Fan blade defect detection method and device based on dynamic inspection of unmanned aerial vehicle

Abstract

The invention provides a fan blade defect detection method and device based on dynamic inspection of an unmanned aerial vehicle, and relates to the technical field of fan detection. The method comprises the following steps: based on a blade motion parameter, a flight state parameter of an unmanned aerial vehicle and a shooting time parameter, carrying out deblurring processing and splicing processing on an initial blade image collected in dynamic inspection of the unmanned aerial vehicle to obtain a second image; performing feature extraction on the second image by adopting an SIFT feature matching algorithm to obtain a third image; and performing defect detection on the third image by adopting a dual-channel convolutional neural network. The initial images are subjected to motion compensation and splicing, the blurring effect caused by blade rotation can be eliminated, and the problem of image dislocation in dynamic shooting is solved. According to the method, feature extraction and defect detection are carried out on the basis of the images after motion compensation and splicing are completed, the dynamic inspection efficiency and the defect detection rate are remarkably improved, and efficient and accurate inspection of the fan blade is achieved when the fan is in the running state.

Description

Technical Field

[0001] The present disclosure relates to the field of wind turbine detection technology, and in particular to a method and device for detecting defects in wind turbine blades based on dynamic inspection by unmanned aerial vehicles. Background Art

[0002] During wind turbine operation, blades are subjected to complex loads such as aerodynamic forces, gravity, and centrifugal forces. Long-term operation can lead to blade damage such as cracks and deformation, compromising safe turbine operation and personnel safety. Furthermore, surface damage can disrupt the blade's aerodynamic shape, reducing wind energy capture efficiency and causing significant economic losses. Therefore, accurately and rapidly detecting defects in wind turbine blades and promptly identifying blade damage is a pressing issue in this field. Summary of the Invention

[0003] The present disclosure aims to solve one of the technical problems in the related art at least to a certain extent.

[0004] To this end, the first embodiment of the present disclosure proposes a method for detecting defects in wind turbine blades based on dynamic inspection by drones, comprising:

[0005] Acquire multiple initial images of the wind turbine blade to be inspected, wherein the multiple initial images are images of multiple areas on the wind turbine blade to be inspected captured by the drone during a dynamic inspection process;

[0006] Based on the blade motion parameters of the wind turbine blade to be inspected, the flight state parameters of the drone, and the shooting time parameters, a motion vector of each point on the blade surface of the wind turbine blade to be inspected in the image plane is obtained through a pre-built blade motion prediction model, wherein the motion vector represents the displacement of each point in the image plane during the exposure time of the drone shooting;

[0007] Constructing a blur kernel function according to the motion vector;

[0008] Performing deblurring processing on each of the initial images based on the blur kernel function and Wiener filtering to obtain multiple first images;

[0009] Performing pixel-level alignment on the plurality of first images according to the motion vector to obtain a second image;

[0010] Performing feature extraction on the second image using a SIFT feature matching algorithm to obtain a third image;

[0011] A dual-channel convolutional neural network is used to perform defect detection on the third image.

[0012] 2. The method according to claim 1, characterized in that the blade motion parameters include blade rotation angular velocity, blade length, and installation yaw angle, the flight state parameters include flight speed and heading angle, and the shooting time parameters include exposure time; based on the rotation speed of the wind turbine blade to be inspected and the flight state parameters and shooting time parameters of the UAV, the blade motion prediction model obtains the motion vector of each point on the blade surface of the wind turbine blade to be inspected in the image plane by the following method:

[0013] Calculating the instantaneous velocity of each point according to the blade rotation angular velocity and the blade length;

[0014] Determining the relative motion speed between the UAV and the wind turbine blade to be inspected according to the instantaneous speed, the flight speed, the installation yaw angle, and the heading angle;

[0015] The motion vector is determined according to the relative motion speed and the exposure time.

[0016] 3. The method according to claim 1, wherein the fuzzy kernel function is expressed as follows:

[0017]

[0018] Wherein, K(x,y) is the blur kernel function, and Δx and Δy are the motion vectors.

[0019] 4. The method according to claim 3, wherein the formula for performing deblurring processing on each of the initial images based on the blur kernel function and the Wiener filter is expressed as follows:

[0020]

[0021] Among them, I blur is the initial image, I restore is the first image obtained after deblurring the initial image, K is the blur kernel function, is the Fourier transform, and SNR is the estimated signal-to-noise ratio.

[0022] 5. The method according to claim 1, wherein the dual-channel convolutional neural network includes a texture feature extraction channel and a deformation feature extraction channel, and the performing defect detection on the third image using the dual-channel convolutional neural network comprises:

[0023] Extracting a texture feature vector from the third image through the texture feature extraction channel;

[0024] extracting a deformation feature vector from the third image through the deformation feature extraction channel;

[0025] The texture feature vector and the deformation feature vector are spliced ​​together, and the spliced ​​vector is input into the fully connected layer of the dual-channel convolutional neural network, and the blade defect information in the second image is determined by the Softmax function.

[0026] 6. The method according to any one of claims 1 to 5, further comprising:

[0027] Determining a three-dimensional model of the wind turbine blade to be inspected;

[0028] The blade defect information is mapped to the three-dimensional model, and an inspection report containing the blade defect information is generated.

[0029] The second embodiment of the present disclosure provides a wind turbine blade defect detection device based on a dynamic inspection by a drone, comprising:

[0030] An acquisition module is used to acquire multiple initial images of the wind turbine blade to be inspected, wherein the multiple initial images are images of multiple areas on the wind turbine blade to be inspected collected by the drone during a dynamic inspection process;

[0031] a determination module, configured to obtain, based on the blade motion parameters of the wind turbine blade to be inspected, the flight state parameters of the drone, and the shooting time parameters, a motion vector of each point on the blade surface of the wind turbine blade to be inspected in the image plane using a pre-established blade motion prediction model, wherein the motion vector represents the displacement of each point in the image plane during the exposure time of the drone shooting;

[0032] A construction module, configured to construct a blur kernel function according to the motion vector;

[0033] a motion compensation module, configured to perform deblurring processing on each of the initial images based on the blur kernel function and Wiener filtering to obtain a plurality of first images;

[0034] a stitching module, configured to perform pixel-level alignment on the plurality of first images according to the motion vector to obtain a second image;

[0035] A feature extraction module, configured to extract features from the second image using a SIFT feature matching algorithm to obtain a third image;

[0036] A defect detection module is used to perform defect detection on the third image using a dual-channel convolutional neural network.

[0037] The third embodiment of the present disclosure proposes a wind turbine blade defect detection system, comprising: a drone equipped with an image acquisition device; a processor, and a memory communicatively connected to the processor; wherein the memory stores computer-executable instructions; and the processor executes the computer-executable instructions stored in the memory to implement the method described in the first aspect above.

[0038] A fourth embodiment of the present disclosure provides an electronic device, comprising: a processor, and a memory communicatively connected to the processor;

[0039] The memory stores computer-executable instructions;

[0040] The processor executes the computer-executable instructions stored in the memory to implement the method described in the first aspect.

[0041] The fifth aspect of the present disclosure provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores computer-executable instructions, which are used to implement the method described in the first aspect when executed by a processor.

[0042] The present disclosure provides a method and device for detecting defects in wind turbine blades based on dynamic inspections by drones. By performing motion compensation and splicing on the initial images collected by the drone during the dynamic inspection process, the blurring effect caused by blade rotation is eliminated, and the problem of image misalignment during dynamic shooting is solved. Based on the images after motion compensation and splicing, feature extraction and defect detection are performed, significantly improving the efficiency of dynamic inspections and the defect detection rate. Since the present disclosure can achieve efficient and accurate inspections of wind turbine blades while the wind turbine is in operation, thus avoiding power generation losses due to downtime, the present disclosure also has significant economic and social benefits.

[0043] Additional aspects and advantages of the present disclosure will be given in part in the description below and in part will be obvious from the description below, or will be learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The above and / or additional aspects and advantages of the present disclosure will become apparent and readily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0045] Figure 1 A schematic flow chart of a method for detecting defects in wind turbine blades based on dynamic inspection by drones provided in an embodiment of the present disclosure;

[0046] Figure 2 A schematic diagram of an initial image of multiple areas on a wind turbine blade to be inspected provided by an embodiment of the present disclosure;

[0047] Figure 3 A schematic diagram of a wind turbine blade defect detection device based on drone dynamic inspection provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0048] The following describes in detail embodiments of the present disclosure, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0049] In traditional wind turbine blade inspection methods, blade defects can be detected through manual inspection or automatic inspection. The manual inspection method relies on visual inspection with a telescope or inspection in a hanging basket after shutdown, which is time-consuming, has limited visual coverage, and is inefficient. For the automatic inspection method, since the blades are stationary after the wind turbine is shut down, some dynamic defects (such as cracks and debonding caused by stress during operation) cannot be exposed, and the power generation loss caused by the wind turbine shutdown is large. Although the automatic inspection method, such as blade defect detection based on drone-collected images, solves the problem of detection efficiency, since the blades are in a rotating state when the image is taken, the captured image is easily distorted, which in turn affects the accuracy of the defect detection results.

[0050] To this end, this disclosure proposes a method and device for detecting wind turbine blade defects based on dynamic drone inspections. This method and device can perform motion compensation on the initial images captured during dynamic inspections, eliminating the blurring effect caused by blade rotation and improving defect detection efficiency and accuracy. Specifically, the following describes, with reference to the accompanying drawings, a method and device for detecting wind turbine blade defects based on dynamic drone inspections according to embodiments of the present disclosure.

[0051] Figure 1 The following is a flow chart of a method for detecting defects in wind turbine blades based on dynamic inspection by drones provided in an embodiment of the present disclosure. Figure 1 As shown, the wind turbine blade defect detection method based on UAV dynamic inspection may include the following steps:

[0052] Step 101 : Acquire multiple initial images of a wind turbine blade to be inspected. The multiple initial images are images of multiple areas on the wind turbine blade to be inspected captured by a drone during a dynamic inspection process.

[0053] It should be noted that dynamic inspection is to inspect the wind turbine blades while the wind turbine is in operation, and the wind turbine blades are in motion. Among them, the drone can be equipped with a high-performance camera (such as 120 million pixels, a global shutter speed of 1 / 16000 seconds, and support for high-speed continuous shooting (3 frames per second)) to capture images of the wind turbine blades to be inspected, use AF autofocus to lock the dynamic position of the blades, and correct the focal length through laser ranging. Optionally, when collecting blade images, the drone's shooting angle needs to be perpendicular to the surface of the wind turbine blades (such as the camera is 30-50 meters away from the blades).

[0054] Figure 2 A schematic diagram of initial images of multiple areas on a wind turbine blade to be inspected provided in an embodiment of the present disclosure. Figure 2 It is composed of initial images of multiple areas on the wind turbine blade to be inspected, covering all positions on the same plane of the wind turbine blade to be inspected.

[0055] During wind turbine operation, the nacelle may yaw due to wind direction. To improve the accuracy of blade image acquisition by drones and ensure that no areas are missed, some embodiments of the present disclosure can use the BeiDou system to obtain the tower coordinates, height, blade length, and rotation speed of each wind turbine. Combined with real-time meteorological data (such as wind speed and direction) and the wind turbine's yaw angle, the drone's orbital path can be dynamically planned to adapt to changes in the wind turbine's yaw attitude and ensure that the drone's shooting angle meets preset requirements.

[0056] Step 102: Based on the blade motion parameters of the wind turbine blade to be inspected, the flight state parameters of the drone, and the shooting time parameters, a motion vector of each point on the blade surface of the wind turbine blade to be inspected in the image plane is obtained through a pre-built blade motion prediction model. The motion vector represents the displacement of each point in the image plane during the exposure time of the drone shooting.

[0057] In some embodiments of the present disclosure, blade motion parameters include blade rotational angular velocity, blade length, and installation yaw angle; flight state parameters include flight speed and heading angle; and capture time parameters include exposure time. A pre-built blade motion prediction model can use these parameters to predict the blade's position and attitude at the moment of capture, calculating the motion vector of each point on the blade surface within the image plane.

[0058] In one implementation, the blade motion prediction model can obtain the motion vector of each point on the blade surface of the wind turbine blade to be inspected in the image plane through the following steps S1-S3:

[0059] S1: Calculate the instantaneous velocity v of each point based on the blade rotation angular velocity ω and the blade length L p :

[0060] vp =ω·r p

[0061] Among them, r p The distance from point P on the blade surface to the center of rotation of the fan blade (related to the blade length L)

[0062] S2: According to the instantaneous speed v p 、UAV flight speed v u 、Installation yaw angle θ yaw and heading angle φ u , determine the relative motion speed v between the UAV and the wind turbine blade to be inspected rel :

[0063] v rel =v u ·cos(φ u -θ yaw )-v p

[0064] S3: Determine the motion vector based on the relative motion speed and exposure time Δt:

[0065] Δx=v rel ·Δt·k x , Δy=v rel ·Δt·k y

[0066] Among them, Δx and Δy are the displacements of point P on the blade surface in the image plane, and k x 、k y is the conversion coefficient between pixel size and space ratio (obtained by internal calibration of the image acquisition device used to take pictures on the UAV).

[0067] Step 103: construct a blur kernel function according to the motion vector.

[0068] The blur kernel function is used to characterize the motion path of the leaf within the image plane during the exposure time. The formula for the blur kernel function is as follows:

[0069]

[0070] Among them, K(x,y) is the blur kernel function, Δx and Δy are motion vectors.

[0071] Step 104 : Deblurring is performed on each of the initial images based on a blur kernel function and Wiener filtering to obtain a plurality of first images.

[0072] In one implementation, the formula for deblurring each initial image based on the blur kernel function and the Wiener filter can be expressed as follows:

[0073]

[0074] Among them, I blur is the initial image, I restore is the first image obtained after deblurring the initial image, K is the blur kernel function, is the Fourier transform, and SNR is the estimated signal-to-noise ratio.

[0075] Through steps 102 to 104 , dynamic compensation and deblurring are performed on each initial image to obtain a plurality of first images, thereby removing the blurring effect caused by the blade rotation in the initial image.

[0076] Step 105 , performing pixel-level alignment on the plurality of first images according to the motion vector to obtain a second image.

[0077] I aligned (x,y)=I restore (x+Δx,y+Δy)

[0078] Among them, I aligned (x, y) represents the pixel position of point P on the leaf surface in the second image, where x is the horizontal coordinate of point P in the second image, and y is the vertical coordinate of point P in the second image. Pixel-level alignment improves image stitching efficiency and enables seamless stitching of dynamic images, eliminating the distortion caused by traditional stitching.

[0079] Step 106: Use the SIFT feature matching algorithm to extract features from the second image to obtain a third image.

[0080] The SIFT feature matching algorithm introduces a multi-scale pyramid structure, combined with Difference of Gaussian (DoG) and the Laplace operator, to enhance the ability to extract subtle texture features (such as hidden cracks), improve defect detection rates, and reduce manual review workload. Feature points are screened using adaptive thresholds to reduce noise interference.

[0081] Step 107: Use a dual-channel convolutional neural network to perform defect detection on the third image.

[0082] The dual-channel convolutional neural network (Dual-ChanneICNN) consists of two parallel convolution channels: a texture feature extraction channel and a deformation feature extraction channel. The texture feature extraction channel extracts a texture feature vector from the third image; the deformation feature extraction channel extracts a deformation feature vector from the third image. The texture feature vector and the deformation feature vector are concatenated and fed into the fully connected layer of the dual-channel convolutional neural network. The softmax function is then used to determine blade defect information in the second image.

[0083] The network architecture of the dual-channel convolutional neural network is as follows:

[0084] 1) Texture feature extraction channel:

[0085] Input: third image;

[0086] Structure: 5 layers of convolution (Conv+ReLU+Pooling), convolution kernel size 3x3, stride 1;

[0087] Output: Texture feature vector (such as stains, degumming, etc.).

[0088] 2) Deformation feature extraction channel:

[0089] Input: third image;

[0090] Structure: 4 layers of convolution (Conv+ReLU+Pooling) with kernel size 5x5 and stride 2;

[0091] Output: deformation feature vector (such as cracks, damage, etc.).

[0092] 3) Feature fusion and classification:

[0093] The output feature vectors of the two channels are concatenated and input to a fully connected layer (2 layers, with 256 and 128 neurons, respectively). A softmax function is used to output leaf defect information. This leaf defect information includes the defect category and the probability of that defect category, such as a leaf degumming probability of 90%.

[0094] In order to display the blade defect information more intuitively, in some embodiments of the present disclosure, the three-dimensional model of the wind turbine blade to be inspected is determined; the blade defect information is mapped to the three-dimensional model (BIM / CAD), and an inspection report containing defect coordinates, dimensions, and risk levels is automatically generated, supporting one-click export.

[0095] By implementing the embodiments of the present disclosure, motion compensation and splicing are performed on the initial images collected by the drone during the dynamic inspection process, eliminating the blurring effect caused by blade rotation and solving the problem of image misalignment during dynamic shooting. Based on the motion-compensated and spliced ​​images, feature extraction and defect detection are performed, significantly improving the efficiency of dynamic inspections and the defect detection rate. Because the present disclosure can achieve efficient and accurate inspections of wind turbine blades while the wind turbine is in operation, avoiding power generation losses due to downtime, the present disclosure also has significant economic and social benefits.

[0096] Through experimental comparison, the wind turbine blade defect detection method based on UAV dynamic inspection provided by the embodiment of the present disclosure was measured in a certain wind farm (single unit capacity 3MW). The inspection time of a single wind turbine was less than 10 minutes, an average of 20-30 wind turbines were inspected per day, and the defect identification false alarm rate was less than 3%.

[0097] The comparison results of the wind turbine blade defect detection method based on UAV dynamic inspection disclosed in this disclosure and the traditional shutdown inspection method are shown in the following table:

[0098] Comparison Item Traditional methods This disclosure Inspection efficiency Single unit ≥2 hours Single unit ≤10 minutes Defect detection rate Relying on manual labor, the missed detection rate is >30% AI automatic recognition, missed detection rate <5% Economical Average daily loss from downtime: RMB 50,000 / unit Zero downtime cost Applicable Scenarios Static detection Dynamic / static full coverage

[0099] Figure 3 This is a schematic diagram of a wind turbine blade defect detection device based on a UAV dynamic inspection provided by an embodiment of the present disclosure. Figure 3 As shown, the wind turbine blade defect detection device based on UAV dynamic inspection may include: an acquisition module 301, a determination module 302, a construction module 303, a motion compensation module 304, a splicing module 305, a feature extraction module 306 and a defect detection module 307.

[0100] The acquisition module 301 is used to acquire multiple initial images of the wind turbine blade to be inspected. The multiple initial images are images of multiple areas on the wind turbine blade to be inspected collected by the UAV during the dynamic inspection process.

[0101] Determination module 302 is used to obtain the motion vector of each point on the blade surface of the wind turbine blade to be inspected in the image plane based on the blade motion parameters of the wind turbine blade to be inspected and the flight state parameters and shooting time parameters of the drone through a pre-built blade motion prediction model. The motion vector represents the displacement of each point in the image plane during the exposure time of the drone shooting.

[0102] The construction module 303 is used to construct a blur kernel function according to the motion vector.

[0103] The motion compensation module 304 is configured to perform deblurring processing on each initial image based on a blur kernel function and Wiener filtering to obtain a plurality of first images.

[0104] The stitching module 305 is configured to perform pixel-level alignment on the plurality of first images according to the motion vector to obtain a second image.

[0105] The feature extraction module 306 is configured to extract features from the second image using a SIFT feature matching algorithm to obtain a third image.

[0106] The defect detection module 307 is configured to perform defect detection on the third image using a dual-channel convolutional neural network.

[0107] In some embodiments of the present disclosure, blade motion parameters include blade rotational angular velocity, blade length, and installation yaw angle; flight state parameters include flight speed and heading angle; and shooting time parameters include exposure time. Based on the rotational velocity of the wind turbine blade to be inspected, as well as the flight state parameters and shooting time parameters of the drone, the blade motion prediction model obtains the motion vector of each point on the blade surface of the wind turbine blade to be inspected within the image plane in the following manner: calculating the instantaneous velocity of each point based on the blade rotational angular velocity and blade length; determining the relative motion velocity between the drone and the wind turbine blade to be inspected based on the instantaneous velocity and flight speed; and determining the motion vector based on the relative motion velocity and exposure time.

[0108] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0109] To implement the above embodiment, the present disclosure further provides a wind turbine blade defect detection system, comprising:

[0110] A drone, wherein the drone is equipped with an image acquisition device;

[0111] A processor, and a memory communicatively connected to the processor; wherein,

[0112] The memory stores computer-executable instructions;

[0113] The processor executes the computer-executable instructions stored in the memory to implement the method provided by any of the aforementioned embodiments.

[0114] In order to implement the above embodiments, the present disclosure also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0115] In order to implement the above embodiments, the present disclosure further proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0116] In order to implement the above embodiments, the present disclosure further provides a computer program product, including a computer program, which implements the methods provided in the above embodiments when executed by a processor.

[0117] In the descriptions of the aforementioned embodiments, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are mutually inconsistent.

[0118] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0119] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present disclosure belong.

[0120] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0121] It should be understood that various parts of the present disclosure can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0122] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0123] In addition, the functional units in the various embodiments of the present disclosure may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0124] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present disclosure have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present disclosure. A person of ordinary skill in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present disclosure.

Claims

1. A method for detecting defects in wind turbine blades based on dynamic inspection by drones, characterized in that: The following steps are involved: Acquire multiple initial images of the wind turbine blade to be inspected, wherein the multiple initial images are images of multiple areas on the wind turbine blade to be inspected captured by the drone during a dynamic inspection process; Based on the blade motion parameters of the wind turbine blade to be inspected, the flight state parameters of the drone, and the shooting time parameters, a motion vector of each point on the blade surface of the wind turbine blade to be inspected in the image plane is obtained through a pre-built blade motion prediction model, wherein the motion vector represents the displacement of each point in the image plane during the exposure time of the drone shooting; Constructing a blur kernel function according to the motion vector; Performing deblurring processing on each of the initial images based on the blur kernel function and Wiener filtering to obtain multiple first images; Performing pixel-level alignment on the plurality of first images according to the motion vector to obtain a second image; Performing feature extraction on the second image using a SIFT feature matching algorithm to obtain a third image; A dual-channel convolutional neural network is used to perform defect detection on the third image.

2. The method according to claim 1, characterized in that The blade motion parameters include blade rotation angular velocity, blade length, and installation yaw angle; the flight state parameters include flight speed and heading angle; and the shooting time parameters include exposure time. Based on the rotation speed of the wind turbine blade to be inspected and the flight state parameters and shooting time parameters of the UAV, the blade motion prediction model obtains the motion vector of each point on the blade surface of the wind turbine blade to be inspected in the image plane in the following manner: Calculating the instantaneous velocity of each point according to the blade rotation angular velocity and the blade length; Determining the relative motion speed between the UAV and the wind turbine blade to be inspected according to the instantaneous speed, the flight speed, the installation yaw angle, and the heading angle; The motion vector is determined according to the relative motion speed and the exposure time.

3. The method according to claim 1, characterized in that The formula of the fuzzy kernel function is expressed as follows: Wherein, K(x,y) is the blur kernel function, and Δx and Δy are the motion vectors.

4. The method according to claim 3, characterized in that The formula for performing deblurring processing on each of the initial images based on the blur kernel function and the Wiener filter is expressed as follows: Among them, I blur is the initial image, I restore is the first image obtained after deblurring the initial image, K is the blur kernel function, is the Fourier transform, and SNR is the estimated signal-to-noise ratio.

5. The method according to claim 1, wherein The dual-channel convolutional neural network includes a texture feature extraction channel and a deformation feature extraction channel. The use of the dual-channel convolutional neural network to perform defect detection on the third image includes: Extracting a texture feature vector from the third image through the texture feature extraction channel; extracting a deformation feature vector from the third image through the deformation feature extraction channel; The texture feature vector and the deformation feature vector are spliced ​​together, and the spliced ​​vector is input into the fully connected layer of the dual-channel convolutional neural network, and the blade defect information in the second image is determined by the Softmax function.

6. The method according to any one of claims 1 to 5, characterized in that Also includes: Determining a three-dimensional model of the wind turbine blade to be inspected; The blade defect information is mapped to the three-dimensional model, and an inspection report containing the blade defect information is generated.

7. A wind turbine blade defect detection device based on UAV dynamic inspection, characterized in that: include: An acquisition module is used to acquire multiple initial images of the wind turbine blade to be inspected, wherein the multiple initial images are images of multiple areas on the wind turbine blade to be inspected collected by the drone during a dynamic inspection process; a determination module, configured to obtain, based on the blade motion parameters of the wind turbine blade to be inspected, the flight state parameters of the drone, and the shooting time parameters, a motion vector of each point on the blade surface of the wind turbine blade to be inspected in the image plane using a pre-established blade motion prediction model, wherein the motion vector represents the displacement of each point in the image plane during the exposure time of the drone shooting; A construction module, configured to construct a blur kernel function according to the motion vector; a motion compensation module, configured to perform deblurring processing on each of the initial images based on the blur kernel function and Wiener filtering to obtain a plurality of first images; a stitching module, configured to perform pixel-level alignment on the plurality of first images according to the motion vector to obtain a second image; A feature extraction module, configured to extract features from the second image using a SIFT feature matching algorithm to obtain a third image; A defect detection module is used to perform defect detection on the third image using a dual-channel convolutional neural network.

8. A fan blade defect detection system, characterized in that: include: A drone, wherein the drone is equipped with an image acquisition device; A processor, and a memory communicatively connected to the processor; wherein, The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

9. An electronic device, characterized in that: include: a processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

Images