Offshore wind power blade crack monitoring method and system based on YOLOv10 optimization model
By introducing a spatial attention mechanism layer and an optimized VIT structure into the YOLOv10 model, combined with the nonmaximum suppression method, the problems of high accuracy and real-time performance in offshore wind turbine blade monitoring were solved, the detection effect was improved, and a reliable guarantee was provided for the safe operation of wind farms.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE POWER INVESTMENT CORPORATION RESEARCH INSTITUTE
- Filing Date
- 2024-10-30
- Publication Date
- 2026-05-01
AI Technical Summary
Existing technologies are insufficient to meet the high precision and real-time requirements of offshore wind turbine blade monitoring, and traditional target detection algorithms have limited detection performance in complex backgrounds.
The YOLOv10 optimized model was used to monitor cracks in offshore wind turbine blades. By introducing a spatial attention mechanism layer and a VIT structure model, and combining the non-maximum suppression method to optimize the crack location box, the detection accuracy was improved.
This improves the accuracy and real-time performance of offshore wind turbine blade crack monitoring, providing a more reliable guarantee for the safe operation of wind farms.
Smart Images
Figure CN121962881A_ABST
Abstract
Description
A method and system for monitoring cracks in offshore wind turbine blades based on a YOLOv10 optimization model Technical Field
[0001] This application relates to the field of offshore wind turbine blade monitoring technology, and in particular to a method and system for monitoring cracks in offshore wind turbine blades based on the YOLOv10 optimization model. Background Technology
[0002] With the transformation of the global energy structure and the development of clean energy, offshore wind power, as an important component of renewable energy, is experiencing rapid development. However, offshore wind turbine blades, as key components of wind power generation systems, are exposed to the harsh marine environment for extended periods, making them susceptible to corrosion, wear, and extreme weather, leading to damage such as cracks. This damage not only reduces power generation efficiency but can also cause serious safety accidents. Therefore, regular and efficient monitoring and maintenance of offshore wind turbine blades is of paramount importance.
[0003] Traditional methods for monitoring offshore wind turbine blades primarily rely on manual inspections, but this approach suffers from low efficiency, high costs, and significant safety risks. In recent years, with the rapid development of computer vision, artificial intelligence, and drone technology, utilizing drones equipped with high-definition cameras and advanced target detection algorithms for online monitoring has become a new solution. This method not only improves monitoring efficiency and reduces labor costs but also effectively avoids the safety hazards associated with manual inspections.
[0004] In existing technologies, several offshore wind turbine blade monitoring solutions based on UAVs and computer vision have been proposed and applied. Most of these solutions employ traditional target detection algorithms, which have limited performance in complex environments and cannot meet the high precision and real-time requirements of offshore wind turbine blade monitoring. Summary of the Invention
[0005] This application provides a method and system for monitoring cracks in offshore wind turbine blades based on the YOLOv10 optimization model, in order to at least solve the technical problem that existing technologies cannot meet the high precision and real-time requirements of offshore wind turbine blade monitoring.
[0006] The first aspect of this application proposes a method for monitoring cracks in offshore wind turbine blades based on a YOLOv10 optimization model, the method comprising:
[0007] Obtain image information of the offshore wind turbine blades to be monitored;
[0008] The image information of the offshore wind turbine blade is input into the pre-trained blade crack monitoring model to obtain the location boxes of each crack and the crack type of the offshore wind turbine blade.
[0009] The non-maximum suppression method was used to optimize the crack location frames and crack types of the offshore wind turbine blades, resulting in optimized crack location frames for each crack type.
[0010] The crack information of the offshore wind turbine blade to be monitored is determined based on the optimized crack location frame corresponding to each crack type.
[0011] The blade crack monitoring model is a YOLOv10 optimized model, which is a model that introduces a spatial attention mechanism layer and a VIT structure into the YOLOv10 model.
[0012] Preferably, acquiring the image information of the offshore wind turbine blade to be monitored includes:
[0013] Based on a preset flight path and shooting angle, drones equipped with cameras are used to capture images of offshore wind turbine blades.
[0014] Furthermore, the training process of the blade crack monitoring model includes:
[0015] Acquire crack image data of multiple offshore wind turbine blades and annotate the crack images;
[0016] The crack image data of multiple labeled offshore wind turbine blades are preprocessed to obtain preprocessed crack image data.
[0017] The preprocessed crack image data are input into the initial YOLOv10 optimization model. The overall loss of classification loss, localization loss and confidence loss is used as the loss function of the model. The model is trained using the adaptive matrix estimation Adam optimization algorithm to obtain the trained YOLOv10 optimization model. The trained YOLOv10 optimization model is used as the blade crack monitoring model.
[0018] Furthermore, the preprocessing of the labeled crack image data of multiple offshore wind turbine blades to obtain preprocessed crack image data includes:
[0019] Enhancement processing was performed on the crack image data of multiple labeled offshore wind turbine blades to obtain an enhanced crack image dataset;
[0020] Each crack image in the crack image dataset is scaled up or down to a preset size, and then the scaled crack image data is normalized to obtain normalized crack image data.
[0021] Furthermore, the optimization of the crack location frames and crack types of the offshore wind turbine blades using the non-maximum suppression method to obtain optimized crack location frames corresponding to each crack type includes:
[0022] Determine the confidence level of each crack location box for the i-th crack type, filter out the crack location box with the highest confidence level among the i-th crack types, and use the crack location box with the highest confidence level as the first crack location box.
[0023] The intersection-union ratios of the first crack location frame and the remaining crack location frames in the i-th crack type are determined respectively, and the intersection-union ratios of each crack location frame in the remaining crack location frames are obtained.
[0024] Determine whether the cross-union ratio (CUI) of each crack location frame in the remaining crack location frames is greater than a preset CUI threshold. If so, retain the crack location frame; otherwise, delete the crack location frame to obtain the optimized crack location frame corresponding to the i-th crack type.
[0025] Where i∈[1, 2, ..., N], and N is the total number of crack types.
[0026] Furthermore, determining the crack information of the offshore wind turbine blade to be monitored based on the optimized crack location boxes corresponding to each crack type includes:
[0027] The optimized crack location boxes corresponding to each crack type are drawn on the image information of the offshore wind turbine blade to obtain the optimized image information of the offshore wind turbine blade.
[0028] Based on the optimized image information of the offshore wind turbine blade, the length and depth of each crack on the offshore wind turbine blade are determined, and the severity level of each crack is determined based on the length and depth of each crack.
[0029] The crack information includes: crack length, depth, and severity level;
[0030] The severity levels include: Level 1 Severity, Level 2 Severity, and Level 3 Severity.
[0031] Furthermore, the method also includes:
[0032] Determine whether the number of third-degree severity levels among the severity levels of each crack is greater than or equal to 1. If so, issue an alarm and send crack information to the staff.
[0033] Furthermore, the method also includes:
[0034] The images, crack information, and alarm records of the offshore wind turbine blades are stored in cloud storage.
[0035] The second aspect of this application proposes a crack monitoring system for offshore wind turbine blades based on the YOLOv10 optimization model, comprising:
[0036] The acquisition module is used to acquire image information of the offshore wind turbine blades to be monitored;
[0037] The positioning module is used to input the image information of the offshore wind turbine blade into the pre-trained blade crack monitoring model to obtain the positioning frames of each crack and the crack type of the offshore wind turbine blade.
[0038] The optimization module is used to optimize the crack location frames and crack types of the offshore wind turbine blades using the non-maximum suppression method, so as to obtain the optimized crack location frames corresponding to each crack type.
[0039] The determination module is used to determine the crack information of the offshore wind turbine blade to be monitored based on the optimized crack location frame corresponding to each crack type.
[0040] The blade crack monitoring model is a YOLOv10 optimized model, which is a model that introduces a spatial attention mechanism layer and a VIT structure into the YOLOv10 model.
[0041] A third aspect of this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.
[0042] The technical solutions provided by the embodiments of this application bring at least the following beneficial effects:
[0043] This application proposes a method and system for monitoring cracks in offshore wind turbine blades based on a YOLOv10 optimized model. The method includes: acquiring image information of the offshore wind turbine blade to be monitored; inputting the image information of the offshore wind turbine blade into a pre-trained blade crack monitoring model to obtain the location boxes and crack types of each crack in the offshore wind turbine blade; optimizing the location boxes and crack types of each crack in the offshore wind turbine blade using the non-maximum suppression method to obtain optimized location boxes corresponding to each crack type; and determining the crack information of the offshore wind turbine blade to be monitored based on the optimized location boxes corresponding to each crack type. The blade crack monitoring model is a YOLOv10 optimized model, which is a model that introduces a spatial attention mechanism layer and a VIT structure into the YOLOv10 model. The technical solution proposed in this application improves monitoring accuracy and provides a more reliable guarantee for the safe operation of wind farms.
[0044] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description
[0045] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:
[0046] Figure 1 is a flowchart of a method for monitoring cracks in offshore wind turbine blades based on a YOLOv10 optimization model according to an embodiment of this application;
[0047] Figure 2 is a first structural diagram of a marine wind turbine blade crack monitoring system based on the YOLOv10 optimization model provided in an embodiment of this application;
[0048] Figure 3 is a second structural diagram of a marine wind turbine blade crack monitoring system based on the YOLOv10 optimization model provided according to an embodiment of this application;
[0049] Figure 4 is a third structural diagram of a marine wind turbine blade crack monitoring system based on the YOLOv10 optimization model provided in an embodiment of this application;
[0050] Figure 5 is a fourth structural diagram of a crack monitoring system for offshore wind turbine blades based on the YOLOv10 optimization model provided in an embodiment of this application. Detailed Implementation
[0051] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.
[0052] This application proposes a method and system for monitoring cracks in offshore wind turbine blades based on a YOLOv10 optimized model. The method includes: acquiring image information of the offshore wind turbine blade to be monitored; inputting the image information of the offshore wind turbine blade into a pre-trained blade crack monitoring model to obtain the location boxes and crack types of each crack in the offshore wind turbine blade; optimizing the location boxes and crack types of each crack in the offshore wind turbine blade using the non-maximum suppression method to obtain optimized location boxes corresponding to each crack type; and determining the crack information of the offshore wind turbine blade to be monitored based on the optimized location boxes corresponding to each crack type. The blade crack monitoring model is a YOLOv10 optimized model, which is a model that introduces a spatial attention mechanism layer and a VIT structure into the YOLOv10 model. The technical solution proposed in this application improves the monitoring effect and provides a more reliable guarantee for the safe operation of wind farms.
[0053] The following description, with reference to the accompanying drawings, illustrates an embodiment of this application of a method and system for monitoring cracks in offshore wind turbine blades based on a YOLOv10 optimization model.
[0054] Example 1
[0055] Figure 1 is a flowchart of a method for monitoring cracks in offshore wind turbine blades based on a YOLOv10 optimization model according to an embodiment of this application. As shown in Figure 1, the method includes:
[0056] Step 1: Obtain image information of the offshore wind turbine blades to be monitored.
[0057] In this embodiment of the disclosure, step 1 specifically includes:
[0058] Based on a preset flight path and shooting angle, drones equipped with cameras are used to capture images of offshore wind turbine blades.
[0059] It should be noted that the flight route and shooting angle of the drone are planned according to factors such as the specific location of the offshore wind farm, weather conditions, and blade distribution. The drone is equipped with a high-definition camera to take high-definition pictures of the offshore wind turbine blades to capture minute cracks on the blade surface.
[0060] Step 2: Input the image information of the offshore wind turbine blade into the pre-trained blade crack monitoring model to obtain the location boxes of each crack and the crack type of the offshore wind turbine blade.
[0061] It should be noted that the blade crack monitoring model is a YOLOv10 optimized model, which is a model that introduces a spatial attention mechanism layer and a VIT structure into the YOLOv10 model.
[0062] The training process of the blade crack monitoring model includes:
[0063] Acquire crack image data of multiple offshore wind turbine blades and annotate the crack images;
[0064] The crack image data of multiple labeled offshore wind turbine blades are preprocessed to obtain preprocessed crack image data.
[0065] The preprocessed crack image data are input into the initial YOLOv10 optimization model. The overall loss of classification loss, localization loss and confidence loss is used as the loss function of the model. The model is trained using the adaptive matrix estimation Adam optimization algorithm to obtain the trained YOLOv10 optimization model. The trained YOLOv10 optimization model is used as the blade crack monitoring model.
[0066] Furthermore, the preprocessing of the labeled crack image data of multiple offshore wind turbine blades to obtain preprocessed crack image data includes:
[0067] Enhancement processing was performed on the crack image data of multiple labeled offshore wind turbine blades to obtain an enhanced crack image dataset;
[0068] Each crack image in the crack image dataset is scaled up or down to a preset size, and then the scaled crack image data is normalized to obtain normalized crack image data.
[0069] It should be noted that a large amount of image data of cracked offshore wind turbine blades was collected and prepared. This data was precisely labeled and used for model learning. During the training phase, the images undergo a series of preprocessing operations. The images are resized to match the model's input requirements, and bilinear interpolation is used to scale the images to a specified size of 416x416 pixels while maintaining the aspect ratio. Next, the images are normalized, scaling the pixel values to a smaller range to accelerate the training process and improve model convergence speed.
[0070] Simultaneously, to increase the diversity of training data and improve the model's generalization ability, the images undergo a series of enhancement operations. These operations include random cropping and scaling, flipping and rotating, color space transformation (converting the image from RGB to HSV color space), adding Gaussian noise, and blending enhancements. These enhancement methods generate new image samples, enabling the model to better adapt to crack detection tasks under different scenes and lighting conditions. The enhanced and preprocessed images are then converted to a format that the model can process, ensuring that their data type is consistent with the model's requirements.
[0071] The processed image is then fed into the YOLOv10 backbone network, where SAM and VIT are applied to the feature map generation process to highlight important spatial regions, improve the accuracy of object detection, and enable the handling of object detection tasks in more different environments. The generated feature maps are then fed into the detection head, which uses these feature maps to predict the bounding boxes and categories of the objects.
[0072] It should be noted that the original YOLOv10 model, by introducing a Spatial Attention Mode (SAM) layer, focuses on the spatial dimension of the image. By assigning different weights to each spatial location, it enhances the model's attention to key regions and improves detection accuracy. At the same time, a VIT structure is added to YOLOv10, which uses the VIT's multi-head self-attention mechanism to capture the global contextual information of the image. Combined with YOLOv10's local feature detection capabilities, it improves the model's adaptability to complex scenes. The optimized YOLOv10 is then deployed on a cloud server, enabling efficient real-time detection. After the image is acquired by the drone, it is processed quickly and the crack detection results can be fed back without delay.
[0073] Step 3: Optimize the crack location frames and crack types of the offshore wind turbine blades using the non-maximum suppression method to obtain the optimized crack location frames corresponding to each crack type.
[0074] In this embodiment of the disclosure, step 3 specifically includes:
[0075] Determine the confidence level of each crack location box for the i-th crack type, filter out the crack location box with the highest confidence level among the i-th crack types, and use the crack location box with the highest confidence level as the first crack location box.
[0076] The intersection-union ratios of the first crack location frame and the remaining crack location frames in the i-th crack type are determined respectively, and the intersection-union ratios of each crack location frame in the remaining crack location frames are obtained.
[0077] Determine whether the cross-union ratio (CUI) of each crack location frame in the remaining crack location frames is greater than a preset CUI threshold. If so, retain the crack location frame; otherwise, delete the crack location frame to obtain the optimized crack location frame corresponding to the i-th crack type.
[0078] Where i∈[1, 2, ..., N], and N is the total number of crack types.
[0079] It should be noted that post-processing steps such as Non-Maximum Suppression (NMS) are used to optimize the detection results. First, for each category (longitudinal crack, transverse crack, fatigue crack, etc.), the bounding boxes are sorted according to their confidence levels. Then, the bounding box with the highest confidence level is selected as the current maximum candidate, i.e., the first crack location box. Next, the intersection-union ratio (IUR) of this maximum candidate with all other bounding boxes is calculated. If the IUR of a bounding box with the maximum candidate is higher than a preset threshold, it is removed from the candidate list to avoid duplicate detection. Finally, this process is repeated until all bounding boxes have been processed. Duplicate bounding boxes are removed, and the best detection result is retained.
[0080] Step 4: Determine the crack information of the offshore wind turbine blade to be monitored based on the optimized crack location box corresponding to each crack type.
[0081] In this embodiment of the disclosure, step 4 specifically includes:
[0082] The optimized crack location boxes corresponding to each crack type are drawn on the image information of the offshore wind turbine blade to obtain the optimized image information of the offshore wind turbine blade.
[0083] Based on the optimized image information of the offshore wind turbine blade, the length and depth of each crack on the offshore wind turbine blade are determined, and the severity level of each crack is determined based on the length and depth of each crack.
[0084] The crack information includes: crack length, depth, and severity level;
[0085] The severity levels include: Level 1 Severity, Level 2 Severity, and Level 3 Severity.
[0086] It should be noted that the crack inspection results plotted on the original images are analyzed in depth to assess the severity of the cracks and their impact on the wind turbine blades. Minor cracks, or first-degree severity cracks, are short and shallow, typically appearing on the blade surface. They have little impact on the structural integrity of the blade but may cause localized stress concentration and potentially propagate over time. Moderate cracks, or second-degree severity cracks, are of moderate length and depth, possibly penetrating part of the blade's thickness, affecting the blade's local strength and stiffness, and potentially increasing deformation under high loads. Severe cracks, or third-degree severity cracks, are long and deep, possibly penetrating the entire thickness of the blade, severely impacting its overall structural strength and stiffness, and potentially causing breakage or failure during operation. Based on the analysis results, intelligent decision support is provided to wind farm maintenance personnel to develop repair or replacement plans.
[0087] Furthermore, the method also includes:
[0088] Determine whether the number of third-degree severity levels among the severity levels of each crack is greater than or equal to 1. If so, issue an alarm and send crack information to the staff.
[0089] It should be noted that when a severe crack, i.e. a crack of the third severity level, or a crack that exceeds the safe range is detected, an alarm mechanism is automatically triggered, and the wind farm maintenance personnel or managers are notified in a timely manner through various means such as SMS, email, and application to handle the situation.
[0090] Furthermore, the method also includes:
[0091] The images, crack information, and alarm records of the offshore wind turbine blades are stored in cloud storage.
[0092] It should be noted that the collected image data, detection results, alarm records, etc. are stored in cloud storage to help wind farm managers fully understand the blade crack situation and optimize maintenance strategies.
[0093] The UAV-based online monitoring method for offshore wind turbine blade cracks proposed in this invention, based on an optimized YOLOv10 model, boasts higher detection accuracy and real-time performance. YOLOv10, the latest version in the YOLO series, achieves more efficient and accurate real-time target detection through improved processing and model structure design. Compared to the original YOLOv10, the optimized YOLOv10 model in this invention enhances the model's ability to capture and process image features and improves the system's adaptability to complex scenarios by introducing an enhanced attention mechanism layer and a VIT structure, thereby increasing the accuracy of blade crack detection. Applying this method to offshore wind turbine blade crack monitoring can further improve monitoring effectiveness and provide a more reliable guarantee for the safe operation of wind farms.
[0094] In summary, the offshore wind turbine blade crack monitoring method based on the YOLOv10 optimization model proposed in this embodiment improves monitoring accuracy and provides a more reliable guarantee for the safe operation of wind farms.
[0095] Example 2
[0096] Figure 2 is a structural diagram of a crack monitoring system for offshore wind turbine blades based on a YOLOv10 optimization model according to an embodiment of this application. As shown in Figure 2, the system includes:
[0097] The acquisition module 100 is used to acquire image information of the offshore wind turbine blades to be monitored;
[0098] The positioning module 200 is used to input the image information of the offshore wind turbine blade into the pre-trained blade crack monitoring model to obtain the positioning frames of each crack and the crack type of the offshore wind turbine blade.
[0099] The blade crack monitoring model is a YOLOv10 optimized model, which is a model that introduces a spatial attention mechanism layer and a VIT structure into the YOLOv10 model.
[0100] The optimization module 300 is used to optimize the crack location frame and crack type of the offshore wind turbine blade using the non-maximum suppression method, so as to obtain the optimized crack location frame corresponding to each crack type.
[0101] The determination module 400 is used to determine the crack information of the offshore wind turbine blade to be monitored based on the optimized crack location frame corresponding to each crack type.
[0102] In this embodiment of the disclosure, the acquisition module 100 is further configured to:
[0103] Based on a preset flight path and shooting angle, drones equipped with cameras are used to capture images of offshore wind turbine blades.
[0104] In this embodiment of the disclosure, as shown in FIG3, the system further includes: a training module 500, used for:
[0105] Acquire crack image data of multiple offshore wind turbine blades and annotate the crack images;
[0106] The crack image data of multiple labeled offshore wind turbine blades are preprocessed to obtain preprocessed crack image data.
[0107] The preprocessed crack image data are input into the initial YOLOv10 optimization model. The overall loss of classification loss, localization loss and confidence loss is used as the loss function of the model. The model is trained using the adaptive matrix estimation Adam optimization algorithm to obtain the trained YOLOv10 optimization model. The trained YOLOv10 optimization model is used as the blade crack monitoring model.
[0108] The preprocessing of the labeled crack image data of multiple offshore wind turbine blades to obtain preprocessed crack image data includes:
[0109] Enhancement processing was performed on the crack image data of multiple labeled offshore wind turbine blades to obtain an enhanced crack image dataset;
[0110] Each crack image in the crack image dataset is scaled up or down to a preset size, and then the scaled crack image data is normalized to obtain normalized crack image data.
[0111] In this embodiment of the disclosure, the optimization module 300 is further configured to:
[0112] Determine the confidence level of each crack location box for the i-th crack type, filter out the crack location box with the highest confidence level among the i-th crack types, and use the crack location box with the highest confidence level as the first crack location box.
[0113] The intersection-union ratios of the first crack location frame and the remaining crack location frames in the i-th crack type are determined respectively, and the intersection-union ratios of each crack location frame in the remaining crack location frames are obtained.
[0114] Determine whether the cross-union ratio (CUI) of each crack location frame in the remaining crack location frames is greater than a preset CUI threshold. If so, retain the crack location frame; otherwise, delete the crack location frame to obtain the optimized crack location frame corresponding to the i-th crack type.
[0115] Where i∈[1, 2, ..., N], and N is the total number of crack types.
[0116] In this embodiment of the disclosure, the determining module 400 is further configured to:
[0117] The optimized crack location boxes corresponding to each crack type are drawn on the image information of the offshore wind turbine blade to obtain the optimized image information of the offshore wind turbine blade.
[0118] Based on the optimized image information of the offshore wind turbine blade, the length and depth of each crack on the offshore wind turbine blade are determined, and the severity level of each crack is determined based on the length and depth of each crack.
[0119] The crack information includes: crack length, depth, and severity level;
[0120] The severity levels include: Level 1 Severity, Level 2 Severity, and Level 3 Severity.
[0121] In this embodiment of the disclosure, as shown in FIG4, the system further includes: an alarm module 600, used for:
[0122] Determine whether the number of third-degree severity levels among the severity levels of each crack is greater than or equal to 1. If so, issue an alarm and send crack information to the staff.
[0123] In this embodiment of the disclosure, as shown in FIG5, the system further includes: a storage module 700, used for:
[0124] The images, crack information, and alarm records of the offshore wind turbine blades are stored in cloud storage.
[0125] In summary, the offshore wind turbine blade crack monitoring system based on the YOLOv10 optimization model proposed in this embodiment improves monitoring accuracy and provides a more reliable guarantee for the safe operation of wind farms.
[0126] Example 3
[0127] To implement the above embodiments, this disclosure also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in Embodiment 1.
[0128] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0129] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.
[0130] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.
Claims
1. A method for monitoring cracks in offshore wind turbine blades based on a YOLOv10 optimization model, characterized in that, The method includes: acquiring image information of the offshore wind turbine blade to be monitored; inputting the image information of the offshore wind turbine blade into a pre-trained blade crack monitoring model to obtain the crack location boxes and crack types of the offshore wind turbine blade; optimizing the crack location boxes and crack types of the offshore wind turbine blade using the non-maximum suppression method to obtain optimized crack location boxes corresponding to each crack type; determining the crack information of the offshore wind turbine blade to be monitored based on the optimized crack location boxes corresponding to each crack type; wherein, the blade crack monitoring model is a YOLOv10 optimized model, and the YOLOv10 optimized model is a model that introduces a spatial attention mechanism layer and a VIT structure into the YOLOv10 model.
2. The method as described in claim 1, characterized in that, The acquisition of image information of the offshore wind turbine blades to be monitored includes: capturing image information of the offshore wind turbine blades using a drone equipped with a camera, based on a preset flight route and shooting angle.
3. The method as described in claim 2, characterized in that, The training process of the blade crack monitoring model includes: acquiring crack image data of multiple offshore wind turbine blades and labeling the crack images; preprocessing the labeled crack image data of multiple offshore wind turbine blades to obtain preprocessed crack image data; inputting the preprocessed crack image data into an initial YOLOv10 optimization model, using the overall loss of classification loss, localization loss and confidence loss as the model's loss function, and training the model using the adaptive matrix estimation Adam optimization algorithm to obtain a trained YOLOv10 optimization model, and using the trained YOLOv10 optimization model as the blade crack monitoring model.
4. The method as described in claim 3, characterized in that, The step of preprocessing the labeled crack image data of multiple offshore wind turbine blades to obtain preprocessed crack image data includes: enhancing the labeled crack image data of multiple offshore wind turbine blades to obtain an enhanced crack image dataset; scaling each crack image data in the crack image dataset according to a preset size, and normalizing the scaled crack image data to obtain normalized crack image data.
5. The method as described in claim 4, characterized in that, The optimization of crack location frames and crack types of the offshore wind turbine blade using the non-maximum suppression method to obtain optimized crack location frames for each crack type includes: determining the confidence level of each crack location frame for the i-th crack type; selecting the crack location frame with the highest confidence level among the i-th crack types and using it as the first crack location frame; determining the intersection-union ratio (CUNR) of the first crack location frame and the remaining crack location frames in the i-th crack type to obtain the CUNR of each crack location frame in the remaining crack location frames; determining whether the CUNR of each crack location frame in the remaining crack location frames is greater than a preset CUNR threshold; if so, retaining the crack location frame; otherwise, deleting the crack location frame to obtain the optimized crack location frame for the i-th crack type; where i ∈ [1, 2, ..., N], and N is the total number of crack types.
6. The method as described in claim 5, characterized in that, The step of determining the crack information of the offshore wind turbine blade to be monitored based on the optimized crack location boxes corresponding to each crack type includes: drawing the optimized crack location boxes corresponding to each crack type on the image information of the offshore wind turbine blade to obtain optimized image information of the offshore wind turbine blade; determining the length and depth of each crack on the offshore wind turbine blade based on the optimized image information of the offshore wind turbine blade, and determining the severity level of each crack based on the length and depth of each crack; wherein, the crack information includes: crack length, depth and severity level; the severity level includes: first severity level, second severity level and third severity level.
7. The method as described in claim 6, characterized in that, The method further includes: determining whether the number of third-degree severity levels among the severity levels of each crack is greater than or equal to 1; if so, issuing an alarm and sending crack information to the staff.
8. The method as described in claim 7, characterized in that, The method also includes storing the image information, crack information, and alarm records of the offshore wind turbine blades in cloud storage.
9. A crack monitoring system for offshore wind turbine blades based on a YOLOv10 optimization model, characterized in that, The system includes: an acquisition module for acquiring image information of the offshore wind turbine blade to be monitored; a positioning module for inputting the image information of the offshore wind turbine blade into a pre-trained blade crack monitoring model to obtain the location boxes and crack types of each crack of the offshore wind turbine blade; an optimization module for optimizing the location boxes and crack types of each crack of the offshore wind turbine blade using the non-maximum suppression method to obtain the optimized location boxes corresponding to each crack type; and a determination module for determining the crack information of the offshore wind turbine blade to be monitored based on the optimized location boxes corresponding to each crack type. The blade crack monitoring model is a YOLOv10 optimized model, which is a model that introduces a spatial attention mechanism layer and a VIT structure into the YOLOv10 model.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-8.