind turbine surface damage detection system, characterized in that: The wind turbine surface damage detection system comprises: An image acquisition unit, used for acquiring an image of a surface to be inspected of a wind turbine; A processing unit is used to input the surface image to be detected of the wind turbine into the trained YOLOv5_EMA model for detection to obtain detection feature data; and according to the detection feature data, determine the probability of judging the surface image to be detected of the wind turbine as a surface damage image, and based on the probability, judge whether the wind turbine has surface damage; The output unit outputs a judgment result on whether the wind turbine has surface damage.
9. A wind turbine surface damage detection device, characterized in that: The wind turbine surface damage detection device comprises a processing component and a storage component, wherein the storage component stores instructions, wherein the instructions are generated according to the wind turbine surface damage detection method according to any one of claims 1 to 7, and the processing component calls the instructions.
10. A storage medium, characterized in that: The storage medium stores instructions, which are generated by the wind turbine surface damage detection method according to any one of claims 1 to 7.