Wind power plant fan blade detection method and system based on three-dimensional point cloud

Abstract

The invention relates to the technical field of fan blade detection, and provides a wind power plant fan blade detection method and system based on three-dimensional point clouds, and the method comprises the following steps: carrying out the multi-angle scanning of a fan blade through a laser scanner, and obtaining the three-dimensional point cloud data of the fan blade; performing registration processing on the three-dimensional point cloud data, and unifying the three-dimensional point cloud data into the same coordinate system to obtain registered point cloud data; performing block acceleration processing on the registered point cloud data by adopting an octree partitioning algorithm to obtain block acceleration point cloud data; performing feature extraction and feature segmentation on the block acceleration point cloud data to obtain a key feature region of the fan blade; performing defect identification analysis on the key feature region based on a point cloud network enhancement model to obtain a defect detection result, the defect detection result comprising a defect position and a defect type; and generating a detection report containing the defect position and the defect type, and visually displaying the detection report. According to the invention, the integrity of data acquisition and the precision of defect identification are improved, and the efficiency of fan blade detection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wind turbine blade detection, and in particular to a wind farm wind turbine blade detection method and system based on three-dimensional point cloud. Background Art

[0002] As an important clean energy source, the safe and stable operation of wind power generation is directly related to power generation efficiency and economic benefits. As the core component of wind turbine generator sets, wind turbine blades are prone to various defects such as cracks, corrosion, delamination, etc. during long-term operation. If these defects are not discovered and handled in time, they may lead to serious safety accidents.

[0003] The existing technology generally uses single-angle laser scanning or simple rigid registration algorithm to detect wind turbine blades. Single-angle scanning is to place the laser scanner at a fixed position and perform a single scan and acquisition of the blade surface; the simple rigid registration algorithm only considers the spatial position relationship of the point cloud data and realizes point cloud alignment by minimizing the Euclidean distance between corresponding points. There are problems of incomplete data collection and low defect recognition accuracy. Summary of the invention

[0004] In view of this, the present invention proposes a wind farm wind turbine blade detection method and system based on three-dimensional point cloud, which solves the problems of incomplete data collection and low defect recognition accuracy in the prior art.

[0005] The technical solution of the present invention is implemented as follows: In a first aspect, the present invention provides a method for detecting wind turbine blades in a wind farm based on a three-dimensional point cloud, comprising the following steps:

[0006] S1, using a laser scanner to scan the fan blades at multiple angles to obtain three-dimensional point cloud data of the fan blades;

[0007] S2, performing registration processing on the three-dimensional point cloud data, unifying the three-dimensional point cloud data into the same coordinate system, and obtaining registered point cloud data;

[0008] S3, using the octree partitioning algorithm to perform block-by-block acceleration processing on the registered point cloud data to obtain block-by-block accelerated point cloud data;

[0009] S4, performing feature extraction and feature segmentation on the block accelerated point cloud data to obtain key feature areas of the wind turbine blades;

[0010] S5, performing defect recognition analysis on the key feature area based on the point cloud network enhancement model to obtain a defect detection result, wherein the defect detection result includes a defect location and a defect type;

[0011] S6, generating a test report including the defect location and the defect type, and visually displaying the test report.

[0012] On the basis of the above technical solution, preferably, step S1 includes:

[0013] S11, determining a plurality of scanning position points according to the size and shape characteristics of the wind turbine blade, wherein the scanning position points include a blade front scanning position point and a blade back scanning position point, wherein the blade front scanning position points are evenly distributed along the blade span direction, and the blade back scanning position point and the blade front scanning position point are arranged opposite to each other in the transverse direction of the blade;

[0014] S12, placing a laser scanner at each scanning position point, adjusting the scanning parameters of the laser scanner, and performing all-round scanning and acquisition of the wind turbine blades, wherein the scanning parameters include scanning resolution, scanning angle, and scanning distance.

[0015] Based on the above technical solution, preferably, step S2 includes:

[0016] S21, using an iterative closest point algorithm to perform coarse registration to align the three-dimensional point cloud data at multiple angles;

[0017] S22, performing precise registration by weighted averaging method to improve the registration accuracy, obtaining unified registration point cloud data, and adjusting the dynamic weight in the weighted averaging method according to the acquisition angle and acquisition distance of each set of registration point cloud data.

[0018] On the basis of the above technical solution, preferably, step S3 includes:

[0019] S31, constructing an octree spatial partitioning structure, including: analyzing the spatial density distribution of the registered point cloud data, determining the spatial range of the initial root node of the octree, and recursively segmenting the child nodes according to a preset spatial resolution standard until a dynamically adjusted depth limit is reached or the number of point clouds contained in each child node is lower than a threshold; dynamically adjusting the depth and number of branches of the octree according to the spatial distribution characteristics of the point cloud data;

[0020] S32, mapping the constructed octree partition structure to multiple processing units, including: dynamically distributing the load to different processing units according to the spatial position and number of point clouds of each octree leaf node, and using a load balancing algorithm to optimize the distribution of computing tasks of each processing unit; and realizing parallel block processing of registration point cloud data.

[0021] Based on the above technical solution, preferably, step S4 includes:

[0022] S41, extracting key features from the unified registration point cloud data, the key features including geometric features and spectral features, including: calculating the main direction and curvature of each registration point based on the principal component analysis method to obtain geometric features, the geometric features are used to characterize the geometric shape of the point, and combining the multispectral imaging data to extract the spectral reflectance characteristics of each point as the spectral features;

[0023] S42, detecting and classifying the key feature areas of the wind turbine blades based on the key features, including: using a support vector machine algorithm to classify points based on the geometric features and the spectral features, identifying potential abnormal feature areas, obtaining a preliminary classification result, and filtering and optimizing the preliminary classification result through a post-processing filtering optimization algorithm to obtain a key feature area;

[0024] The calculation formula for the fan blade classification score is:

[0025]

[0026] Among them, Score q is the classification score of the fan at the qth point, f g (q) is the geometric eigenvalue of the qth point, f s (q) is the spectral characteristic value of the qth point, w g and w s are the weight coefficients of geometric features and spectral features, w g and w s The sum of is 1, γ is the post-processing adjustment coefficient, m is the parameter of the control function slope, and θ0 is the classification probability threshold;

[0027] The calculation formula of the post-processing filtering optimization algorithm is:

[0028]

[0029] Among them, P final (q) is the key feature region division result of the qth point, 1 represents the corresponding point is in the key feature region, 0 represents the corresponding point is not in the key feature region, θ' is the post-processing classification threshold, and δ is the neighbor point number threshold.

[0030] Based on the above technical solution, preferably, step S5 includes:

[0031] S51, based on the PointNet++ model, integrates the attention mechanism, adds a multi-scale feature extraction module and a residual connection structure to build a point cloud network enhancement model;

[0032] S52, outputting a feature vector of the key feature area based on the point cloud network enhancement model, classifying the feature vector into defect types using a defect type classification algorithm to obtain defect types, locating the spatial position of each defect in the key feature area according to the three-dimensional spatial coordinate information to obtain the defect position, and screening and verifying the defect type and the defect position based on the confidence score;

[0033] The calculation formula of the defect type classification algorithm is:

[0034]

[0035] Among them, P type (c|n) is the probability that the nth point belongs to the cth type of defect, z c (n) is the score of the nth point on the cth defect, z c' (n) is the score of the nth point on the c'th defect type, and C is the number of defect types;

[0036] The confidence score is calculated as:

[0037]

[0038] Among them, Confidence t is the confidence score of the t-th defect location, is the spatial distance of the t-th defect point, Score t is the classification score of the wind turbine at the t-th point, σ(·) is the sigmoid activation function, and β1 and β2 are the weight coefficients of the spatial distance and the classification score of the wind turbine, respectively.

[0039] Based on the above technical solution, preferably, step S6 includes:

[0040] S61, collating and summarizing the coordinate data of the detected defect positions, associating each defect position with the corresponding point defect type information, integrating the position information and the type information to obtain a test report in a standard format, wherein the test report includes the defect position information and the defect type data;

[0041] S62, visualizing the inspection report to generate a defect distribution map of the three-dimensional model of the wind turbine blade, displaying the defect locations, using different colors or different icons to identify different types of defects, and using an interactive interface to support viewing and screening of defect information.

[0042] In a second aspect, the present invention further provides a wind farm wind turbine blade detection system based on three-dimensional point cloud, the system comprising:

[0043] Point cloud data module, used to scan the fan blades at multiple angles using a laser scanner to obtain three-dimensional point cloud data of the fan blades;

[0044] A data registration module is used to perform registration processing on the three-dimensional point cloud data, unify the three-dimensional point cloud data into the same coordinate system, and obtain registered point cloud data;

[0045] A block acceleration module is used to perform block acceleration processing on the registration point cloud data using an octree partitioning algorithm to obtain block accelerated point cloud data;

[0046] A feature extraction module, used to extract and segment the block acceleration point cloud data to obtain key feature areas of the wind turbine blades;

[0047] A defect detection module is used to perform defect recognition analysis on the key feature area based on the point cloud network enhancement model to obtain a defect detection result, wherein the defect detection result includes a defect location and a defect type;

[0048] The defect display module is used to generate a detection report including the defect location and the defect type, and to visually display the detection report.

[0049] In a third aspect, the present invention further provides an electronic device, comprising: at least one processor, at least one memory, a communication interface and a bus;

[0050] Among them, the processor, memory, and communication interface communicate with each other through the bus, the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement steps such as a wind farm wind turbine blade detection method based on three-dimensional point cloud.

[0051] In a fourth aspect, the present invention further provides a computer-readable storage medium storing computer instructions, wherein the computer instructions enable a computer to implement steps of a method for detecting wind turbine blades in a wind farm based on three-dimensional point clouds.

[0052] The wind farm wind turbine blade detection method and system based on three-dimensional point cloud of the present invention has the following beneficial effects compared with the prior art:

[0053] (1) Complete three-dimensional point cloud data of wind turbine blades is acquired through multi-angle laser scanning, and accurate registration is performed using the iterative closest point algorithm and weighted average method. The octree partitioning algorithm is used to accelerate data processing, and defect identification is performed by combining geometric features and spectral feature extraction and point cloud network enhancement model. Finally, a visual inspection report is generated, which improves the integrity of data acquisition and the accuracy of defect identification, and improves the efficiency of wind turbine blade inspection.

[0054] (2) By extracting geometric features and spectral features from the unified registered point cloud data and combining principal component analysis with multispectral imaging data, the key feature areas of wind turbine blades are accurately identified and classified. The support vector machine algorithm is used for classification, and the preliminary classification results are optimized through the post-processing filter optimization algorithm, ensuring the accuracy of feature extraction and classification, and improving the ability to identify potential abnormal feature areas of wind turbine blades;

[0055] (3) By building a point cloud network enhancement model based on the PointNet++ model, integrating the attention mechanism, adding a multi-scale feature extraction module and a residual connection structure, the key feature areas of the wind turbine blades can be more accurately extracted and analyzed. Combined with the defect type classification algorithm and the confidence scoring mechanism, the defect type and its spatial position can be accurately classified and located, reducing the possibility of false detection and missed detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 A flow chart of a method for detecting wind turbine blades in a wind farm based on three-dimensional point cloud according to the present invention;

[0058] Figure 2 This is a structural diagram of a wind farm wind turbine blade detection system based on three-dimensional point cloud according to the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] See also Figure 1 The present invention provides a method for detecting wind turbine blades in a wind farm based on three-dimensional point clouds, comprising the following steps:

[0061] S1, using a laser scanner to scan the fan blades at multiple angles to obtain three-dimensional point cloud data of the fan blades;

[0062] S2, performing registration processing on the three-dimensional point cloud data, unifying the three-dimensional point cloud data into the same coordinate system, and obtaining registered point cloud data;

[0063] S3, using the octree partitioning algorithm to perform block-by-block acceleration processing on the registered point cloud data to obtain block-by-block accelerated point cloud data;

[0064] S4, performing feature extraction and feature segmentation on the block accelerated point cloud data to obtain key feature areas of the wind turbine blades;

[0065] S5, performing defect recognition analysis on the key feature area based on the point cloud network enhancement model to obtain a defect detection result, wherein the defect detection result includes a defect location and a defect type;

[0066] S6, generating a test report including the defect location and the defect type, and visually displaying the test report.

[0067] Specifically, this embodiment acquires complete three-dimensional point cloud data of wind blades through multi-angle laser scanning, adopts iterative nearest point algorithm and weighted average method for precise alignment, uses octree partitioning algorithm to accelerate data processing, and combines geometric features and spectral feature extraction and point cloud network enhancement model for defect identification, and finally generates a visual inspection report, which improves the integrity of data collection and the accuracy of defect identification, and improves the efficiency of wind blade inspection.

[0068] Step S1 includes:

[0069] S11, determining a plurality of scanning position points according to the size and shape characteristics of the wind turbine blade, wherein the scanning position points include a blade front scanning position point and a blade back scanning position point, wherein the blade front scanning position points are evenly distributed along the blade span direction, and the blade back scanning position point and the blade front scanning position point are arranged opposite to each other in the transverse direction of the blade;

[0070] S12, placing a laser scanner at each scanning position point, adjusting the scanning parameters of the laser scanner, and performing all-round scanning and acquisition of the wind turbine blades, wherein the scanning parameters include scanning resolution, scanning angle, and scanning distance.

[0071] In a specific embodiment, the number of scanning position points is not less than 4, the scanning resolution is set to 0.5-2 mm, the scanning angle is 120°-180°, and the scanning distance is 3-8 meters.

[0072] Specifically, this embodiment sets multiple scanning position points on the front and back of the wind blade and performs all-round scanning and collection, thereby ensuring the comprehensiveness and integrity of the blade surface data, reducing the blind spots and risks of missed detection in data collection, and using a high-resolution (0.5-2mm) and large scanning angle (120°-180°) laser scanner to capture subtle features of the blade surface. According to the size and shape characteristics of the wind blade, the scanning parameters (such as a scanning distance of 3-8 meters) are flexibly adjusted to adapt to wind blades of different types and sizes, thereby improving the flexibility of detection.

[0073] Step S2 includes:

[0074] S21, using an iterative closest point algorithm to perform coarse registration to align the three-dimensional point cloud data at multiple angles;

[0075] S22, performing precise registration by weighted average method to improve registration accuracy, obtaining unified registration point cloud data, and adjusting the dynamic weight in the weighted average method according to the acquisition angle and acquisition distance of each set of registration point cloud data;

[0076] The calculation formula of the dynamic weight is:

[0077]

[0078] Among them, w i is the dynamic weight of the weighted average method, θ i is the acquisition angle of the i-th group of point clouds relative to the reference surface, d i is the collection distance;

[0079] The calculation formula of the weighted average method is:

[0080]

[0081] Among them, P j is the three-dimensional coordinate of the jth registration point, P ji is the 3D coordinate of the jth point in the i-th point cloud group, and N is the number of point cloud groups involved in the registration.

[0082] In a specific embodiment, the error threshold of the coarse registration is set to 0.01 meters, and the dynamic weight range is between 0.1 and 1.

[0083] Specifically, this embodiment uses an iterative nearest point algorithm for coarse registration and sets the error threshold to 0.01 meters to ensure the preliminary alignment of multi-angle three-dimensional point cloud data. It uses a weighted average method for fine registration and dynamically adjusts the weight (ranging from 0.1 to 1) according to the acquisition angle and distance of each set of point cloud data to improve the accuracy of registration. It takes into account the impact of different scanning conditions on data quality and optimizes the uniformity of point cloud data. Through precise registration processing, the multi-angle scanning data is unified into the same coordinate system to obtain high-precision registered point cloud data.

[0084] Step S3 includes:

[0085] S31, constructing an octree spatial partitioning structure, including: analyzing the spatial density distribution of the registered point cloud data, determining the spatial range of the initial root node of the octree, and recursively segmenting the child nodes according to a preset spatial resolution standard until a dynamically adjusted depth limit is reached or the number of point clouds contained in each child node is lower than a threshold; dynamically adjusting the depth and number of branches of the octree according to the spatial distribution characteristics of the point cloud data;

[0086] S32, mapping the constructed octree partition structure to multiple processing units, including: dynamically allocating the load to different processing units according to the spatial position and point cloud quantity of each octree leaf node, optimizing the computing task allocation of each processing unit by using a load balancing algorithm; and realizing parallel block processing of registration point cloud data;

[0087] The calculation formula for dynamic load distribution is:

[0088]

[0089] Among them, LU k is the load assigned to the kth processing unit, N k is the number of point clouds in the kth octet leaf node, D k is the spatial density of the kth octet leaf node, C k is the main frequency of the kth processing unit, α and β are the point cloud quantity adjustment coefficient and spatial density adjustment coefficient respectively;

[0090] The calculation formula of the load balancing algorithm is:

[0091]

[0092] Among them, LB k is the load balancing ratio of the kth processing unit, and M is the number of processing units.

[0093] Specifically, this embodiment constructs an octree spatial partition structure to accelerate the processing of the registration point cloud data in blocks. The octree algorithm recursively divides the child nodes according to the spatial density distribution of the point cloud data until the dynamically adjusted depth limit is reached or the number of point clouds contained in each child node is lower than the threshold, reducing the complexity of data processing.

[0094] The constructed octree partition structure is mapped to multiple processing units, and the load balancing algorithm is used to optimize the distribution of computing tasks of each processing unit. According to the spatial position and number of point clouds of each octree leaf node, the load is dynamically distributed to different processing units to ensure that the computing tasks of each processing unit are evenly distributed, which improves the efficiency of parallel processing. By dynamically adjusting the depth and number of branches of the octree, it can adapt to point cloud data sets of different sizes and complexities.

[0095] Step S4 includes:

[0096] S41, extracting key features from the unified registration point cloud data, the key features including geometric features and spectral features, including: calculating the main direction and curvature of each registration point based on the principal component analysis method to obtain geometric features, the geometric features are used to characterize the geometric shape of the point, and combining the multispectral imaging data to extract the spectral reflectance characteristics of each point as the spectral features;

[0097] S42, detecting and classifying the key feature areas of the wind turbine blades based on the key features, including: using a support vector machine algorithm to classify points based on the geometric features and the spectral features, identifying potential abnormal feature areas, obtaining a preliminary classification result, and filtering and optimizing the preliminary classification result through a post-processing filtering optimization algorithm to obtain a key feature area;

[0098] The calculation formula for the fan blade classification score is:

[0099]

[0100] Among them, Score q is the classification score of the fan at the qth point, f g (q) is the geometric eigenvalue of the qth point, f s (q) is the spectral characteristic value of the qth point, w g and w s are the weight coefficients of geometric features and spectral features, w g and w s The sum of is 1, γ is the post-processing adjustment coefficient, m is the parameter of the control function slope, and θ0 is the classification probability threshold;

[0101] The calculation formula of the post-processing filtering optimization algorithm is:

[0102]

[0103] Among them, P final (q) is the key feature region division result of the qth point, 1 represents the corresponding point is in the key feature region, 0 represents the corresponding point is not in the key feature region, θ' is the post-processing classification threshold, and δ is the neighbor point number threshold.

[0104] Specifically, this embodiment extracts geometric features and spectral features from the unified registration point cloud data, calculates the main direction and curvature of each registration point using principal component analysis, and extracts spectral reflectance characteristics using multispectral imaging data, thereby ensuring the accuracy of key features of wind turbine blades.

[0105] The support vector machine algorithm is used to classify the extracted geometric features and spectral features, identify potential abnormal feature areas, and optimize the preliminary classification results through a post-processing filtering optimization algorithm to obtain the key feature areas. The key feature areas of the wind turbine blades are segmented, thereby improving the efficiency of feature segmentation.

[0106] Step S5 includes:

[0107] S51, based on the PointNet++ model, integrates the attention mechanism, adds a multi-scale feature extraction module and a residual connection structure to build a point cloud network enhancement model;

[0108] S52, outputting a feature vector of the key feature area based on the point cloud network enhancement model, classifying the feature vector into defect types using a defect type classification algorithm to obtain defect types, locating the spatial position of each defect in the key feature area according to the three-dimensional spatial coordinate information to obtain the defect position, and screening and verifying the defect type and the defect position based on the confidence score;

[0109] The calculation formula of the defect type classification algorithm is:

[0110]

[0111] Among them, P type (c|n) is the probability that the nth point belongs to the cth type of defect, z c (n) is the score of the nth point on the cth defect, z c' (n) is the score of the nth point on the c'th defect type, and C is the number of defect types;

[0112] The confidence score is calculated as:

[0113]

[0114] Among them, Confidence t is the confidence score of the t-th defect location, is the spatial distance of the t-th defect point, Score t is the classification score of the wind turbine at the t-th point, σ(·) is the sigmoid activation function, and β1 and β2 are the weight coefficients of the spatial distance and the classification score of the wind turbine, respectively.

[0115] Specifically, this embodiment uses a point cloud network enhancement model based on the PointNet++ model, integrates an attention mechanism, adds a multi-scale feature extraction module and a residual connection structure, thereby improving the point cloud network enhancement model's ability to extract feature vectors of key feature areas of wind blades, and can better capture subtle features in point cloud data.

[0116] The extracted feature vectors are classified using the defect type classification algorithm, and the spatial position of each defect is located by combining the three-dimensional spatial coordinate information. The defect type and position are screened and verified through the confidence scoring mechanism, thereby improving the accuracy of wind turbine blade defect detection.

[0117] Step S6 includes:

[0118] S61, collating and summarizing the coordinate data of the detected defect positions, associating each defect position with the corresponding point defect type information, integrating the position information and the type information to obtain a test report in a standard format, wherein the test report includes the defect position information and the defect type data;

[0119] S62, visualizing the inspection report to generate a defect distribution map of the three-dimensional model of the wind turbine blade, displaying the defect locations, using different colors or different icons to identify different types of defects, and using an interactive interface to support viewing and screening of defect information.

[0120] Specifically, this embodiment organizes and summarizes the coordinate data of the detected defect positions, associates each defect position with the corresponding defect type information, and generates a test report in a standard format. By visualizing the test report, a defect distribution map of the three-dimensional model of the wind turbine blade is generated, and different colors or icons are used to identify different types of defects, so that users can quickly identify and locate defects, improving the readability of defect detection results.

[0121] The interactive interface supports users to view and filter defect information, enhancing their ability to understand and analyze test results.

[0122] See also Figure 2 The present invention also provides a wind farm wind turbine blade detection system based on three-dimensional point cloud, the system comprising:

[0123] Point cloud data module, used to scan the fan blades at multiple angles using a laser scanner to obtain three-dimensional point cloud data of the fan blades;

[0124] A data registration module is used to perform registration processing on the three-dimensional point cloud data, unify the three-dimensional point cloud data into the same coordinate system, and obtain registered point cloud data;

[0125] A block acceleration module is used to perform block acceleration processing on the registration point cloud data using an octree partitioning algorithm to obtain block accelerated point cloud data;

[0126] A feature extraction module, used to extract and segment the block acceleration point cloud data to obtain key feature areas of the wind turbine blades;

[0127] A defect detection module is used to perform defect recognition analysis on the key feature area based on the point cloud network enhancement model to obtain a defect detection result, wherein the defect detection result includes a defect location and a defect type;

[0128] The defect display module is used to generate a detection report including the defect location and the defect type, and to visually display the detection report.

[0129] Specifically, a wind farm wind turbine blade detection system based on three-dimensional point cloud in this embodiment uses a laser scanner to perform multi-angle scanning through a point cloud data module to obtain three-dimensional point cloud data of the wind turbine blades, and unifies the multi-angle point cloud data into the same coordinate system based on the data alignment module to ensure data consistency.

[0130] The block acceleration module uses the octree partitioning algorithm to process the registered point cloud data in blocks, which improves the speed of data processing. By using parallel processing technology, the system can quickly process large-scale point cloud data and adapt to the inspection needs of wind turbine blades of different sizes.

[0131] The feature extraction module extracts and segments the block acceleration point cloud data to obtain key feature areas. The defect detection module performs defect recognition and analysis based on the point cloud network enhancement model to provide high-precision defect detection results, including defect location and type.

[0132] The defect display module generates a test report containing the defect location and type, and displays it visually. Through the interactive interface, users can easily view and filter defect information, which improves the readability of the test results and user experience.

[0133] The present invention also discloses an electronic device, comprising: at least one processor, at least one memory communication interface and a bus: wherein the processor, memory and communication interface communicate with each other through the bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to implement a wind farm wind turbine blade detection method based on three-dimensional point cloud.

[0134] The present invention also discloses a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to implement all or part of the steps of a wind turbine blade detection method based on a three-dimensional point cloud in an embodiment of the present invention. The storage medium includes: a U disk, a mobile hard disk, a read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk, and other media that can store program codes.

[0135] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A wind farm wind turbine blade detection method based on three-dimensional point cloud, characterized in that: The following steps are involved: S1, using a laser scanner to scan the fan blades at multiple angles to obtain three-dimensional point cloud data of the fan blades; S2, performing registration processing on the three-dimensional point cloud data, unifying the three-dimensional point cloud data into the same coordinate system, and obtaining registered point cloud data; S3, using the octree partitioning algorithm to perform block-by-block acceleration processing on the registered point cloud data to obtain block-by-block accelerated point cloud data; S4, performing feature extraction and feature segmentation on the block accelerated point cloud data to obtain key feature areas of the wind turbine blades; S5, performing defect recognition analysis on the key feature area based on the point cloud network enhancement model to obtain a defect detection result, wherein the defect detection result includes a defect location and a defect type; S6, generating a detection report including the defect location and the defect type, and visually displaying the detection report.

2. A wind farm wind turbine blade detection method based on three-dimensional point cloud as claimed in claim 1, characterized in that: Step S1 includes: S11, determining a plurality of scanning position points according to the size and shape characteristics of the wind turbine blade, wherein the scanning position points include a blade front scanning position point and a blade back scanning position point, wherein the blade front scanning position points are evenly distributed along the blade span direction, and the blade back scanning position point and the blade front scanning position point are arranged opposite to each other in the transverse direction of the blade; S12, placing a laser scanner at each scanning position point, adjusting the scanning parameters of the laser scanner, and performing all-round scanning and acquisition of the wind turbine blades, wherein the scanning parameters include scanning resolution, scanning angle, and scanning distance.

3. The wind farm wind turbine blade detection method based on three-dimensional point cloud according to claim 1, characterized in that: Step S2 includes: S21, using an iterative closest point algorithm to perform coarse registration to align the three-dimensional point cloud data at multiple angles; S22, performing precise registration by weighted averaging method to improve the registration accuracy, obtaining unified registration point cloud data, and adjusting the dynamic weight in the weighted averaging method according to the acquisition angle and acquisition distance of each set of registration point cloud data.

4. The wind farm wind turbine blade detection method based on three-dimensional point cloud according to claim 1, characterized in that: Step S3 includes: S31, constructing an octree spatial partitioning structure, including: analyzing the spatial density distribution of the registered point cloud data, determining the spatial range of the initial root node of the octree, and recursively segmenting the child nodes according to a preset spatial resolution standard until a dynamically adjusted depth limit is reached or the number of point clouds contained in each child node is lower than a threshold; dynamically adjusting the depth and number of branches of the octree according to the spatial distribution characteristics of the point cloud data; S32, mapping the constructed octree partition structure to multiple processing units, including: dynamically distributing the load to different processing units according to the spatial position and number of point clouds of each octree leaf node, and using a load balancing algorithm to optimize the distribution of computing tasks of each processing unit; and realizing parallel block processing of registration point cloud data.

5. The wind farm wind turbine blade detection method based on three-dimensional point cloud according to claim 1, characterized in that: Step S4 includes: S41, extracting key features from the unified registration point cloud data, the key features including geometric features and spectral features, including: calculating the main direction and curvature of each registration point based on the principal component analysis method to obtain geometric features, the geometric features are used to characterize the geometric shape of the point, and combining the multispectral imaging data to extract the spectral reflectance characteristics of each point as the spectral features; S42, detecting and classifying the key feature areas of the wind turbine blades based on the key features, including: using a support vector machine algorithm to classify points based on the geometric features and the spectral features, identifying potential abnormal feature areas, obtaining a preliminary classification result, and filtering and optimizing the preliminary classification result through a post-processing filtering optimization algorithm to obtain a key feature area; The calculation formula for the fan blade classification score is: Among them, Score q is the classification score of the fan at the qth point, f g (q) is the geometric eigenvalue of the qth point, f s (q) is the spectral characteristic value of the qth point, w g and w s are the weight coefficients of geometric features and spectral features, w g and w s The sum of is 1, γ is the post-processing adjustment coefficient, m is the parameter of the control function slope, and θ0 is the classification probability threshold; The calculation formula of the post-processing filtering optimization algorithm is: Among them, P final (q) is the key feature region division result of the qth point, 1 represents the corresponding point is in the key feature region, 0 represents the corresponding point is not in the key feature region, θ' is the post-processing classification threshold, and δ is the neighbor point number threshold.

6. A wind farm wind turbine blade detection method based on three-dimensional point cloud as claimed in claim 5, characterized in that: Step S5 includes: S51, based on the PointNet++ model, integrates the attention mechanism, adds a multi-scale feature extraction module and a residual connection structure to build a point cloud network enhancement model; S52, outputting a feature vector of the key feature area based on the point cloud network enhancement model, classifying the feature vector into defect types using a defect type classification algorithm to obtain defect types, locating the spatial position of each defect in the key feature area according to the three-dimensional spatial coordinate information to obtain the defect position, and screening and verifying the defect type and the defect position based on the confidence score; The calculation formula of the defect type classification algorithm is: Among them, P type (c|n) is the probability that the nth point belongs to the cth type of defect, z c (n) is the score of the nth point on the cth defect, z c' (n) is the score of the nth point on the c'th defect type, and C is the number of defect types; The confidence score is calculated as: Among them, Confidence t is the confidence score of the tth defect location, d t zx is the spatial distance of the t-th defect point, Score t is the classification score of the wind turbine at the t-th point, σ(·) is the sigmoid activation function, and β1 and β2 are the weight coefficients of the spatial distance and the classification score of the wind turbine, respectively.

7. The wind farm wind turbine blade detection method based on three-dimensional point cloud according to claim 1, characterized in that: Step S6 includes: S61, collating and summarizing the coordinate data of the detected defect positions, associating each defect position with the corresponding point defect type information, integrating the position information and the type information to obtain a test report in a standard format, wherein the test report includes the defect position information and the defect type data; S62, visualizing the inspection report to generate a defect distribution map of the three-dimensional model of the wind turbine blade, displaying the defect locations, using different colors or different icons to identify different types of defects, and using an interactive interface to support viewing and screening of defect information.

8. A wind farm wind turbine blade detection system based on three-dimensional point cloud, characterized in that: The system comprises: Point cloud data module, used to scan the fan blades at multiple angles using a laser scanner to obtain three-dimensional point cloud data of the fan blades; A data registration module is used to perform registration processing on the three-dimensional point cloud data, unify the three-dimensional point cloud data into the same coordinate system, and obtain registered point cloud data; A block acceleration module is used to perform block acceleration processing on the registration point cloud data using an octree partitioning algorithm to obtain block accelerated point cloud data; A feature extraction module, used to perform feature extraction and feature segmentation on the block acceleration point cloud data to obtain key feature areas of the wind turbine blades; A defect detection module is used to perform defect recognition analysis on the key feature area based on the point cloud network enhancement model to obtain a defect detection result, wherein the defect detection result includes a defect location and a defect type; The defect display module is used to generate a detection report including the defect location and the defect type, and to visually display the detection report.

9. An electronic device, characterized in that: include: at least one processor, at least one memory, a communication interface, and a bus; The processor, memory, and communication interface communicate with each other via the bus, the memory stores program instructions executable by the processor, and the processor calls the program instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions enable a computer to implement the method according to any one of claims 1 to 7.

Images