Image-based leaf inspection method and system

Abstract

The invention discloses an image-based blade inspection method and system, and relates to the technical field of wind power equipment operation and maintenance, and the method comprises the steps: employing an unmanned plane to inspect and collect blade image data; performing denoising, enhancement and feature extraction on the acquired leaf image; the blade damage development trend is analyzed and predicted through machine learning, and early warning information is provided; and operation and maintenance suggestions are given in combination with the blade operation environment. According to the method, the SLAM algorithm is combined with an unmanned aerial vehicle system of a GPS and an IMU sensor, so that precise positioning and comprehensive coverage shooting of the blade surface are realized; the image quality is enhanced through technologies such as Gaussian filtering and HDR imaging, and damage types such as cracks and corrosion are intelligently classified by using a deep learning model; the development trend of blade damage is predicted by combining an ARIMA model with finite element analysis, and early warning information and operation suggestions are provided based on a multi-modal time sequence diagnosis model. According to the method, the inspection efficiency, accuracy and intelligence can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind power equipment operation and maintenance, and in particular to an image-based blade inspection method and system. Background Art

[0002] With the transformation of the global energy structure and the improvement of the utilization rate of renewable energy, wind power generation has gradually become an important part of the clean energy field. As the core component of energy conversion, the health status of the blades in wind power equipment directly affects the power generation efficiency and equipment life. However, due to the long-term exposure of the blades to complex and changeable natural environments (such as strong winds, high humidity, dust, ice and snow, etc.), cracks, corrosion, ice accumulation and other damages are prone to appear on their surfaces, which not only increases the operation and maintenance costs, but also poses certain safety hazards. Traditional blade inspection methods mainly rely on manual climbing inspections and long-distance observations on the ground. This method is difficult to operate and has a high risk of personnel safety. In addition, due to the low inspection efficiency and insufficient detection accuracy, it is often difficult to detect minor damage in time, resulting in potential failures not being effectively prevented. In recent years, with the rapid development of drone technology, image processing technology and machine learning, combining these technologies for blade inspection has become a hot topic in research and application. The existing drone inspection methods have solved the limitations of manual inspections to a certain extent, and can achieve preliminary inspections by collecting blade surface images from high altitudes. However, these methods still have problems such as poor stability, insufficient environmental adaptability, and limited data processing capabilities.

[0003] The main problems of the existing technology are concentrated in the following aspects: First, the comprehensiveness and accuracy of the inspection are insufficient. Due to the complexity of the blade surface and irregular distribution of positions, the existing drone inspection path planning algorithm often cannot cover the entire blade surface. The existence of missed inspection areas may cause key problems to not be discovered in time; second, the intelligence level of the detection results is not high. The existing technology is mainly based on traditional image analysis methods, and the classification and identification capabilities of damage such as cracks and corrosion are limited. Especially when the lighting conditions are complex or the blade materials are diverse, the detection accuracy is further reduced; third, there is a lack of damage trend prediction capabilities. Existing technologies are difficult to accurately predict the expansion trend of damage based on blade operation data and historical inspection records, and thus cannot effectively guide operation and maintenance strategies. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the prior art has weak operation and maintenance decision-making optimization capability in a multi-wind turbine scenario, resulting in unreasonable allocation of maintenance resources and low efficiency.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a blade inspection method based on an image, comprising:

[0007] Use drone inspection to collect blade image data;

[0008] De-noising, enhancing and feature extracting the collected leaf images;

[0009] Predict the development trend of blade damage through machine learning analysis and provide early warning information;

[0010] Operation and maintenance recommendations are given based on the blade operating environment.

[0011] As a preferred solution of the image-based blade inspection method described in the present invention, the inspection using a drone includes: using a SLAM algorithm in combination with GPS and IMU sensors to achieve stable flight and precise positioning of the drone; a laser radar provides depth information to detect the distance between the blade surface and obstacles; automatically planning a path to achieve full coverage shooting of the blade through an optimal coverage algorithm;

[0012] Multi-angle shooting, set the shooting angle θ and distance d to meet the blade width coverage formula:

[0013]

[0014] Where W is the blade width and θ is the camera field of view.

[0015] As a preferred solution of the image-based blade inspection method described in the present invention, the collected blade image data includes environmental adaptability, using HDR imaging technology, multiple images I1, I2, ..., I n According to the exposure weight w i Synthesize a high dynamic range image:

[0016]

[0017] Data transmission and storage module, data transmission, uses the ultra-low latency characteristics of the 5G network to perform real-time transmission of video streams through the UDP protocol;

[0018] Satellite communication is an alternative solution for remote areas. Data volume control formula:

[0019] R=W·H·F·C

[0020] Among them, R is the data transmission rate, W, H are the image resolution, F is the frame rate, and C is the number of color channels.

[0021] As a preferred solution of the image-based blade inspection method described in the present invention, the feature extraction includes distributed storage, using the Hadoop distributed file system, and the storage structure is designed as metadata storage, blade inspection task identification, timestamp; image data storage, original image, processed image and analysis results;

[0022] Perform image preprocessing and Gaussian filtering to denoise:

[0023]

[0024] Where σ is the standard deviation of the filter window. Image enhancement: contrast-limited adaptive histogram equalization;

[0025] Eigenvector construction formula:

[0026] f=[f1,f2,…,f n ]

[0027] Among them, f i Including crack length, width, direction and other parameters; trend prediction, based on the ARIMA model to predict the crack extension trend:

[0028] y t =α+β1y t-1 +β2y t-2 +…+∈ t

[0029] Among them, y t is the crack extension at time t.

[0030] As a preferred solution of the image-based blade inspection method described in the present invention, wherein: the prediction of blade damage development trend through machine learning analysis includes fault detection, convolutional neural network architecture: using ResNet, YOLO or UNet to detect cracks and corrosion;

[0031] Classification loss function:

[0032]

[0033] 3D modeling, using the SfM algorithm, constructs blade point cloud data through multi-view images; point cloud matching formula:

[0034]

[0035] Among them, d ij Represents the Euclidean distance between two points; performs intelligent classification and uses support vector machine (SVM) or random forest algorithm to classify faults.

[0036] As a preferred solution of the image-based blade inspection method described in the present invention, the providing of early warning information includes constructing a fault prediction model based on a physical field, combining finite element analysis, and simulating the stress distribution and crack propagation path of the blade in the operating state to improve the physical accuracy of the fault trend prediction;

[0037] Multimodal time series diagnostic model, crack propagation formula:

[0038] Δa=C(ΔK) m

[0039] Among them, Δa is the crack extension length, ΔK is the stress intensity factor amplitude, and C,m is the material parameter. The multimodal sensor data, including images, thermal data, and material acoustic emission signals, are combined with the time series model to perform fault time series analysis. The digital twin of the blade is created to assist in diagnosis and prediction through real-time simulation comparison of inspection data and design standards. The digital twin model formula is expressed as:

[0040] D(t)=f(I sensor (t),P simulation (t))

[0041] Where D(t) is the diagnosis result, I sensor (t) is the sensor input data, P simulation (T) is the predicted value of the simulation model.

[0042] As a preferred solution of the image-based blade inspection method of the present invention, the operation and maintenance suggestions given in combination with the blade operation environment include an augmented reality interactive system, which uses AR technology to project the inspection results onto the blade digital model in real time, so that on-site personnel can view the fault location and suggested repair solutions in real time through smart glasses or tablets;

[0043] Based on the decision optimization of game theory, in a multi-wind turbine system, according to the cost function of operation and maintenance, the game theory optimization strategy is used to determine the best maintenance time:

[0044]

[0045] Among them, R i (t) is the wind turbine operating income, C i (t) is the maintenance cost; the UAV self-repair module integrates a local repair device on the UAV to achieve simple automatic repair of the blade surface, and at the same time issues maintenance requirements and blade equipment warnings.

[0046] As a preferred solution of the image-based blade inspection system described in the present invention, it includes:

[0047] The image acquisition module uses a multi-rotor drone equipped with a high-definition camera and a thermal imager to inspect the blade surface. It is also equipped with a laser radar for precise positioning and obstacle avoidance.

[0048] Data transmission and storage module, which transmits inspection data to the cloud in real time via 5G network or satellite communication;

[0049] The data processing and analysis module performs denoising, enhancement and feature extraction on the collected blade images; uses a deep learning model to identify and classify fault types such as cracks, corrosion, and ice accumulation on the blade surface;

[0050] The fault diagnosis module diagnoses the degree of blade damage and possible causes based on the historical fault sample library; predicts the development trend of blade damage through time series analysis and machine learning, and provides early warning information; automatically generates inspection reports including blade status, fault type, specific location and recommended maintenance plans

[0051] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a step of an image-based blade inspection method.

[0052] A computer-readable storage medium stores a computer program, which implements the steps of an image-based blade inspection method when executed by a processor.

[0053] Beneficial effects of the present invention: The image-based blade inspection method provided by the present invention realizes accurate positioning and comprehensive coverage shooting of the blade surface by using SLAM algorithm combined with GPS and IMU sensor drone system; enhances image quality by using Gaussian filtering and HDR imaging and other technologies, and uses deep learning models (such as ResNet, YOLO or UNet) to intelligently classify damage types such as cracks and corrosion; predicts the development trend of blade damage by combining ARIMA model with finite element analysis, and provides early warning information and operation suggestions based on multimodal time series diagnosis model. This method can significantly improve the efficiency, accuracy and intelligence level of inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0055] Figure 1 An overall flow chart of an image-based blade inspection method provided for the first embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0057] Example 1, reference Figure 1 , as an embodiment of the present invention, provides an image-based blade inspection method, comprising:

[0058] S1: Use drone inspection to collect blade image data.

[0059] Furthermore, the drone can cruise automatically, using PID controller or SLAM algorithm combined with GPS and IMU sensors to achieve stable flight and precise positioning of the drone. LiDAR provides depth information to detect the distance between the blade surface and obstacles. The path is automatically planned, and the optimal coverage algorithm (such as SpanningTree Coverage Algorithm) is used to achieve full coverage shooting of the blade.

[0060] Multi-angle shooting: Set the shooting angle θ and distance d to satisfy the blade width coverage formula:

[0061]

[0062] Where W is the blade width and θ is the camera field of view.

[0063] Environmental adaptability, using HDR imaging technology, multiple images I1, I2, …, I taken under different lighting conditions n According to the exposure weight w i Synthesize a high dynamic range image:

[0064]

[0065] Data transmission and storage module,data transmission, uses the ultra-low latency characteristics of the 5G network to perform real-time transmission of video streams through the UDP protocol.

[0066] Satellite communication is an alternative solution for remote areas. Data volume control formula:

[0067] R=W·H·F·C

[0068] Among them, R is the data transmission rate, W, H are image resolutions, F is the frame rate, and C is the number of color channels.

[0069] Distributed storage,adopts Hadoop Distributed File System (HDFS), and the storage structure is designed as,metadata storage: blade inspection task identification,timestamp. Image data storage: original image, processed image and analysis results.

[0070] It should be noted that the optional technology is the 5G communication protocol (eMBB enhanced mobile broadband). Distributed file systems: Hadoop, Ceph. The SLAM algorithm is combined with GPS and IMU sensors to achieve precise positioning, ensuring the stable flight of the drone in complex environments. The depth information provided by the lidar significantly improves the obstacle avoidance capability, which is especially important in the multi-obstacle area of ​​the wind farm, reducing the risk of accidental collision.

[0071] S2: De-noise, enhance and extract features from the collected leaf images.

[0072] Further, image preprocessing and Gaussian filtering for denoising:

[0073]

[0074] Where σ is the standard deviation of the filter window. Image enhancement: Contrast Limited Adaptive Histogram Equalization (CLAHE). Fault detection, Convolutional Neural Network (CNN) architecture: Detect cracks and corrosion using ResNet, YOLO, or UNet.

[0075] Classification loss function:

[0076]

[0077] 3D modeling uses the Structure-from-Motion (SfM) algorithm to construct blade point cloud data from multi-view images. Point cloud matching formula:

[0078]

[0079] Among them, d ij Represents the Euclidean distance between two points.

[0080] Intelligent classification, using support vector machine (SVM) or random forest algorithm to classify faults. Feature vector construction formula:

[0081] f=[f1,f2,…,f n ]

[0082] Among them, f i Including parameters such as crack length, width, and direction.

[0083] Trend prediction, based on the ARIMA model to predict the crack growth trend:

[0084] yt =α+β1y t-1 +β2y t-2 +…+∈ t

[0085] Among them, y t is the crack extension at time t.

[0086] Real-time monitoring + dynamic adjustment, using GANs to generate simulated fault images, enhance the data set, and improve the deep learning model's ability to identify small sample faults; GAN loss function:

[0087]

[0088] It should be noted that the adaptive learning algorithm is introduced to dynamically adjust the feature extraction parameters under different blade materials and environmental conditions to improve the accuracy of the analysis. Image processing and point cloud analysis are integrated, combined with deep learning and geometric algorithms to achieve automatic reconstruction of point clouds based on image features and improve the efficiency of 3D modeling. The improved ICP algorithm (lterative Closest Point) is introduced to improve the accuracy of point cloud registration.

[0089] S3: Predict the development trend of blade damage through machine learning analysis and provide early warning information.

[0090] Furthermore, the fault prediction model based on physical fields, combined with finite element analysis (FEA), improves the physical accuracy of fault trend prediction by simulating the stress distribution and crack propagation path of the blade under operating conditions.

[0091] Multimodal time series diagnostic model, crack propagation formula:

[0092] Δa=C(ΔK) m

[0093] Among them, Δa is the crack extension length, ΔK is the stress intensity factor amplitude, and C,m is the material parameter. Combine multimodal sensor data (image + thermal data + material acoustic emission signal) with a time series model (such as LSTM or Transformer) to perform fault time series analysis. 3. Digital twin system support: Create a digital twin of the blade, compare the inspection data with the design standard through real-time simulation, and assist in diagnosis and prediction. Digital twin model formula:

[0094] D(t)=f(I sensor (t),P simulation (t))

[0095] Where D(t) is the diagnosis result, I sensor (t) is the sensor input data, P simulation (t) is the predicted value of the simulation model.

[0096] Distributed storage and real-time data transmission: Hadoop distributed file system optimizes the organization of inspection data by storing task identification, timestamp and other information through metadata, facilitating the rapid retrieval of historical data. Combined with the ultra-low latency characteristics of 5G communication, real-time transmission of video streams is achieved through the UDP protocol, providing technical support for remote diagnosis and emergency processing, especially when switching to satellite communication in remote areas, data compression and quantization strategies ensure the stability of transmission.

[0097] S4: Provide operation and maintenance recommendations based on the blade operating environment.

[0098] Furthermore, the augmented reality (AR) interactive system uses AR technology to project the inspection results onto the digital model of the blade in real time, allowing on-site personnel to view the fault location and suggested repair plans in real time through smart glasses or tablets.

[0099] Based on the decision optimization of game theory, in a multi-wind turbine system, according to the cost function of operation and maintenance, the game theory optimization strategy is used to determine the best maintenance time:

[0100]

[0101] Among them, R i (t) is the wind turbine operating income, C i (t) is the maintenance cost.

[0102] The UAV self-repair module integrates local repair devices on the UAV, such as a spraying device (to repair the blade surface coating) or a laser heating device (to eliminate minor cracks), to achieve simple automatic repair of the blade surface.

[0103] Cloud-edge collaborative storage, combined with quantum key distribution (QKD) technology, ensures the security of inspection data during transmission. Build an "edge + cloud" collaborative architecture, determine the data storage location through a hierarchical algorithm, and store key data (such as suspected fault areas) on drone edge devices. Historical and redundant data are stored in the cloud. Optimize transmission bandwidth formula:

[0104]

[0105] Among them, R eff is the actual transmission rate of the cloud, R edge (i) is the amount of data processed by the edge device.

[0106] Blockchain data storage uses blockchain technology to decentrally store inspection data, record historical data of each inspection, and ensure that the data cannot be tampered with.

[0107] Furthermore, convolutional neural networks (such as ResNet and YOLO) can accurately identify various fault types such as cracks and corrosion through classification loss function optimization, and the detection accuracy is improved by more than 20% compared with traditional methods. Trend prediction based on the ARIMA model combined with finite element analysis (FEA) not only performs statistical modeling on historical data, but also simulates the stress distribution of the blade under operation, significantly improving the physical accuracy of crack propagation prediction.

[0108] It should be noted that AR technology projects the inspection results onto the digital model of the blade in real time, providing visual repair suggestions for on-site personnel and reducing the difficulty of understanding traditional text reports. Game theory optimization strategy dynamically adjusts decisions based on wind turbine operating benefits and maintenance costs, especially in multi-wind turbine systems, and can be used through the formula max U ∑[R i -C i ]Achieve optimal allocation of resources and significantly reduce maintenance costs.

[0109] Example 2: The following is an embodiment of the present invention, which provides an image-based blade inspection method. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0110] In order to verify the innovation and beneficial effects of the image-based blade inspection method, a set of comparative experiments was designed. The experimental objects were six wind turbine blades (numbered T1 to T6) running in a wind farm, and the traditional manual inspection method and the drone image inspection method described in the invention were applied respectively. The drone inspection system configuration includes: a multi-rotor drone equipped with a high-definition camera, a lidar, and an IMU sensor; flight control using the SLAM algorithm combined with GPS; data transmission using a 5G network; crack detection using the ResNet model; and data storage using the Hadoop distributed file system.

[0111] The experiment is carried out in the following steps: UAV inspection image acquisition: Through the SLAM algorithm combined with LiDAR, the UAV can achieve precise flight and automatically plan the shooting path to ensure coverage of the blade surface.

[0112] The collected images use HDR technology to synthesize high dynamic range images under different lighting conditions. The data is transmitted to the processing center in real time via the 5G network. If the network is unstable, it switches to satellite communication. Gaussian filtering is used to denoise the image, and then the image is enhanced by contrast limited adaptive histogram equalization (CLAHE). The crack features (length, width, direction, etc.) of the blade are extracted and the feature vector is constructed.

[0113] Machine learning analysis, crack detection uses the ResNet deep learning model, and the detection effect is optimized through the classification loss function. The crack extension trend analysis uses the ARIMA model, combined with finite element analysis to calculate the stress distribution.

[0114] Result output and optimization decision: The inspection results are projected in real time in the blade digital twin system to display the crack location and expansion risk. Combined with the operating environment and game theory strategy, operation and maintenance recommendations are generated.

[0115] Table 1 Tabular data records

[0116]

[0117]

[0118] Through the analysis of experimental data, it can be clearly seen that the image-based blade inspection method is significantly better than the traditional manual inspection method in many key performance indicators. The image acquisition success rate of drone inspection image acquisition success rate is 96% on average, while manual inspection is only 76.5%. This is due to the SLAM algorithm combined with LiDAR technology, which ensures flight stability and comprehensive coverage.

[0119] The system achieved an average crack detection accuracy of 92% through a deep learning model (ResNet), which is much higher than the 61.5% of manual inspection. This shows that automated image processing technology has significantly improved fault detection capabilities. The crack propagation prediction error of drone inspections is controlled within 5%, while the error of manual inspections is 14.5%. This is because the ARIMA model combined with finite element analysis achieves more accurate trend prediction.

[0120] Embodiment 3, the following is an embodiment of the present invention, which provides an image-based blade inspection system, including:

[0121] Image acquisition module, drone inspection: a multi-rotor drone equipped with a high-definition camera and thermal imager is used to inspect the blade surface; it is also equipped with a laser radar for precise positioning and obstacle avoidance.

[0122] Real-time monitoring: Combined with GPS and IMU, automatic cruise shooting of blades and multi-angle scanning of key areas can be achieved.

[0123] Environmental adaptability: The module supports image acquisition under different lighting conditions (such as night and haze) and improves image quality through HDR camera.

[0124] The data transmission and storage module transmits the inspection data to the cloud in real time via 5G network or satellite communication. Data storage is based on a distributed storage architecture, supporting the rapid storage and retrieval of large-scale image and video data.

[0125] Data processing and analysis module, pre-processing: denoising, enhancing and feature extraction of collected blade images. Intelligent identification: using deep learning models such as convolutional neural networks (CNN) to identify and classify fault types such as cracks, corrosion, ice accumulation, etc. on the blade surface. 3D modeling: combining multi-angle images to generate a 3D model of the blade, and accurately measuring blade deformation or damage through point cloud analysis.

[0126] Fault diagnosis module, intelligent classification: Based on the historical fault sample library, diagnose the degree of blade damage (such as minor, severe) and possible causes. Trend prediction: Predict the development trend of blade damage through time series analysis and machine learning, and provide early warning information. Automatically generate inspection reports, including the status of the blade, fault type, specific location and recommended maintenance plan. Combined with the blade operating environment (wind speed, humidity, etc.), it is recommended to shut down for maintenance or continue operation. Support remote control of maintenance robots or drones to perform simple surface repairs (such as clearing ice accumulation).

[0127] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0128] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0129] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a 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, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0130] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0131] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A blade inspection method based on an image, characterized in that: include: Use drone inspection to collect blade image data; De-noising, enhancing and feature extracting the collected leaf images; Predict the development trend of blade damage through machine learning analysis and provide early warning information; Operation and maintenance recommendations are given based on the blade operating environment.

2. The image-based blade inspection method according to claim 1, characterized in that: The drone inspection includes using SLAM algorithm combined with GPS and IMU sensors to achieve stable flight and precise positioning of the drone; laser radar provides depth information to detect the distance between the blade surface and obstacles; automatically plans the path and achieves full coverage shooting of the blade through the optimal coverage algorithm; Multi-angle shooting, set the shooting angle θ and distance d to meet the blade width coverage formula: Where W is the blade width and θ is the camera field of view.

3. The image-based blade inspection method according to claim 2, characterized in that: The leaf image data collected includes environmental adaptability, using HDR imaging technology, multiple images I1, I2, ..., I1 taken under different lighting conditions n According to the exposure weight w i Synthesize a high dynamic range image: Data transmission and storage module, data transmission, uses the ultra-low latency characteristics of the 5G network to perform real-time transmission of video streams through the UDP protocol; Satellite communication is an alternative solution and is suitable for remote areas; data volume control formula: R=W·H·F·C Among them, R is the data transmission rate, W, H are the image resolution, F is the frame rate, and C is the number of color channels.

4. The image-based blade inspection method according to claim 3, characterized in that: The feature extraction includes: Distributed storage, using Hadoop distributed file system, the storage structure is designed as follows: metadata storage, blade inspection task identification, timestamp; image data storage, original image, processed image and analysis results; Perform image preprocessing and Gaussian filtering to denoise: Where σ is the standard deviation of the filter window; image enhancement and contrast-limited adaptive histogram equalization are performed; Eigenvector construction formula: f=[f1,f2,...,f n ] Among them, f i Including crack length, width, direction and other parameters; trend prediction, based on the ARIMA model to predict the crack extension trend: y t =α+β1y t-1 +β2y t-2 +…+∈ t Among them, y t is the crack extension at time t.

5. The image-based blade inspection method according to claim 4, characterized in that: The prediction of blade damage development trends through machine learning analysis includes fault detection, convolutional neural network architecture: using ResNet, YOLO or UNet to detect cracks and corrosion; Classification loss function: 3D modeling, using the SfM algorithm, constructs blade point cloud data through multi-view images; point cloud matching formula: Among them, d ij Represents the Euclidean distance between two points; performs intelligent classification and uses support vector machine (SVM) or random forest algorithm to classify faults.

6. The image-based blade inspection method according to claim 5, characterized in that: Providing early warning information includes building a fault prediction model based on physical fields, combining finite element analysis, and simulating the stress distribution and crack propagation path of the blade under operation to improve the physical accuracy of fault trend prediction; Multimodal time series diagnostic model, crack propagation formula: Δa=C(ΔK) m Among them, Δa is the crack extension length, ΔK is the stress intensity factor amplitude, and C, m are material parameters. The multimodal sensor data, including images, thermal data, and material acoustic emission signals, are combined with the time series model to perform fault time series analysis. The digital twin of the blade is created to assist in diagnosis and prediction through real-time simulation comparison of inspection data and design standards. The digital twin model formula is expressed as: D(t)=f(I sensor (t),P simulation (t)) Where D(t) is the diagnosis result, I sensor (t) is the sensor input data, P simulation (T) is the predicted value of the simulation model.

7. The image-based blade inspection method according to claim 6, characterized in that: The operation and maintenance suggestions given in combination with the blade operation environment include an augmented reality interactive system, which uses AR technology to project the inspection results onto the blade digital model in real time, so that on-site personnel can view the fault location and recommended repair solutions in real time through smart glasses or tablets; Based on the decision optimization of game theory, in a multi-wind turbine system, according to the cost function of operation and maintenance, the game theory optimization strategy is used to determine the best maintenance time: Among them, R i (t) is the wind turbine operating income, C i (t) is the maintenance cost; the UAV self-repair module integrates a local repair device on the UAV to achieve simple automatic repair of the blade surface, and at the same time issues maintenance requirements and blade equipment warnings.

8. A system using the image-based blade inspection method according to any one of claims 1 to 7, characterized in that: The image acquisition module uses a multi-rotor drone equipped with a high-definition camera and a thermal imager to inspect the blade surface. It is also equipped with a laser radar for precise positioning and obstacle avoidance. Data transmission and storage module, which transmits inspection data to the cloud in real time via 5G network or satellite communication; The data processing and analysis module performs denoising, enhancement and feature extraction on the collected blade images; uses a deep learning model to identify and classify fault types such as cracks, corrosion, and ice accumulation on the blade surface; The fault diagnosis module diagnoses the degree of blade damage and possible causes based on the historical fault sample library; predicts the development trend of blade damage through time series analysis and machine learning, and provides early warning information; and automatically generates inspection reports including the blade status, fault type, specific location and recommended maintenance plan.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the image-based blade inspection method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the image-based blade inspection method according to any one of claims 1 to 7 are implemented.

Images