A wind blade defect detection system based on event camera driving

Abstract

The application discloses a kind of wind force blade defect detection systems based on event camera driving, belong to wind power generation equipment state monitoring and intelligent detection technical field.The system includes unmanned aerial vehicle control unit, event sensing unit, data processing unit, defect detection unit and result output unit.By being equipped with the unmanned aerial vehicle of event camera along the preset route cruise, the visual event stream of blade surface is collected, and the time alignment with IMU data is realized by means of hardware synchronization mode.The system filters event stream, motion compensation and feature construction, generates event feature map and inputs neural network model, completes the classification, positioning and three-dimensional coordinate mapping of defect, and finally outputs the visualization report containing the multi-dimensional information of defect.The application effectively overcomes the limitations of traditional visual detection under high-speed motion and complex illumination, realizes efficient, accurate identification and quantitative evaluation of wind force blade defect.

Description

Technical Field

[0001] This invention belongs to the field of wind power equipment condition monitoring and intelligent detection technology, specifically relating to a wind turbine blade defect detection system based on event camera driving. Background Technology

[0002] Wind turbine blades are the core components for capturing wind energy, and their structural integrity directly determines the unit's power generation efficiency and operational safety. Exposed to extreme outdoor environments for extended periods, blades are subject to wind and sand erosion, alternating loads, ultraviolet aging, and strong sunlight radiation, making them prone to defects such as surface cracks, localized corrosion, and gel coat peeling. Industry statistics indicate that single-incident breakage accidents caused by undetected blade defects can result in economic losses of several million yuan, and blade failures account for more than 30% of total wind farm downtime, making them a key bottleneck restricting the reliability of wind power equipment.

[0003] Current blade defect detection technologies primarily rely on traditional visual sensors (frame exposure cameras) or single-modal devices, which have the following core limitations: Traditional frame exposure cameras are prone to motion blur due to the mismatch between exposure time and movement speed during high-speed drone inspections, leading to missed detection of minute cracks as small as 1mm. Furthermore, their dynamic range cannot cover the brightness difference between the "strong light reflection area" and the "shadow area" on the blade surface; overexposure or underexposure can completely obscure defect features. While 3D sensing devices such as LiDAR can acquire depth information, their feature recognition of static defects (such as stable corrosion areas) is low, failing to distinguish between real defects and surface stains, requiring secondary manual interpretation. Single-blade inspection can take over 60 minutes, resulting in low efficiency. Current technologies for fusing visual and 3D data often remain at a simple stitching level, lacking a dynamic correlation mechanism for defect features, leading to inaccuracies in the "2D positioning - 3D quantization" process, making it difficult to support maintenance decisions.

[0004] Event cameras, as a novel bio-inspired sensor, output asynchronous event streams (including spatial coordinates, timestamps, and brightness change polarity) only when pixel brightness changes within a preset threshold. They offer advantages such as high dynamic range, microsecond-level temporal resolution, no motion blur, and low data redundancy, providing a new paradigm for solving defect detection challenges in complex environments. However, event cameras output unstructured event streams, and mature solutions for feature extraction, dynamic correlation with 3D data, and long-term tracking mechanisms are still lacking. Especially in dynamic scenarios with curved surfaces like wind turbine blades, the mapping relationship between event features and the physical properties of defects still requires further development.

[0005] Therefore, developing a defect detection system and method based on event camera, combined with virtual-real fusion perception, dynamic feature enhancement and multimodal fusion verification technology, to break through the environmental adaptability and quantitative accuracy bottlenecks of traditional detection has become a key technical problem that urgently needs to be solved in the field of wind power operation and maintenance. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, this invention provides a wind turbine blade defect detection system based on event camera driving. The objective of this invention can be achieved through the following technical solutions:

[0007] A wind turbine blade defect detection system based on event camera-driven operation, comprising:

[0008] The UAV control unit is used to cruise based on a preset spiral route via the UAV positioning module and the lidar obstacle avoidance module.

[0009] The event sensing unit includes an event camera module, an IMU inertial measurement module, and a synchronization module. The event camera module acquires visual event streams on the blade surface through a dynamic visual sensor. The IMU inertial measurement module acquires inertial motion data of the UAV. The synchronization module achieves time synchronization between the event stream and the IMU data through hardware triggering.

[0010] The data processing unit is used to filter the event stream, perform motion compensation and feature construction, and output an event feature map.

[0011] The defect detection unit classifies and locates defects in the event feature map, and processes the event feature map using a neural network model to obtain the defect type, confidence level, and two-dimensional bounding box.

[0012] The result output unit is used to display the detection results in real time and map the two-dimensional position of the defect to the three-dimensional coordinate system of the blade to obtain the three-dimensional coordinates and size information, generating a visualization report containing multi-dimensional information about the defect.

[0013] Specifically, the event sensing unit also includes an adaptive laser illumination module, which dynamically adjusts the power according to the event density to maintain the integrity of event characteristics in low-light environments.

[0014] Specifically, the motion compensation of the data processing unit performs rotation and translation transformations on the event coordinates based on the attitude data provided by the IMU inertial measurement module to compensate for image offset caused by the movement of the UAV.

[0015] Specifically, the motion compensation of the data processing unit is also used to periodically perform cumulative error calibration based on fixed features of the wind turbine blade surface.

[0016] Specifically, the feature construction of the data processing unit adopts a dynamic time window strategy to accumulate asynchronous event streams and convert them into multi-channel event feature maps, which include time surfaces, polarity histograms, and event density information.

[0017] Specifically, the defect detection unit employs a neural network model that includes an event feature attention mechanism to enhance defect-related features and suppress background features.

[0018] Specifically, when the synchronization module detects that the event density in the static area is continuously lower than the threshold, it triggers the modulation light source to generate an event stream by exciting brightness fluctuations, thereby enhancing the event response of static defects.

[0019] Specifically, the result output module further includes: predicting the growth trend of defect size based on historical inspection data and outputting graded maintenance suggestions.

[0020] Specifically, the event feature map includes multiple information channels, including at least: a time surface channel with the most recent event timestamp as the pixel value, a polarity histogram channel that records the counts of positive and negative polarity events, and an event density map channel that reflects the number of events per unit area.

[0021] Specifically, the hardware triggering method is as follows:

[0022] The IMU inertial measurement module outputs a periodic hardware pulse signal to the synchronization module, and the synchronization module adds a uniform timestamp to each event acquired by the event camera module based on the pulse signal.

[0023] Furthermore, the event feature attention mechanism includes:

[0024] During the feature extraction stage of the network, the event density map of each candidate region on the feature map is calculated and input in real time. The event density map is obtained based on the event density information channel generated by the data processing unit.

[0025] For any spatial location on a feature map, its attention weight is dynamically determined by the event density corresponding to the location and the average event density of its neighborhood background.

[0026] The calculated attention weights are multiplied element-wise with the original feature map along the channel dimension to enhance the features of high-response defect regions and suppress the features of low-response background regions.

[0027] The parameters of the event feature attention mechanism are trained together with the neural network model. The optimization goal is to introduce a weight sparsity regularization term into the classification and localization loss function, so that the network learns to focus on regions with significantly abnormal event density during training.

[0028] Furthermore, the visualization report generated by the result output unit also includes a structured defect list generated according to the standard operation and maintenance specifications for wind turbine blades. The list includes at least: blade number, detection time, defect type, three-dimensional location coordinates, equivalent two-dimensional size, estimated depth, confidence level, and recommended maintenance priority.

[0029] The beneficial effects of this invention are as follows:

[0030] By employing an event camera with high dynamic range and microsecond-level resolution, motion blur and overexposure / underexposure problems caused by traditional frame cameras in high-speed drone cruising and strong light / backlight environments are fundamentally avoided. Combined with IMU-based real-time motion compensation and periodic feature calibration, pixel shift and cumulative errors introduced by carrier motion are effectively suppressed, ensuring a high detection rate and 3D positioning accuracy for minute defects such as cracks and corrosion even in dynamic scenes.

[0031] The system innovatively integrates an active excitation mechanism with adaptive laser illumination. When the event density in a static area is detected to be too low, the system automatically triggers a modulation light source to excite a change in brightness, transforming the "invisibility" of static defects into the "detectability" of the event flow, thereby increasing the generation rate of static defect events. The adaptive illumination is dynamically adjusted according to the ambient light intensity, ensuring the integrity and consistency of event characteristics under low light conditions such as dusk and shadows, thus achieving all-weather defect perception.

[0032] Based on an asynchronous sensing mechanism using event cameras, the system outputs only brightness change information, significantly reducing data volume compared to traditional video streams and greatly alleviating transmission and storage pressure. The system not only performs real-time defect identification and localization but also accurately correlates deep features such as event density and spatiotemporal patterns with defect types through multi-channel event feature maps and an improved YOLO network with integrated attention mechanisms. The output module further provides a defect growth trend prediction report including 3D quantified dimensions and depth, as well as historical data. This report can be directly integrated with the wind farm asset health management system, providing a structured and actionable data foundation for full lifecycle health management, from detection to early warning and maintenance planning. Attached Figure Description

[0033] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0034] Figure 1 This is a system architecture diagram of the present invention;

[0035] Figure 2 This is a schematic diagram of the hardware layout of the event sensing unit in this invention. Detailed Implementation

[0036] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of example embodiments to those skilled in the art. Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring aspects of this disclosure. The blocks shown in the drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices. The flowcharts shown in the drawings are merely illustrative and do not necessarily include all contents and operations / steps, nor do they necessarily have to be performed in the order described. For example, some operations / steps can be broken down, while others can be combined or partially combined. Therefore, the actual execution order may change depending on the actual situation.

[0037] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0038] Please see Figures 1-2 A wind turbine blade defect detection system based on event camera-driven operation, comprising:

[0039] The UAV control unit is used to cruise based on a preset spiral route via the UAV positioning module and the lidar obstacle avoidance module.

[0040] like Figure 2 As shown, the event sensing unit includes an event camera module, an IMU inertial measurement module, and a synchronization module. The event camera module acquires visual event streams on the blade surface through a dynamic visual sensor. The IMU inertial measurement module acquires the inertial motion data of the UAV. The synchronization module achieves time synchronization between the event stream and the IMU data through hardware triggering.

[0041] The data processing unit is used to filter the event stream, perform motion compensation and feature construction, and output an event feature map.

[0042] The defect detection unit classifies and locates defects in the event feature map, and processes the event feature map using a neural network model to obtain the defect type, confidence level, and two-dimensional bounding box.

[0043] The result output unit is used to display the detection results in real time and map the two-dimensional position of the defect to the three-dimensional coordinate system of the blade to obtain the three-dimensional coordinates and size information, generating a visualization report containing multi-dimensional information about the defect.

[0044] Specifically, the event sensing unit also includes an adaptive laser illumination module, which dynamically adjusts the power according to the event density to maintain the integrity of event characteristics in low-light environments.

[0045] Specifically, the motion compensation of the data processing unit performs rotation and translation transformations on the event coordinates based on the attitude data provided by the IMU inertial measurement module to compensate for image offset caused by the movement of the UAV.

[0046] Specifically, the motion compensation of the data processing unit is also used to periodically perform cumulative error calibration based on fixed features of the wind turbine blade surface.

[0047] Specifically, the feature construction of the data processing unit adopts a dynamic time window strategy to accumulate asynchronous event streams and convert them into multi-channel event feature maps, which include time surfaces, polarity histograms, and event density information.

[0048] Specifically, the defect detection unit employs a neural network model that includes an event feature attention mechanism to enhance defect-related features and suppress background features.

[0049] Specifically, when the synchronization module detects that the event density in the static area is continuously lower than the threshold, it triggers the modulation light source to generate an event stream by exciting brightness fluctuations, thereby enhancing the event response of static defects.

[0050] Specifically, the result output module further includes: predicting the growth trend of defect size based on historical inspection data and outputting graded maintenance suggestions.

[0051] Specifically, the event feature map includes multiple information channels, including at least: a time surface channel with the most recent event timestamp as the pixel value, a polarity histogram channel that records the counts of positive and negative polarity events, and an event density map channel that reflects the number of events per unit area.

[0052] Specifically, the hardware triggering method is as follows:

[0053] The IMU inertial measurement module outputs a periodic hardware pulse signal to the synchronization module, and the synchronization module adds a uniform timestamp to each event acquired by the event camera module based on the pulse signal.

[0054] Furthermore, the event feature attention mechanism includes:

[0055] During the feature extraction stage of the network, the event density map of each candidate region on the feature map is calculated and input in real time. The event density map is obtained based on the event density information channel generated by the data processing unit.

[0056] For any spatial location on a feature map, its attention weight is dynamically determined by the event density corresponding to the location and the average event density of its neighborhood background.

[0057] The calculated attention weights are multiplied element-wise with the original feature map along the channel dimension to enhance the features of high-response defect regions and suppress the features of low-response background regions.

[0058] The parameters of the event feature attention mechanism are trained together with the neural network model. The optimization goal is to introduce a weight sparsity regularization term into the classification and localization loss function, so that the network learns to focus on regions with significantly abnormal event density during training.

[0059] Furthermore, the visualization report generated by the result output unit also includes a structured defect list generated according to the standard operation and maintenance specifications for wind turbine blades. The list includes at least: blade number, detection time, defect type, three-dimensional location coordinates, equivalent two-dimensional size, estimated depth, confidence level, and recommended maintenance priority.

[0060] In this embodiment, a wind turbine blade defect detection method and system based on event camera driving is presented, which aims to solve the problems of motion blur, insufficient dynamic range, and high defect false negative rate in traditional visual inspection under complex environments such as high-speed movement and strong light / backlight.

[0061] The drone is equipped with an RTK-GPS positioning unit (positioning accuracy ±2cm) and a lidar obstacle avoidance unit (minimum obstacle avoidance distance 1.5m). It cruises along a preset spiral route (lateral spacing 0.5-0.8m, height above the blade surface 3-5m), with a maximum cruise speed of 8m / s and a wind resistance level ≥8.

[0062] The event camera module is a 1280×720 resolution dynamic vision sensor (DVS, dynamic range ≥120dB), and the IMU inertial measurement module is a 200Hz sampling frequency with an attitude accuracy of ±0.1°. The synchronization module achieves time synchronization between the event stream (spatial coordinates (x,y), timestamp t, polarity p) and the IMU data through hardware triggering, with a synchronization error ≤10μs.

[0063] The data processing unit performs filtering, motion compensation, and feature construction on the event stream, and outputs a 5-channel event feature map;

[0064] The defect detection unit uses an improved YOLOv11 network to process the 5-channel feature map to achieve defect classification and localization;

[0065] Results output unit: Displays the detection results in real time and generates a visual report containing the three-dimensional coordinates of the defects.

[0066] The supplementary lighting unit dynamically adjusts its power based on event density, using the following formula:

[0067] ,

[0068] Where P is the adjusted output power, and P0 = 10W is the reference power. This represents the current detection time density. The event / ms represents the target density, the power adjustment range is 5-15W, and the response delay is ≤5ms, ensuring the integrity of event characteristics in low-light environments.

[0069] The motion compensation unit of the data processing module removes noise events by using spatiotemporal consistency verification and performs rotation and translation compensation on the event coordinates. The compensation formula is as follows:

[0070] ,

[0071] Where θ is the rotation compensation angle (calculated from the IMU yaw angle), and t x t y The value represents the translation compensation amount; the coordinate offset error after compensation is ≤0.5 pixels; x, y are the event pixel coordinates before compensation. The event pixel coordinates are compensated; the cumulative error is calibrated every 5 seconds using fixed features on the blade (such as bolt holes). After calibration, the angle error is ≤0.1° and the translation error is ≤1 pixel. The calibration formula is:

[0072] ,

[0073] ,

[0074] ,

[0075] Where, θ 特征 t x,特征 t y,特征 These are the true attitude parameters based on blade feature matching;

[0076] The event feature construction unit of the data processing module adopts a dynamic time window, which adaptively adjusts the time window. The calculation formula is:

[0077] ,

[0078] Wherein, the event space density is denoted as Adaptive adjustment of the time window is denoted as Generate a 5-channel feature map: a 1-channel time surface (pixel values ​​are the timestamps of the most recent events), a 2-channel polarity histogram (counts of positive / negative polarity events), and a 2-channel event density map (number of events per unit area).

[0079] The improved YOLOv11 network of the defect detection module includes an event feature attention mechanism, which assigns feature weights to defect regions. Enhancement:

[0080] ,

[0081] in As the benchmark weight, This is the gain coefficient. and These represent the event densities of the defect and background areas, respectively.

[0082] As one embodiment of the present invention, the implementation steps specifically include:

[0083] 1. System initialization: Calibrate the event camera and IMU extrinsic parameters (rotation matrix R, translation vector T), and plan a spiral flight path based on the 3D model of the blade;

[0084] 2. Event Data Acquisition: The UAV cruises along the flight path, and the event camera outputs an event stream (x, y, t, p). The synchronization unit synchronizes the event data with the IMU attitude data (roll angle). Pitch angle Yaw angle Aligned by timestamp;

[0085] 3. Event filtering: Retain events that satisfy the criteria. And the number of valid events in a 3×3 neighborhood is ≥2;

[0086] 4. Motion compensation: Transform the event coordinates and correct accumulated errors using calibration methods;

[0087] 5. Dynamic Feature Construction: Generate a 5-channel event feature map based on a dynamic window strategy;

[0088] 6. Defect detection: Improve the YOLOv11 network to process the feature map and output the defect type (crack / corrosion / gel coat peeling), confidence score (≥0.95 is a valid defect) and two-dimensional coordinates (x1, y1, x2, y2).

[0089] 7. 3D Positioning: The 2D coordinates are converted into 3D coordinates (X, Y, Z) in the blade coordinate system using camera extrinsic parameters. The conversion formula is as follows: Where s is the scale factor (calculated from the camera focal length and flight altitude);

[0090] 8. Output Results: Generates data including defect location, type, size, and depth. ,area ,length A visual report.

[0091] Event data acquisition also includes an active incentive mechanism: when the static defect area (event density) When the duration is ≥1s, a 10-50Hz modulated LED light source is triggered, which excites the brightness fluctuation to generate an event stream, thus increasing the static defect event generation rate.

[0092] The output also includes defect trend prediction, based on the defect size from n consecutive inspections. Predicting size over the next 6 months using an exponential smoothing model : in The smoothing coefficient (value 0.3-0.5) is used to output graded maintenance suggestions.

[0093] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0094] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0095] The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof. The computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0096] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A wind turbine blade defect detection system based on event camera driving, characterized in that, include: The UAV control unit is used to cruise based on a preset spiral route via the UAV positioning module and the lidar obstacle avoidance module. The event sensing unit includes an event camera module, an IMU (Inertial Measurement Unit) module, and a synchronization module. The event camera module acquires visual event streams on the blade surface through a dynamic visual sensor. The IMU module acquires inertial motion data of the UAV. The synchronization module achieves time synchronization between the event stream and the IMU data through hardware triggering. The event sensing unit also includes an adaptive laser illumination module, which dynamically adjusts the power according to the event density to maintain the integrity of event characteristics in low-light environments. When the synchronization module detects that the event density in a static area is continuously below a threshold, it triggers a modulated light source to generate an event stream by exciting brightness fluctuations, thereby enhancing the event response to static defects. The hardware triggering method is as follows: the IMU inertial measurement module outputs a periodic hardware pulse signal to the synchronization module, and the synchronization module adds a unified timestamp to each event collected by the event camera module based on the pulse signal; The data processing unit is used to filter, perform motion compensation, and construct features from the event stream, and output an event feature map. The motion compensation of the data processing unit is based on the attitude data provided by the IMU inertial measurement module to perform rotation and translation transformations on the event coordinates to compensate for image offset caused by UAV movement. The motion compensation of the data processing unit is also used to periodically perform cumulative error calibration based on fixed features on the surface of the wind turbine blades. The feature construction of the data processing unit adopts a dynamic time window strategy to accumulate asynchronous event streams and convert them into multi-channel event feature maps. The event feature maps include time surfaces, polarity histograms, and event density information. The event feature map contains multiple information channels, including at least: a time surface channel with the most recent event timestamp as the pixel value, a polarity histogram channel that records the counts of positive and negative polarity events, and an event density map channel that reflects the number of events per unit area. The defect detection unit classifies and locates defects in the event feature map, processes the event feature map using a neural network model, and obtains the defect type, confidence level, and two-dimensional bounding box. The defect detection unit employs a neural network model that includes an event feature attention mechanism to enhance defect features and suppress background features. The event feature attention mechanism includes: in the feature extraction stage of the network, calculating and inputting the event density map of each candidate region on the feature map in real time, the event density map being obtained based on the event density information channel generated by the data processing unit; for any spatial location on the feature map, its attention weight is dynamically determined by the event density corresponding to the location and the average event density of its neighborhood background; the calculated attention weight is multiplied element-wise with the original feature map in the channel dimension to enhance the features of high-response defect regions and suppress the features of low-response background regions; the parameters of the event feature attention mechanism are trained together with the neural network model, and the optimization objective is to introduce a weight sparsity regularization term into the classification and localization loss function, prompting the network to learn to focus on regions with significantly abnormal event density during training; The result output unit is used to display the detection results in real time and map the two-dimensional location of the defect to the three-dimensional coordinate system of the blade to obtain the three-dimensional coordinates and size information, generating a visualization report containing multi-dimensional information about the defect; the result output unit also includes: predicting the growth trend of the defect size based on historical detection data and outputting graded maintenance suggestions; The visualization report generated by the result output unit also includes a structured defect list generated according to the standard operation and maintenance specifications for wind turbine blades. The list includes at least: blade number, detection time, defect type, three-dimensional location coordinates, equivalent two-dimensional size, estimated depth, confidence level, and recommended maintenance priority.

Images