Multi-viewpoint planning fan blade intelligent inspection method and system

Abstract

The application discloses a multi-viewpoint planning intelligent inspection method and system for a fan blade, which first analyzes the basic element type of defects based on preliminary abnormal images in-depth, and extracts the unique observation requirements; then, the three-dimensional position information of the defects is used to perform multi-dimensional value evaluation on the generated candidate viewpoints, and environmental factors are considered in the evaluation to ensure that each viewpoint can provide high-quality and high-value defect information; on this basis, an advanced combination pruning algorithm is used to intelligently select an optimal viewpoint combination with the strongest information complementarity and no redundancy, and the flight trajectory of the unmanned aerial vehicle is accurately generated according to the optimal viewpoint combination. In this way, the problems of information redundancy and lack of pertinence in traditional inspection are effectively solved, the quality and effectiveness of the defect images are significantly improved, and a more efficient, accurate and reliable intelligent inspection new paradigm is provided for the fan blade, and the safety and economic benefits of fan operation are comprehensively improved.

Description

Technical Field

[0001] This application relates to the field of intelligent inspection, and more specifically, to a method and system for intelligent inspection of wind turbine blades based on multi-viewpoint planning. Background Technology

[0002] Wind power generation, as a crucial component of clean energy, relies heavily on its core equipment, the wind turbine. However, wind turbine blades are constantly exposed to harsh natural environments, making them susceptible to damage from wind and sand erosion, ultraviolet radiation, lightning strikes, and freezing, resulting in defects such as cracks, peeling, and corrosion. These defects not only affect the turbine's operating efficiency and power generation but can also lead to safety hazards or even catastrophic accidents. Traditional blade inspection methods primarily rely on manual climbing or telescope observation, which suffers from high operational risks, low efficiency, and significant weather-related influences, failing to meet the growing demand for refined and routine inspections. Therefore, developing an efficient, safe, and precise intelligent inspection technology is imperative.

[0003] Despite the rapid development of drone inspection technology and image recognition algorithms, making them an effective alternative to traditional methods and achieving significant progress in automated data acquisition, existing technologies still face numerous challenges in the in-depth investigation phase after the initial discovery of suspected defects. Specifically, defect information acquisition often suffers from high redundancy. For example, existing viewpoint planning methods generally lack prior guidance on the morphological characteristics of the defects themselves and fail to intelligently adjust the optimal observation angle according to the specific type of defect, resulting in an inability to fully display key defect information. Simultaneously, environmental factors, especially lighting, have a significant asymmetric impact on the image quality of blade surfaces, and current viewpoint planning strategies often fail to effectively consider these dynamic environmental factors. More critically, finding the optimal combination of viewpoints with the strongest information complementarity and the maximum total information from a massive pool of candidate viewpoints is an exponentially increasing computational complexity problem. It is highly likely to select seemingly good viewpoints that actually overlap in information, severely limiting the efficiency and diagnostic accuracy of intelligent inspection.

[0004] Therefore, an optimized multi-viewpoint planning method for intelligent inspection of wind turbine blades is needed. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a multi-viewpoint planning intelligent inspection method and system for wind turbine blades. First, based on preliminary anomaly images, the method deeply analyzes the primitive types of defects and extracts their unique observation requirements. Then, using the three-dimensional location information of the defects, a multi-dimensional value assessment is performed on the generated candidate viewpoints. This assessment cleverly incorporates environmental factors to ensure that each viewpoint provides high-quality, high-value defect information. Based on this, an advanced combinatorial pruning algorithm intelligently selects the optimal viewpoint combination with the strongest information complementarity and no redundancy, and accurately generates the flight trajectory of a UAV accordingly. This approach effectively solves the problems of information redundancy and lack of specificity in traditional inspections, significantly improving the quality and effectiveness of defect images. Therefore, it provides a more efficient, accurate, and reliable new paradigm for intelligent inspection of wind turbine blades, comprehensively improving the safety and economic benefits of wind turbine operation and maintenance.

[0006] According to one aspect of this application, a method for intelligent inspection of wind turbine blades based on multi-viewpoint planning is provided, comprising:

[0007] Based on the preliminary anomaly images, preliminary analysis and observation aspects of defects are extracted to obtain a list of defect primitive types and required observation aspects;

[0008] Based on the abnormal three-dimensional location, candidate viewpoints are generated and their value is evaluated from the list of required observation aspects to obtain a valuable set of candidate viewpoints.

[0009] The optimal viewpoint combination is obtained by pruning the valuable candidate viewpoint set.

[0010] The flight trajectory is generated based on the drone's status and the optimal viewpoint combination.

[0011] According to another aspect of this application, a multi-viewpoint planning intelligent inspection system for wind turbine blades is provided, comprising:

[0012] The anomaly and defect analysis module is used to perform preliminary defect analysis and observation aspect extraction based on preliminary anomaly images to obtain a list of defect primitive types and required observation aspects.

[0013] The candidate viewpoint value assessment module is used to generate and assess the value of candidate viewpoints based on the list of required observation aspects, based on the abnormal 3D location, to obtain a set of valuable candidate viewpoints.

[0014] The optimal viewpoint pruning module is used to prune a set of valuable candidate viewpoints to obtain the optimal viewpoint combination.

[0015] The flight trajectory generation module is used to generate flight trajectories based on the UAV's status and the optimal viewpoint combination.

[0016] Compared with existing technologies, this application provides a multi-viewpoint planning intelligent inspection method and system for wind turbine blades. First, it analyzes the primitive types of defects based on preliminary anomaly images and extracts their unique observation requirements. Then, using the three-dimensional location information of the defects, it performs a multi-dimensional value assessment of the generated candidate viewpoints. This assessment cleverly incorporates environmental factors to ensure that each viewpoint provides high-quality, high-value defect information. Based on this, an advanced combinatorial pruning algorithm intelligently selects the optimal viewpoint combination with the strongest information complementarity and no redundancy, and accurately generates the flight trajectory of a UAV accordingly. This approach effectively solves the problems of information redundancy and lack of specificity in traditional inspections, significantly improving the quality and effectiveness of defect images. Therefore, it provides a more efficient, accurate, and reliable intelligent inspection paradigm for wind turbine blades, comprehensively improving the safety and economic benefits of wind turbine operation and maintenance. Attached Figure Description

[0017] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 This is a flowchart of a multi-viewpoint planning intelligent inspection method for wind turbine blades according to an embodiment of this application;

[0019] Figure 2 This is a data flow diagram of a multi-viewpoint planning intelligent inspection method for wind turbine blades according to an embodiment of this application;

[0020] Figure 3 This is a block diagram of a wind turbine blade intelligent inspection system based on multi-viewpoint planning according to an embodiment of this application. Detailed Implementation

[0021] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0022] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0023] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0024] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0025] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0026] The technical solution of this application proposes an intelligent inspection method for wind turbine blades based on multi-viewpoint planning. Figure 1 This is a flowchart of a wind turbine blade intelligent inspection method based on multi-viewpoint planning according to an embodiment of this application. Figure 2 This is a data flow diagram illustrating the intelligent inspection method for wind turbine blades based on multi-viewpoint planning according to an embodiment of this application. Figure 1 and Figure 2 As shown, the intelligent inspection method for wind turbine blades based on multi-viewpoint planning according to an embodiment of this application includes the following steps: S1, performing preliminary defect analysis and observation aspect extraction based on preliminary anomaly images to obtain defect primitive types and a list of required observation aspects; S2, generating candidate viewpoints and evaluating their value based on the anomaly's three-dimensional location to obtain a set of valuable candidate viewpoints; S3, pruning the set of valuable candidate viewpoints to obtain an optimal viewpoint combination; S4, generating a flight trajectory based on the UAV's state and the optimal viewpoint combination.

[0027] Specifically, in step S1, preliminary defect analysis and observation aspect extraction are performed based on preliminary anomaly images to obtain a list of defect primitive types and required observation aspects. Wind turbine blades exhibit a wide variety of defects with diverse morphologies; for example, elongated cracks and blocky spalling may require different observation angles or aspects to obtain complete and valuable information. Failure to differentiate defects will lead to information redundancy or missing key information, significantly wasting flight time and battery life, and ultimately affecting the accurate diagnosis of defects. Therefore, in the technical solution of this application, preliminary defect analysis and observation aspect extraction are performed based on preliminary anomaly images to quickly identify and classify initially discovered anomaly areas, and determine the optimal observation aspect based on their inherent characteristics, thereby improving inspection efficiency and data quality.

[0028] Among them, preliminary anomaly images refer to image data containing suspected defects obtained through preliminary inspection (such as wide-area scanning); defect primitive types refer to simplifying complex defects into basic geometric features that can be identified and classified, such as elongated, blocky, etc. These types provide important prior guidance for subsequent viewpoint planning; the list of required observation aspects refers to the set of observation dimensions or angles that must be considered for a specific defect primitive type in order to obtain comprehensive and valuable information.

[0029] In practice, the initial anomaly image is first preprocessed to obtain a binary image. That is, through image preprocessing, noise is eliminated, defect features are enhanced, and a binary image is finally obtained. This processing aims to effectively separate the defect region from the background in the image, laying the foundation for subsequent contour extraction.

[0030] Next, the binary image is input into the contour lookup module to obtain a contour point set. This contour point set represents the boundary of the defect on the two-dimensional image plane and is the basis for quantifying the morphological features of the defect.

[0031] Next, the morphological parameters of the contour point set are calculated, including aspect ratio and compactness. Specifically, the morphological parameters of the contour point set can be calculated through the following steps: First, calculate the minimum bounding rectangle of the contour point set; then, extract the width and height of the minimum bounding rectangle, and calculate the aspect ratio using the following formula:

[0032]

[0033] Where Width is the width, Height is the height, max(·,·) is the maximum value function, and min(·,·) is the minimum value function; this aspect ratio parameter helps distinguish elongated defects (such as cracks) from defects that are closer to square or circular. Then, based on the contour vertices in the contour point set, the contour area and contour perimeter are calculated; where the contour area is the area of ​​the region enclosed by the contour, and the contour perimeter is the length of the contour. Then, based on the contour area and contour perimeter, the compactness is calculated using the following formula:

[0034]

[0035] Among them, Contour area For the contour area, Contour peri The perimeter of the contour is given. Here, the compactness value characterizes how closely the shape of the contour resembles a circle. Obtaining these two morphological parameters provides a quantitative basis for subsequent defect type determination.

[0036] Subsequently, a type determination is performed based on aspect ratio and compactness to obtain the defect primitive type. For example, in practical applications, contours with a large aspect ratio and low compactness can be classified as elongated defects (such as cracks) using a preset threshold or classification algorithm, while contours with an aspect ratio close to 1 and high compactness can be classified as blocky defects (such as peeling or corrosion spots). This determination process can transform abstract image features into specific, operable defect types.

[0037] Finally, the defect primitive types are input into an aspect rule mapping table to obtain the list of required observation aspects. This mapping table predefines the optimal observation strategies for different defect primitive types. For example, for a defect identified as a slender crack, the required observation aspects may include lateral confirmation and depth estimation, requiring the acquisition of images perpendicular to the crack's direction and depth information on both sides of its normal direction. For blocky spalling, contour confirmation and texture details are required, demanding that the viewpoint fully cover its edge contour and capture surface texture. In this way, precise and personalized observation guidance can be provided for subsequent UAV viewpoint planning, ensuring that the most critical and valuable information about the defect is efficiently obtained in each inspection.

[0038] Specifically, S2, based on the abnormal three-dimensional location, generates and evaluates the value of candidate viewpoints from the list of required observation aspects to obtain a valuable set of candidate viewpoints. That is, based on the determined defect type and corresponding observation requirements, as well as the precise location of the defect in three-dimensional space, a series of potential observation viewpoints are automatically generated. These viewpoints are then objectively and multidimensionally evaluated for value. This addresses the problems in traditional inspections, such as poor image quality, information redundancy or omissions, and severe influence from ambient lighting, caused by the arbitrary selection of viewpoints and lack of quantitative standards. This provides optimal observation positions with high-quality information potential for subsequent UAV trajectory planning, ensuring that the most comprehensive and useful defect image data can be obtained at the lowest cost.

[0039] The abnormal 3D location refers to the precise spatial coordinates and local orientation of the defect on the wind turbine blade surface. Specifically, the abnormal 3D location includes the 3D coordinates of the defect center point and the unit normal vector of the defect center point on the blade surface, providing a solid spatial reference for subsequent viewpoint generation. The candidate viewpoint set refers to each potential camera shooting location after initial generation and screening, which will undergo further value assessment. The valuable candidate viewpoint set is the set of quantified scores assigned to each candidate viewpoint after the value assessment, reflecting its observational utility and environmental adaptability.

[0040] In practice, firstly, based on the anomaly's three-dimensional location, a sampling space is defined to obtain a virtual observation hemisphere. This virtual hemisphere is centered on the three-dimensional coordinates of the defect's center point, and its orientation is determined by the unit normal vector of the defect's center point on the blade surface. The normal vector determines the orientation of the defect surface; therefore, the hemisphere extends outward from the defect surface, encompassing all possible free-space viewpoints capable of observing the defect. This process ensures that candidate viewpoints are distributed within the effective observation range of the defect.

[0041] Next, discretization sampling is performed on the surface of the virtual observation hemisphere to obtain a set of candidate viewpoints. That is, on the defined surface of the virtual observation hemisphere, a series of discrete viewpoint positions are generated according to a certain strategy (e.g., uniformly spaced latitude and longitude sampling or density-based non-uniform sampling). Each discrete point is a candidate viewpoint, representing a possible shooting posture and position for a camera mounted on a UAV. These viewpoints form a preliminary set of candidate viewpoints, providing a foundation for subsequent detailed evaluation.

[0042] Furthermore, a multi-dimensional value evaluation is performed on each candidate viewpoint in the candidate viewpoint set to obtain the valuable candidate viewpoint set. This evaluation process is comprehensive and aims to quantify the potential of each candidate viewpoint to acquire high-quality defect information. In this process, firstly, a first candidate viewpoint is extracted from the candidate viewpoint set; then, the viewpoint vector and reflection vector of the first candidate viewpoint are calculated; wherein, the viewpoint vector refers to the direction vector from the first candidate viewpoint to the center point of the defect, representing the imaging direction from the camera to the defect; the reflection vector refers to the direction vector of light reflected from a specific light source (such as the sun) onto the defect surface according to the optical laws. It can be calculated by combining the known or estimated light source position and the normal vector of the defect surface; then, based on the viewpoint vector and reflection vector of the first candidate viewpoint, the glare penalty factor of the first candidate viewpoint is calculated; wherein, glare is a key environmental factor affecting image quality, and the calculation of its penalty factor aims to quantify the degree to which the viewpoint is affected by glare. In the technical solution of this application, the glare penalty factor of the first candidate viewpoint is calculated using the following formula:

[0043]

[0044] Where, δ glare Thershold is the angle between the viewpoint vector and the reflection vector. g-a A preset glare angle threshold is defined, and min(·,·) is the minimum value function. When the angle between the viewpoint vector and the reflection vector is less than or equal to the preset glare angle threshold, the glare penalty factor is usually 0, indicating no significant glare. When the angle between the viewpoint vector and the reflection vector is greater than the preset glare angle threshold, the penalty factor will linearly increase between 0 and 1 based on the difference between the angle between the viewpoint vector and the reflection vector and the preset glare angle threshold, thereby quantifying the penalty for viewpoints severely affected by glare. Then, the original score of the first candidate viewpoint is calculated. Here, the original score represents the degree to which the viewpoint meets the various observation requirements (such as "contour confirmation", "depth estimation", and "texture detail") in the required observation aspect list under ideal lighting conditions. For example, based on the defect type and the required observation aspects, a viewpoint value function can be defined to comprehensively consider the viewpoint's observation angle and distance relative to the defect, as well as whether it can meet the geometric requirements of a specific observation task, to calculate its original score. This score does not consider environmental influences such as glare, but only evaluates its theoretical value as an observation point. Further, the glare penalty factor of the first candidate viewpoint is multiplied by the original score of the first candidate viewpoint to obtain its final value. In the technical solution of this application, the final value of the candidate viewpoint is obtained by multiplying the original score by the glare penalty factor or combining it in other ways, thus providing a quantitative indicator that comprehensively considers observation capability and environmental adaptability. In a specific example, the higher the final value, the more valuable the viewpoint is as a preferred choice.

[0045] Taking the scheme of this application as an example, suppose a suspected crack is found on a wind turbine blade. Its three-dimensional location and orientation have been determined, and the previous step identified the need for two observation aspects: depth estimation and profile confirmation. This step defines a virtual observation hemisphere centered on the crack within its observable area. The system uniformly samples the surface of this hemisphere to generate hundreds of candidate viewpoints, each representing a potential UAV camera position. For one candidate viewpoint, firstly, its viewpoint vector pointing towards the crack is calculated, and then the reflection vector is calculated based on the current solar position and the crack surface normal. If the angle between the viewpoint vector and the reflection vector is close to zero (indicating possible glare), a high glare penalty factor is calculated. Simultaneously, based on the spatial geometric relationship between the viewpoint and the crack, its original scores for depth estimation (usually requiring a grazing angle) and profile confirmation (usually requiring forward observation) are evaluated. Finally, the original scores are multiplied by the glare penalty factor to obtain the final value of the candidate viewpoint. This process is repeated for all candidate viewpoints, forming a set containing each candidate viewpoint and its quantified final value.

[0046] Specifically, S3 involves performing optimal viewpoint combination pruning on the valuable candidate viewpoint set to obtain the optimal viewpoint combination. That is, through optimal viewpoint combination pruning, the smallest, most effective, and most information-complementary viewpoint set is intelligently selected from the valuable candidate viewpoint set generated in the previous step, which may contain a large amount of redundant information. Its core purpose is to solve the complexity problem of combination optimization, avoid selecting viewpoints that seem good but have overlapping information, thereby effectively reducing flight costs, improving inspection efficiency, and ensuring that the final acquired image data is high-quality, non-redundant, and has the highest diagnostic value. Here, information overlap measures the degree of similarity or repetition of defect information that can be obtained between different viewpoints; for example, two viewpoints taken from very similar angles will have a high degree of information overlap.

[0047] The optimal viewpoint combination refers to the final set of viewpoints selected after pruning. This set of viewpoints, while covering all necessary observation aspects, has the fewest number of viewpoints, the most complementary information, and the highest overall value. It represents the best set of shooting positions for subsequent precise and detailed UAV surveys. Here, pruning refers to an optimization strategy that iteratively evaluates and removes redundant or inefficient viewpoints to streamline the set.

[0048] In practice, firstly, for each valuable candidate viewpoint in the valuable candidate viewpoint set, the viewpoint with the highest score is selected from the set to obtain a set of viewpoint pairs. This process can be accomplished by comparing the capabilities of viewpoints in observing different coverage defects, or by calculating the similarity of the image features they can capture. For example, for each candidate viewpoint A, the system will find a viewpoint B in its valuable candidate viewpoint set that is most similar to it in terms of the provided information or has the highest information overlap, and form a viewpoint pair (A, B). The establishment of these viewpoint pairs lays the foundation for the next step of accurately evaluating information overlap and implementing intelligent pruning.

[0049] Subsequently, based on the information overlap of each viewpoint pair in the set, a combination pruning is performed to obtain the optimal viewpoint combination. That is, after obtaining the potential set of viewpoint pairs, the system will perform a detailed analysis of these viewpoint pairs, evaluating the degree of overlap in the information they provide. If the information overlap between two viewpoints in a pair is too high, it means that most of the information provided by one viewpoint has been covered by the other (usually the more valuable one), or its marginal value is diminishing (i.e., the addition of this viewpoint contributes almost zero to the total information). In this case, the system will remove redundant viewpoints with relatively small contributions or low value based on preset rules or optimization algorithms, thereby pruning the set. It is worth noting that this process will continue iteratively until all viewpoints in the set make a significant and non-redundant contribution to the total information, and the number of viewpoints in the set is minimized while satisfying the requirements of all required observation aspect lists. Finally, the optimal viewpoint combination that can obtain the most comprehensive and valuable defect information with the least flight cost is obtained.

[0050] Taking the scheme of this application as an example, suppose that after preliminary processing, a set of valuable candidate viewpoints is obtained, among which two viewpoints (e.g., viewpoint A and viewpoint B) can observe a specific crack in the defect with extremely high clarity (i.e., they both contribute highly to depth estimation or contour confirmation). Calculations show that the information provided by viewpoints A and B regarding the crack overlaps by 90%. Meanwhile, viewpoint A's final value score is slightly higher than viewpoint B's. In the pruning stage, the system will identify this high degree of overlap and may ultimately retain only viewpoint A while discarding viewpoint B, as the addition of viewpoint B does not provide additional valuable information for diagnosis. By iteratively pruning all such viewpoints with high information overlap, a refined and efficient optimal viewpoint combination can be obtained, ensuring that the UAV only needs to fly the most concise path to acquire all the key information required for defect diagnosis.

[0051] Specifically, step S4 generates a flight trajectory based on the drone's state and the optimal viewpoint combination. In the technical solution of this application, generating a flight trajectory ensures that the drone can efficiently, safely, and accurately reach each optimized viewpoint position and complete the image acquisition task with the correct attitude. Effective flight trajectory generation maximizes the drone's endurance, minimizes flight time, avoids obstacles, and guarantees the quality of the captured data, thereby achieving comprehensive intelligent inspection.

[0052] The UAV status refers to the current physical and operational parameters of the UAV, including but not limited to its starting position, real-time attitude (such as attitude, pitch, roll), current and maximum speed, acceleration, rate of climb, rate of descent, endurance (battery charge), sensor payload (such as camera type, lens focal length, gimbal degrees of freedom and range of motion), and environmental adaptability parameters such as obstacle avoidance capability, effective communication range, and wind resistance. The flight trajectory refers to the complete three-dimensional path of the UAV starting from its current position, sequentially visiting all optimal viewpoints, and finally returning or completing its mission. This path consists of a series of discrete waypoints and the action commands that the UAV should execute at each waypoint.

[0053] In practical implementation, firstly, all viewpoints in the optimal viewpoint combination are efficiently sorted. Since a defectos system may need to observe from multiple optimal viewpoints, the system needs to determine the best order for visiting these viewpoints based on the UAV's current state (e.g., starting position) and the spatial relationships between these optimal viewpoints, using path planning algorithms (such as variations of the Traveling Salesman Problem (TSP) or heuristic search algorithms like A*). This order should aim to minimize the total flight distance and energy consumption, thereby maximizing the UAV's effective working time. Next, based on the determined viewpoint access order and the UAV's own kinematic constraints, a specific flight path is generated. This means that between each optimal viewpoint, the system needs to plan smooth and executable path segments, avoiding sharp ascents, descents, or turns, ensuring that the UAV flies within its performance limits such as maximum speed and acceleration. Simultaneously, path planning must consider the UAV's inertia, camera stability, and image acquisition frame rate to ensure high-quality image capture during movement; furthermore, environmental and safety constraints must be incorporated. The generation of the flight trajectory must fully consider the complex three-dimensional geometry of the wind turbine blades themselves, as well as other potential obstacles in the surrounding area, ensuring that the UAV does not collide during the entire flight. This requires leaving sufficient safety distance during trajectory planning. Furthermore, the technical solution in this application also needs to consider the impact of real-time environmental factors such as wind speed and direction on the drone's flight, and reserve sufficient battery capacity to cope with emergencies or return-to-home situations. Finally, the planned 3D path is converted into a sequence of commands that the drone can recognize and execute. This includes decomposing the path into a series of precise waypoints (including latitude, longitude, and altitude), and setting accompanying action commands for each waypoint, such as triggering the camera shutter at a specific waypoint, adjusting the gimbal's pitch and yaw angles to align with defects, and pausing at specific points for longer shooting or data transmission. These commands will be directly uploaded to the drone's flight control system to guide its automated flight.

[0054] Taking the scheme of this application as an example, assume that the system has identified a crack on a wind turbine blade and calculated two optimal viewpoints: one is a horizontal observation point V1 for acquiring the overall morphology of the crack, and the other is a vertical top-down viewpoint V2 for accurately measuring the crack depth. The drone is currently located at the bottom of the wind turbine tower. The flight trajectory generation process first determines the path sequence from the drone's current position, visiting V1 and V2 in sequence, and finally returning (e.g., HOME->V1->V2->HOME). Then, the system calculates the specific waypoint sequence from HOME to V1, from V1 to V2, and then from V2 back to HOME based on the drone's speed and battery capacity. Upon reaching V1, the flight path command includes adjusting the drone's gimbal to precisely align the camera with the crack and triggering an image capture; similarly, at point V2, the command will again adjust the gimbal attitude to capture crack depth information from a vertical perspective and trigger another image capture. The entire trajectory ensures that the drone maintains a safe distance from the blade throughout the entire flight and avoids any potential collision risks.

[0055] In summary, the intelligent inspection method for wind turbine blades based on multi-viewpoint planning according to the embodiments of this application is explained. First, it analyzes the primitive types of defects based on preliminary anomaly images and extracts their unique observation requirements. Then, using the three-dimensional location information of the defects, it performs a multi-dimensional value assessment of the generated candidate viewpoints. This assessment cleverly incorporates environmental factors to ensure that each viewpoint provides high-quality, high-value defect information. Based on this, an advanced combinatorial pruning algorithm intelligently selects the optimal viewpoint combination with the strongest information complementarity and no redundancy, and accurately generates the flight trajectory of the UAV accordingly. This approach effectively solves the problems of information redundancy and lack of specificity in traditional inspections, significantly improving the quality and effectiveness of defect images. Therefore, it provides a more efficient, accurate, and reliable new paradigm for intelligent inspection of wind turbine blades, comprehensively improving the safety and economic benefits of wind turbine operation and maintenance.

[0056] Furthermore, a multi-viewpoint planning intelligent inspection system for wind turbine blades is also provided.

[0057] Figure 3 This is a block diagram of a multi-viewpoint planning intelligent inspection system for wind turbine blades according to an embodiment of this application. Figure 3As shown, the wind turbine blade intelligent inspection system 300 with multi-viewpoint planning according to an embodiment of this application includes: an anomaly defect analysis module 310, used to perform preliminary defect analysis and observation aspect extraction based on preliminary anomaly images to obtain defect primitive types and a list of required observation aspects; a candidate viewpoint value evaluation module 320, used to generate and evaluate candidate viewpoints based on the anomaly three-dimensional position and the list of required observation aspects to obtain a set of candidate viewpoints with value; an optimal viewpoint pruning module 330, used to perform optimal viewpoint combination pruning on the set of candidate viewpoints with value to obtain an optimal viewpoint combination; and a flight trajectory generation module 340, used to generate a flight trajectory based on the UAV status and the optimal viewpoint combination.

[0058] As described above, the multi-viewpoint planning intelligent wind turbine blade inspection system 300 according to the embodiments of this application can be implemented in various wireless terminals, such as servers with multi-viewpoint planning intelligent wind turbine blade inspection algorithms. In one possible implementation, the multi-viewpoint planning intelligent wind turbine blade inspection system 300 according to the embodiments of this application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the multi-viewpoint planning intelligent wind turbine blade inspection system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the multi-viewpoint planning intelligent wind turbine blade inspection system 300 can also be one of many hardware modules of the wireless terminal.

[0059] Alternatively, in another example, the multi-viewpoint planning wind turbine blade intelligent inspection system 300 and the wireless terminal can also be separate devices, and the multi-viewpoint planning wind turbine blade intelligent inspection system 300 can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with the agreed data format.

[0060] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for intelligent inspection of wind turbine blades using multi-viewpoint planning, characterized in that, include: Based on the preliminary anomaly images, preliminary analysis and observation aspects of defects are extracted to obtain a list of defect primitive types and required observation aspects; Based on the abnormal three-dimensional location, candidate viewpoints are generated and their value is evaluated from the list of required observation aspects to obtain a valuable set of candidate viewpoints. The optimal viewpoint combination is obtained by pruning the valuable candidate viewpoint set. Generate flight trajectory based on UAV status and optimal viewpoint combination; Furthermore, a multi-dimensional value evaluation is performed on each candidate viewpoint in the candidate viewpoint set to obtain the valued candidate viewpoint set, which includes: Extract a first candidate viewpoint from the candidate viewpoint set; Calculate the viewpoint vector and reflection vector of the first candidate viewpoint; Calculate the glare penalty factor for the first candidate viewpoint based on the viewpoint vector and reflection vector of the first candidate viewpoint. Calculate the original score of the first candidate viewpoint; The glare penalty factor of the first candidate viewpoint is multiplied by the original score of the first candidate viewpoint to obtain the final value of the first candidate viewpoint; where... The calculation of the glare penalty factor for the first candidate viewpoint based on its viewpoint vector and reflection vector includes: calculating the glare penalty factor for the first candidate viewpoint using the following formula: ； in, The angle between the viewpoint vector and the reflection vector. To preset the glare angle threshold, This is a function that takes the minimum value.

2. The intelligent inspection method for wind turbine blades based on multi-viewpoint planning according to claim 1, characterized in that, Based on the preliminary anomaly images, preliminary defect analysis and observation aspect extraction are performed to obtain a list of defect primitive types and required observation aspects, including: Preprocess the initial anomaly image to obtain a binary image; The binary image is input into the contour lookup module to obtain a set of contour points; Calculate the morphological parameters of the contour point set, including aspect ratio and compactness; The type of defect primitive is determined based on aspect ratio and compactness.

3. The intelligent inspection method for wind turbine blades based on multi-viewpoint planning according to claim 2, characterized in that, Calculate the morphological parameters of the contour point set, including aspect ratio and compactness, including: Calculate the minimum bounding rectangle of the set of contour points; Extract the width and height of the minimum bounding rectangle, and calculate the aspect ratio using the following formula: ； in, For width, For height, To find the maximum value function, The function is for finding the minimum value; Based on the contour vertices in the contour point set, calculate the contour area and contour perimeter. The compactness is calculated based on the contour area and contour perimeter using the following formula: ； in, For the outline area, This is the perimeter of the outline.

4. The intelligent inspection method for wind turbine blades based on multi-viewpoint planning according to claim 2, characterized in that, The preliminary analysis and observation aspect extraction based on the preliminary anomaly image to obtain the defect primitive type and the list of required observation aspects also includes: inputting the defect primitive type into an aspect rule mapping table to obtain the list of required observation aspects.

5. The intelligent inspection method for wind turbine blades based on multi-viewpoint planning according to claim 1, characterized in that, The abnormal three-dimensional location includes the three-dimensional coordinates of the defect center point and the unit normal vector of the defect center point on the blade surface; Specifically, based on the anomaly's 3D location, candidate viewpoints are generated and their value assessed from the list of required observation aspects to obtain a valuable set of candidate viewpoints, including: Based on the abnormal three-dimensional location, a sampling space is defined to obtain a virtual observation hemisphere; Discretization sampling is performed on the surface of the virtual observation hemisphere to obtain the candidate viewpoint set.

6. The intelligent inspection method for wind turbine blades based on multi-viewpoint planning according to claim 1, characterized in that, To obtain the optimal viewpoint combination, the valuable candidate viewpoint set is pruned using the following methods: For each valuable candidate viewpoint in the set of valuable candidate viewpoints, select the viewpoint with the highest score from the set of valuable candidate viewpoints to obtain a set of viewpoint pairs. Based on the information overlap of each viewpoint pair in the set of viewpoint pairs, they are combined and pruned to obtain the optimal viewpoint combination.

7. A multi-viewpoint planning intelligent inspection system for wind turbine blades, characterized in that, include: The anomaly and defect analysis module is used to perform preliminary defect analysis and observation aspect extraction based on preliminary anomaly images to obtain a list of defect primitive types and required observation aspects. The candidate viewpoint value assessment module is used to generate and assess the value of candidate viewpoints based on the list of required observation aspects, based on the abnormal 3D location, to obtain a set of valuable candidate viewpoints. The optimal viewpoint pruning module is used to prune a set of valuable candidate viewpoints to obtain the optimal viewpoint combination. The flight trajectory generation module is used to generate flight trajectories based on the UAV's status and the optimal viewpoint combination; Furthermore, the candidate viewpoint value assessment module is also used to perform multi-dimensional value assessments on each candidate viewpoint in the candidate viewpoint set to obtain the valuable candidate viewpoint set, which includes: Extract a first candidate viewpoint from the candidate viewpoint set; Calculate the viewpoint vector and reflection vector of the first candidate viewpoint; Calculate the glare penalty factor for the first candidate viewpoint based on the viewpoint vector and reflection vector of the first candidate viewpoint. Calculate the original score of the first candidate viewpoint; The glare penalty factor of the first candidate viewpoint is multiplied by the original score of the first candidate viewpoint to obtain the final value of the first candidate viewpoint; where... The calculation of the glare penalty factor for the first candidate viewpoint based on its viewpoint vector and reflection vector includes: calculating the glare penalty factor for the first candidate viewpoint using the following formula: ； in, The angle between the viewpoint vector and the reflection vector. To preset the glare angle threshold, This is a function that takes the minimum value.

Images