A method and system for detecting defects of a shell of a wind power control cabinet

Abstract

The present disclosure provides a wind power control cabinet shell defect detection method and system, which belongs to the technical field of wind power equipment detection. The method comprises the following steps: determining target data corresponding to the wind power control cabinet shell based on detection requirements, wherein the target data comprises image data and point cloud data; processing the image data to obtain a target shell image, and processing the point cloud data to obtain target point cloud data; and obtaining a wind power control cabinet shell defect detection result according to the target shell image and the target point cloud data. The wind power control cabinet shell defect detection method and system provided by the present disclosure can obtain clearer target data through image data processing and point cloud data processing, and the target data can comprehensively and accurately detect various defects of the wind power control cabinet shell, thereby improving the detection efficiency and accuracy.

Description

TECHNICAL FIELD

[0001] The present disclosure belongs to the technical field of wind power equipment detection, and more particularly to a wind power control cabinet shell defect detection method and system. BACKGROUND

[0002] As an important part of the wind power generation system, the integrity of the wind power control cabinet shell is crucial to the normal operation of the internal electrical equipment. Currently, the wind power control cabinet shell defect detection method has certain limitations; single data source cannot accurately reflect the defect condition of the shell. SUMMARY

[0003] The present disclosure aims to provide a wind power control cabinet shell defect detection method and system to improve the speed and accuracy of wind power control cabinet shell defect detection.

[0004] The first aspect of the present disclosure embodiment provides a wind power control cabinet shell defect detection method, comprising:

[0005] Based on the detection requirement, the target data corresponding to the wind power control cabinet shell is determined, and the target data includes image data and point cloud data;

[0006] The image data is processed to obtain the target shell image, and the point cloud data is processed to obtain the target point cloud data;

[0007] According to the target shell image and the target point cloud data, the wind power control cabinet shell defect detection result is obtained.

[0008] The second aspect of the present disclosure embodiment provides a wind power control cabinet shell defect detection system, comprising:

[0009] An image data processing module is configured to process the image data of the wind power control cabinet shell to obtain a target shell image;

[0010] A point cloud data processing module is configured to process the point cloud data of the wind power control cabinet shell to obtain target point cloud data;

[0011] A monitoring result output module is configured to input the target shell image and the target point cloud data into a target classification model to obtain a wind power control cabinet shell defect detection result.

[0012] The wind power control cabinet shell defect detection method and system provided by the present disclosure embodiment have the beneficial effect that through image data processing and point cloud data processing, clearer target data is obtained, which can comprehensively and accurately detect various defects of the wind power control cabinet shell, improving the detection efficiency and accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description only constitute some embodiments of the present disclosure, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0014] Figure 1 A flowchart of a shell defect detection method of a wind power control cabinet according to an embodiment of the present disclosure is provided.

[0015] Figure 2 A structural block diagram of a shell defect detection system of a wind power control cabinet according to an embodiment of the present disclosure is provided.

[0016] Figure 3 A schematic block diagram of an electronic device according to an embodiment of the present disclosure is provided. DETAILED DESCRIPTION

[0017] In the following description, specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. However, persons skilled in the art will understand that the present disclosure can be implemented in other embodiments without these specific details. In other cases, well-known systems, devices, circuits, and methods have not been described in detail in order not to obscure the description of the present disclosure with unnecessary details.

[0018] In order to make the objects, technical solutions and advantages of the present disclosure clearer, the following will be described with specific embodiments in conjunction with the drawings.

[0019] Reference will be made to Figure 1 , Figure 1 A flowchart of a shell defect detection method of a wind power control cabinet according to an embodiment of the present disclosure is provided. The method comprises:

[0020] S101: Determine the target data corresponding to the shell of the wind power control cabinet based on the detection requirement; the target data includes image data and point cloud data.

[0021] In the present embodiment, the detection requirement includes: appearance defect detection requirement and structural integrity detection requirement; wherein the appearance defect detection requirement includes: detecting fine cracks, depressions and coating integrity on the surface of the shell; the structural integrity detection requirement includes: the overall structure, weld, deformation, depression and hole position deviation of the control cabinet shell. The detection requirement also includes the requirement of the user for detecting multiple or one wind power control cabinet shell;

[0022] A multi-sensor array is deployed based on detection requirements, the multi-sensor array comprising: a high-resolution industrial camera array, a three-dimensional laser radar, and a structured light scanner; in this embodiment, image data and point cloud data are taken as target data, the image data of the wind power control cabinet shell to be detected is acquired through the high-resolution industrial camera array; the point cloud data of the wind power control cabinet shell to be detected is acquired through the three-dimensional laser radar and the structured light scanner.

[0023] S102: The image data is processed to obtain a target shell image, and the point cloud data is processed to obtain target point cloud data.

[0024] In this embodiment, the image data needs to adjust the definition of the image data to obtain target image data; the image data is easily affected by light, environment, shooting angle, and the like, and thus the image quality is affected, so the image needs to be processed, for example, the definition of the image is adjusted in the manner of image strengthening, image weakening, and the like, so as to obtain the target image data. The surface texture and visible defect information of the shell are collected through the industrial camera; for example, surface defects (scratches, rust), texture abnormalities (paint peeling), and micro-size changes (weld width fluctuation).

[0025] In this embodiment, the point cloud data of the shell is acquired through the three-dimensional laser scanner, and represents deformation, depression, and hole deviation; since there may be noise points and outliers in the point cloud data collection process, the collected point cloud data needs to be filtered. Common filtering methods include statistical filtering, radius filtering, bilateral filtering, and the like. Through filtering processing, the noise points and outliers in the point cloud data can be effectively removed, and the quality of the point cloud data is improved.

[0026] If multiple scanners are used or scanning is performed from different angles, multiple point cloud data needs to be registered to obtain complete shell point cloud data. The point cloud registration method includes an iterative closest point algorithm, a feature-based registration algorithm, and the like. Through point cloud registration, point cloud data from different sources can be aligned to the same coordinate system.

[0027] Different parts in the point cloud data also need to be segmented, for example, the shell surface is segmented from the surrounding environment, and the area where defects may exist is marked. A clustering-based segmentation method can be used, such as a density-based spatial clustering application algorithm, to segment the point cloud data. Through point cloud segmentation, the target area can be separated from the complex point cloud data, which is convenient for subsequent geometric feature extraction.

[0028] S103: According to the target shell image and the target point cloud data, a shell defect detection result of the wind power control cabinet is obtained.

[0029] In this embodiment, the image features of the target shell image and the geometric features of the target point cloud data are fused to obtain more comprehensive and accurate feature information. The feature-level fusion method can be used to splice the image features and the geometric features together to form a new feature vector. The decision-level fusion method can be used to judge the defects according to the image features of the target shell image and the geometric features of the target point cloud data, respectively, and then fuse the judgment results. The machine learning or deep learning model is used to classify and identify the fused features to judge whether the wind power control cabinet shell has defects and the type of the defects. For example, the target shell image and the target point cloud data are spatio-temporally aligned to generate a fusion feature matrix. The fusion feature matrix is input into a pre-trained defect classification model (for example, a random forest model, a support vector machine model) to output the shell defect type, position, and severity level.

[0030] As can be seen from the above, the shell defect detection method of the wind power control cabinet of the present disclosure can obtain clearer target data by processing image data and point cloud data, which can comprehensively and accurately detect various defects of the wind power control cabinet shell, and improve the detection efficiency and accuracy.

[0031] In an embodiment of the present disclosure, the image data of the wind power control cabinet shell is processed to obtain a target shell image, including:

[0032] The image data is processed by an adaptive grayscale algorithm to obtain a first grayscale image;

[0033] The first grayscale image is processed by a filtering algorithm to obtain the target shell image.

[0034] In this embodiment, the adaptive grayscale considers the color distribution, contrast, and other characteristics of different regions of the first plastic pipe image, and can divide the image into multiple sub-regions, calculate the grayscale parameter for each sub-region, so that the converted grayscale image can better preserve details and features in each local region. Grayscale processing is the process of converting a color image into a grayscale image. The first grayscale image is a grayscale image obtained by processing the image data by the adaptive grayscale algorithm, which converts the color information in the original color image into a grayscale value, while preserving the details of different regions by using the adaptive grayscale algorithm.

[0035] The filtering algorithm can reduce or eliminate the influence of noise by processing the pixel values in the image. The filtering algorithm can use mean filtering, Gaussian filtering, or median filtering, etc. Different filtering algorithms are suitable for different types of noise. The denoising processing is to remove the noise introduced in the image during acquisition, transmission, etc., and improve the image quality. Noise can manifest as random bright spots, dark spots or salt and pepper interference in the image, affecting the analysis and understanding of the image content. The target shell image is an image obtained by denoising the first gray image by the filtering algorithm. After denoising, the noise interference in the image is reduced, and the real characteristics of the shell are more clear, providing cleaner image data for subsequent processing.

[0036] In an embodiment of the present disclosure, the image data of the wind power control cabinet shell is processed to obtain a target shell image, further comprising: performing gray scale on the original image in the image data based on a brightness partition weighting algorithm to generate a first gray image;

[0037] Specifically, the RGB channel of the original image is divided into a high brightness area , a medium brightness area , and a low brightness area ; and the gray weight coefficient is dynamically adjusted according to the brightness partition.

[0038] The gray scale calculation formula is:

[0039]

[0040] wherein, is a first dynamic weight coefficient, is a second dynamic weight coefficient, is a third dynamic weight coefficient, .

[0041] For example, the second brightness threshold is greater than the first brightness threshold; the gray scale of the low brightness area is less than the first brightness threshold; the gray scale of the medium brightness area is greater than or equal to the first brightness threshold and less than the second brightness threshold; the gray scale of the high brightness area is greater than the second brightness threshold or equal to the second brightness threshold; the first brightness threshold and the second brightness threshold of the high brightness area, the medium brightness area and the low brightness area can be divided according to the gray scale of each pixel of the original image; the first brightness threshold and the second brightness threshold can also be determined by the three quantiles or adaptive method (such as Otsu multi-threshold segmentation).

[0042] In this embodiment, the weight of each brightness partition is calculated based on the contrast (standard deviation) of the R, G, and B channels of the brightness area:

[0043] The weight calculation formula is:

[0044] w F =(σ R +σ G+σ B ) / σ F

[0045] where F e {R, G, B}, σ R is the contrast of the high-light region, σ G is the contrast of the middle-light region, σ B is the contrast of the low-light region, w R +w G +w B = 1.

[0046] The embodiment can better preserve image information by enhancing the contrast of the high-light region and the dark details compared with the traditional gray-scale method.

[0047] In an embodiment of the present disclosure, the first gray-scale image is denoised based on a filtering algorithm to obtain a target shell image, including:

[0048] The first Gaussian kernel is determined based on the resolution of the first gray-scale image;

[0049] The target standard deviation is determined based on the noise intensity and the noise suppression coefficient of the first gray-scale image;

[0050] The target Gaussian kernel is obtained by adjusting the first Gaussian kernel based on the target standard deviation;

[0051] The first gray-scale image is denoised based on the target Gaussian kernel to obtain the target shell image.

[0052] In the embodiment, the first gray-scale image after gray-scale processing may contain noise due to factors such as sensor noise and light changes during image acquisition. Noise will interfere with subsequent feature extraction and defect detection, so denoising is needed. The present disclosure adopts an adaptive filtering method based on Gaussian kernel, which dynamically adjusts the Gaussian kernel parameters according to the resolution and noise intensity of the image. The resolution of the image reflects the degree of detail of the image, and different resolution images need different size Gaussian kernels when denoising. An image with high resolution contains more details, and a larger Gaussian kernel can be used to smooth the image; an image with low resolution uses a smaller Gaussian kernel to avoid excessive blurring of details.

[0053] Specifically, the initial Gaussian kernel size is calculated according to the resolution of the first gray-scale image

[0054] The calculation formula is:

[0055]

[0056] where W and H are the width and height of the image, respectively, and is an odd number; and ​​Construct the first Gaussian kernel with an initial standard deviation , where R is a preset value, such as 0.3.

[0057] Calculate the noise intensity of the first grayscale image by the local variance method , the calculation formula is:

[0058]

[0059] in, is the number of non-overlapping local regions in the image, For the i The variance of the region;

[0060] According to the noise intensity and the preset noise suppression coefficient , calculate the target standard deviation , the calculation formula is:

[0061] ;

[0062] The target Gaussian kernel is

[0063]

[0064] in, x and y are the pixel coordinates within the Gaussian kernel.

[0065] This embodiment uses the adjusted target Gaussian kernel to perform a convolution operation on the first grayscale image, replacing the value of each pixel in the image with the weighted average of the pixels in its neighborhood, with the weights determined by the Gaussian kernel. This effectively smoothes the image and removes noise while preserving as much image detail as possible.

[0066] In one embodiment of the present disclosure, performing denoising on the first grayscale image based on a filtering algorithm to obtain a target shell image further includes:

[0067] In response to the noise intensity of the first grayscale image being greater than or equal to the first noise intensity, increasing the noise suppression coefficient by a first step;

[0068] In response to the noise intensity of the first grayscale image being less than the first noise intensity, the noise suppression coefficient is reduced with a second step size.

[0069] In this embodiment, the noise intensity of the first grayscale image is greater than or equal to the first noise intensity, and the noise suppression coefficient is increased with the first step. When the noise intensity is large, increasing the noise suppression coefficient can enhance the denoising effect and make the image smoother.

[0070] The noise intensity of the first gray-scale image is less than the first noise intensity, and the noise suppression coefficient is reduced by the second step length. When the noise intensity is small, reducing the noise suppression coefficient can reduce the impact on image details and avoid over-smoothing.

[0071] In the embodiment, the first step length is calculated according to a first step length base value, the noise intensity, the first noise intensity, and a preset maximum noise intensity.

[0072] The second step length is calculated according to a second step length base value, the noise intensity, and the first noise intensity.

[0073] The first step length calculation formula is:

[0074]

[0075] wherein, the first step length base value is, the preset maximum noise intensity is, the first step length is, the first noise intensity is;

[0076] The second step length calculation formula is:

[0077]

[0078] wherein, the second step length base value is, the second step length is.

[0079] Specifically, the first step length and the second step length adaptively change with the difference between the noise intensity and the threshold value, quickly increase the suppression degree in strong noise, and finely adjust in weak noise; the image can be effectively smoothed, the noise can be removed, and the details of the image can be retained.

[0080] The embodiment also includes adjusting the first step length and the second step length according to the average of the gray-scale differences between adjacent pixels (local contrast) in the first gray-scale image; wherein, if the average of the gray-scale differences between adjacent pixels is greater than a preset value, the first step length and the second step length are calculated according to a decay coefficient; the decay coefficient can be obtained according to experience. The disclosure avoids the edge blurring problem of traditional Gaussian filtering, improves the defect detection accuracy, reduces the step length adjustment range in high-contrast areas (which may be defect edges), prevents feature loss caused by over-smoothing, and balances noise suppression and image feature retention.

[0081] In one embodiment of the disclosure, point cloud data of a wind power control cabinet shell is processed to obtain target point cloud data, including:

[0082] The point cloud data is preprocessed to obtain first point cloud data;

[0083] Calculate curvature information of the first point cloud data, add the curvature information to the first point cloud data to obtain second point cloud data;

[0084] Register the second point cloud data by using a normal distribution transformation algorithm to obtain target point cloud data.

[0085] In the present embodiment, the point cloud data may have uneven data distribution such as noise points and outliers. The preprocessing improves the quality of the point cloud data, providing high-quality point cloud data for subsequent curvature calculation.

[0086] Curvature is an important feature for describing the local geometry of a point cloud surface, and can reflect the concave-convex degree of the point cloud surface. In the defect detection of a wind power control cabinet shell, curvature information can help us better identify defects on the surface, such as cracks and holes.

[0087] The calculated curvature information is added as a new attribute to the first point cloud data to obtain second point cloud data containing curvature information. This can enable dynamic optimization of voxels during registration of the second point cloud data, improving the registration efficiency and defect area alignment accuracy of the normal distribution transformation algorithm for the second point cloud data.

[0088] In one embodiment of the present disclosure, the second point cloud data is registered by using a normal distribution transformation algorithm to obtain point cloud data, comprising:

[0089] Determine the voxel size according to the point cloud density of the target point cloud data;

[0090] Determine a size adjustment coefficient of the voxel size according to the curvature information;

[0091] Adjust the voxel size based on the size adjustment coefficient to obtain a target voxel size;

[0092] Register the second point cloud data based on the normal distribution transformation algorithm and the target voxel size to obtain the target point cloud data.

[0093] In the present example, point cloud data of different densities requires different voxel sizes for effective processing. When the point cloud density is high, using a smaller voxel size can retain more detailed information. When the point cloud density is low, a larger voxel size can improve the calculation efficiency while avoiding the problem of inaccurate statistical information due to too few points in the voxel.

[0094] In areas with large curvature, the point cloud surface changes more dramatically and may have more detailed information. In areas with small curvature, the point cloud surface is relatively smooth. Therefore, in order to better retain the details of the area with large curvature during registration, the voxel size needs to be adjusted according to the curvature information.

[0095] In an embodiment of the present disclosure, in response to the curvature value in the curvature information being less than a first curvature threshold, the size adjustment coefficient is increased by a third step size;

[0096] In response to the curvature value in the curvature information being greater than a second curvature threshold, the size adjustment coefficient is decreased by a fourth step size.

[0097] In the present embodiment, the average number of points of the point cloud in a certain region is calculated to determine the point cloud density. According to a pre-set density-voxel size mapping relationship, a suitable voxel size is determined. The curvature value of each point of the point cloud is calculated, and then the size adjustment coefficient is determined according to the curvature distribution.

[0098] When the curvature value is small, it indicates that the surface of the region is relatively smooth, for example, a large-area flat surface part of the wind power control cabinet shell. At this time, the size adjustment coefficient is increased, which can enable the subsequent processing (such as the meshing, filtering and other operations of the point cloud) to be performed in a larger range, and help to capture the subtle defects or feature changes that may exist in the region;

[0099] When the curvature value is large, it indicates that the surface of the region is curved to a large extent, and there may be features such as edges, corners or defects, for example, around the mounting hole or at the weld of the wind power control cabinet shell. Reducing the size adjustment coefficient can make the processing more fine and focus on these local features.

[0100] The present disclosure realizes the balance between global features and local details through intelligent dynamic optimization of the size adjustment coefficient in the shell defect detection of the wind power control cabinet, and improves the accuracy and efficiency of defect detection.

[0101] In an embodiment of the present disclosure, the shell defect detection method of the wind power control cabinet further comprises:

[0102] For each point in the second point cloud data, the curvature gradient of the point is calculated , which is defined as the curvature change rate of the point and the adjacent points in the neighborhood:

[0103]

[0104] wherein N is the number of points in the neighborhood, is a point of the second point cloud data; is the curvature value of the first point;

[0105] According to the curvature value and the curvature gradient , the point cloud region is classified as:

[0106] a smooth mutation region ( ≥ and ≥ ​), the curvature is high and changes sharply, and can be a defect edge;

[0107] a smooth transition zone (a curvature value is less than a second curvature value threshold, and a curvature gradient is low and changes gently, and can be a normal plane; and a smooth transition zone (a curvature value is less than a second curvature value threshold, and a curvature gradient is low and changes gently, and can be a normal plane;

[0108] a noise disturbance zone (a curvature value is less than a second curvature value threshold, and a curvature gradient is low and changes gently, and can be noise or a small structure; and a noise disturbance zone (a curvature value is less than a second curvature value threshold, and a curvature gradient is low and changes gently, and can be noise or a small structure; is a curvature gradient threshold value, is a first curvature threshold value.

[0109] In the smooth mutation zone, the size adjustment coefficient is increased by a fifth step length in response to the curvature value in the curvature information being greater than the first curvature threshold value;

[0110] In the smooth transition zone, the size adjustment coefficient is decreased by a sixth step length in response to the curvature value in the curvature information being less than the second curvature value threshold.

[0111] The fifth step length calculation formula is:

[0112]

[0113] is a fifth step length, is a fifth step length base value.

[0114] The sixth step length calculation formula is:

[0115]

[0116] wherein, is a sixth step length, is a sixth step length base value, is a second curvature value threshold.

[0117] The present disclosure introduces a curvature gradient as a basis for region classification, distinguishing between real defect edges (high curvature + high gradient) and noise disturbances (low curvature + high gradient), avoiding the misjudgment problem of traditional single curvature threshold, suppressing noise interference while retaining defect features, improving registration robustness, and improving the registration accuracy of the defect region.

[0118] In an embodiment of the present disclosure, according to the target shell image and the target point cloud data, a shell defect detection result of the wind power control cabinet is obtained, comprising:

[0119] According to the target shell image and the target point cloud data, a fusion matrix is constructed, and the matrix dimensions of the fusion matrix include: texture abnormality, curvature mutation coefficient and three-dimensional deformation parameter;

[0120] Input the fusion matrix into the target classification model to obtain a shell defect detection result of the wind power control cabinet.

[0121] The target classification model is obtained by training a random forest model based on first training data.

[0122] In this embodiment, a fusion matrix is constructed according to the target image and the target point cloud data. The fusion matrix includes information of multiple dimensions, which can comprehensively and accurately describe the characteristics of the target wind power control cabinet shell. The matrix dimensions of the fusion matrix include: texture abnormality degree, curvature mutation coefficient and three-dimensional deformation parameter; the texture abnormality degree is determined by detailed texture analysis of the target image. The image processing algorithm is used to extract the texture features of the target image. The calculation of the curvature mutation coefficient depends on the target point cloud data. The point cloud data is an accurate record of the three-dimensional space information of the target object surface. Through processing and analysis of these data, the curvature information of the object surface can be obtained; the curvature mutation coefficient is calculated for the defect candidate region in the target point cloud data. The three-dimensional deformation parameter is used to describe the shape change of the target wind power control cabinet shell in the three-dimensional space. The registered target point cloud data is aligned with the standard model of the target wind power control cabinet shell, and the deformation distance of each point is calculated to obtain the three-dimensional deformation of the target wind power control cabinet shell.

[0123] The target classification model is constructed based on the random forest algorithm; the random forest is an ensemble learning method, which combines multiple decision trees to perform classification and regression tasks, and has high accuracy and robustness. The constructed fusion matrix is input into the trained target classification model, and the model analyzes and judges the fusion matrix according to the learned feature pattern; the defect detection result of the target wind power control cabinet shell is output, which can be a specific defect type, such as scratch, indentation, deformation, etc., or a severity level of the defect. For example, if the model judges that the fusion matrix corresponds to the wind power control cabinet shell with obvious texture abnormality and curvature mutation, it can output the defect detection result of surface scratch and local indentation, and give the corresponding severity level according to the specific situation of the defect.

[0124] In one embodiment of the present disclosure, according to the target shell image and the target point cloud data, the shell defect detection result of the wind power control cabinet is obtained, further comprising:

[0125] The first ratio is calculated according to the number of matrix dimensions of the fusion matrix and the ratio formula, and the compensation value is determined according to the first ratio, which is used to compensate the base depth of the decision tree in the target classification model.

[0126] The target depth of the decision tree in the target classification model is calculated by weighting the compensation value and the base depth.

[0127] In this embodiment, the first ratio is calculated according to a ratio formula of the number of the maximum feature categories and the number of the minimum feature categories in the matrix dimension of the fusion matrix; when the first ratio is greater than a ratio threshold, a compensation value is calculated according to a preset compensation value calculation formula; a target depth of a decision tree in the target classification model is calculated by weighting the compensation value and the base depth; and the target classification model is constrained according to the target depth.

[0128] The features include scratches, depressions, deformations, etc.

[0129] In this embodiment, the compensation value calculation formula is:

[0130]

[0131] wherein, is the compensation value, is the first ratio, is an intermediate parameter, used to suppress the surge of the compensation amount under a high is the number of the minimum feature categories.

[0132] This embodiment can dynamically adjust the depth of the decision tree in the target classification model according to the feature information in the fusion matrix, thereby improving the performance of the model in detecting the shell defects of the target wind power control cabinet and better adapting to different data features and actual application scenarios.

[0133] The shell defect detection method of the wind power control cabinet corresponding to the above embodiment, Figure 2 is a structural block diagram of a shell defect detection device of a wind power control cabinet provided by an embodiment of the present disclosure. For ease of illustration, only parts related to the embodiments of the present disclosure are shown. For parts not related to the embodiments of the present disclosure, refer to the description of the prior art. Figure 2 The shell defect detection device 20 of the wind power control cabinet includes a target data determination module 21, a target data processing module 22, and a monitoring result output module 23.

[0134] The target data determination module 21 is configured to determine target data corresponding to the shell of the wind power control cabinet based on detection requirements; the target data includes image data and point cloud data.

[0135] The target data processing module 22 is configured to process the image data to obtain a target shell image, and is further configured to process the point cloud data to obtain target point cloud data.

[0136] The monitoring result output module 23 is configured to obtain a shell defect detection result of the wind power control cabinet according to the target shell image and the target point cloud data.

[0137] In an embodiment of the present disclosure, the target data processing module 22 is specifically configured to: ​

[0138] The image data is subjected to grayscale processing based on an adaptive grayscale algorithm to obtain a first grayscale image;

[0139] The first grayscale image is subjected to denoising processing based on a filtering algorithm to obtain a target shell image.

[0140] In an embodiment of the present disclosure, the target data processing module 22 is specifically configured to:

[0141] determine a first Gaussian kernel based on a resolution of the first grayscale image;

[0142] determine a target standard deviation based on a noise intensity of the first grayscale image and a noise suppression coefficient;

[0143] adjust the first Gaussian kernel based on the target standard deviation to obtain a target Gaussian kernel;

[0144] subject the first grayscale image to denoising processing based on the target Gaussian kernel to obtain the target shell image.

[0145] In an embodiment of the present disclosure, the target data processing module 22 is specifically configured to:

[0146] in response to the noise intensity of the first grayscale image being greater than or equal to a first noise intensity, increase the noise suppression coefficient by a first step size;

[0147] in response to the noise intensity of the first grayscale image being less than the first noise intensity, decrease the noise suppression coefficient by a second step size.

[0148] In an embodiment of the present disclosure, the target data processing module 22 is specifically configured to:

[0149] subject the point cloud data to preprocessing to obtain first point cloud data;

[0150] calculate curvature information of the first point cloud data, and add the curvature information to the first point cloud data to obtain second point cloud data;

[0151] subject the second point cloud data to registration by using a normal distribution transformation algorithm to obtain target point cloud data.

[0152] In an embodiment of the present disclosure, the target data processing module 22 is specifically configured to:

[0153] determine a voxel size according to a point cloud density of the target point cloud data;

[0154] determine a size adjustment coefficient of the voxel size according to the curvature information;

[0155] adjust the voxel size based on the size adjustment coefficient to obtain a target voxel size;

[0156] The second point cloud data is registered based on a normal distribution transformation algorithm and a target voxel size, to obtain target point cloud data.

[0157] In an embodiment of the present disclosure, the target data processing module 22 is specifically configured to:

[0158] In response to the curvature value in the curvature information being less than the first curvature threshold, increasing the size adjustment coefficient by a third step size;

[0159] In response to the curvature value in the curvature information being greater than the second curvature threshold, decreasing the size adjustment coefficient by a fourth step size.

[0160] In an embodiment of the present disclosure, the monitoring result output module 23 is specifically configured to:

[0161] According to the target shell image and the target point cloud data, a fusion matrix is constructed, and the matrix dimensions of the fusion matrix include a texture abnormality degree, a curvature mutation coefficient and a three-dimensional deformation parameter;

[0162] The fusion matrix is input into the target classification model to obtain the shell defect detection result of the wind power control cabinet.

[0163] The target classification model is obtained by training the random forest model based on first training data. The first training data is fusion matrix data of historical defect samples.

[0164] In an embodiment of the present disclosure, the monitoring result output module 23 is specifically configured to:

[0165] The feature quantity of the fusion matrix is used to determine a base depth of a decision tree in the target classification model.

[0166] A first ratio is calculated according to the number of matrix dimensions of the fusion matrix and a ratio formula, and a compensation value is determined according to the first ratio, the compensation value being used to compensate the base depth of the decision tree in the target classification model.

[0167] The compensation value and the base depth are weighted to obtain a target depth of the decision tree in the target classification model.

[0168] The shell defect detection system of the wind power control cabinet provided in the embodiments of the present disclosure can comprehensively and accurately detect various defects of the shell of the wind power control cabinet by comprehensively considering image features and geometric features, and combining advanced image processing and point cloud data processing technology, thereby improving the detection efficiency and accuracy.

[0169] Referring to Figure 3 , Figure 3 The schematic block diagram of the electronic device provided in an embodiment of the present disclosure is shown in FIG. 1. As shown in FIG. 1, the electronic device includes a processor 10, a memory 20 and a communication interface 30. Figure 3The electronic device 300 in the embodiment shown can include one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The above-mentioned processors 301, input devices 302, output devices 303, and memories 304 complete communication with each other through a communication bus 305. The memory 304 is configured to store a computer program, and the computer program includes program instructions. The processor 301 is configured to execute the program instructions stored in the memory 304. Specifically, the processor 301 is configured to invoke the program instructions to execute the functions of the modules in the above-mentioned device embodiments, for example Figure 2 The functions of the target data determination module 22, the target data processing module 22, and the monitoring result output module 23 are shown.

[0170] It should be understood that, in the embodiments of the present disclosure, the processor 301 can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0171] The input device 302 can include a touchpad, a fingerprint collection sensor (used to collect fingerprint information and direction information of a fingerprint of a user), a microphone, etc., and the output device 303 can include a display (LCD, etc.), a speaker, etc.

[0172] The memory 304 can include read-only memory and random access memory, and provide instructions and data for the processor 301. A portion of the memory 304 can also include non-volatile random access memory. For example, the memory 304 can also store device type information.

[0173] In specific implementations, the processor 301, the input device 302, and the output device 303 described in the embodiments of the present disclosure can execute the implementation manners described in the first and second embodiments of the shell defect detection method of the wind power control cabinet provided by the embodiments of the present disclosure, and can also execute the implementation manners of the electronic device described in the embodiments of the present disclosure, which will not be described here.

[0174] In another embodiment of the present disclosure, a computer readable storage medium is provided, which stores a computer program. The computer program includes program instructions, which, when executed by a processor, implement all or part of the processes of the above-mentioned embodiment methods. The computer program can also instruct related hardware to complete the above-mentioned processes. The computer program can be stored in a computer readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0175] The computer readable storage medium can be an internal storage unit of the electronic device of any of the above-mentioned embodiments, such as a hard disk or a memory of the electronic device. The computer readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the computer readable storage medium can include both the internal storage unit and the external storage device of the electronic device. The computer readable storage medium is used to store the computer program and other programs and data required by the electronic device. The computer readable storage medium can also be used to temporarily store data that has been output or will be output.

[0176] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in the above description in general terms. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present disclosure.

[0177] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the electronic device and the units described above can refer to the corresponding processes in the above-mentioned method embodiments, which will not be described here.

[0178] In several embodiments provided in the present application, it should be understood that the disclosed electronic device and method can be implemented in other manners. For example, the embodiments of the apparatus described above are merely illustrative. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling or direct coupling or communication connection between the units can be indirect coupling or communication connection through some interfaces, or can be in electrical, mechanical or other forms.

[0179] The units described as separated components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purposes of the embodiments of the present disclosure.

[0180] In addition, each functional unit in the various embodiments of the present disclosure can be integrated in one processing unit, or each unit can exist physically as a separate unit, or two or more units can be integrated in one unit. The integrated unit can be implemented in the form of hardware, or in the form of a software functional unit.

[0181] The above is merely specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto, and any skilled person in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present disclosure, and these modifications or replacements should be covered in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the protection scope of the claims.

Claims

1. A method of detecting defects in an outer shell of a wind power control cabinet, characterized in that The method comprises the following steps: determining target data corresponding to the wind power control cabinet shell based on detection requirements, the target data including image data and point cloud data; The image data is grayscaled based on an adaptive grayscale algorithm, and the original image in the image data is grayscaled according to a brightness partition weighted algorithm to generate a first grayscale image. The grayscale calculation formula is: ,in, is the first dynamic weight coefficient, is the second dynamic weight coefficient, is the third dynamic weight coefficient, , the weight of each brightness partition is determined based on the R, G, and B channel contrast of the brightness partition. The weight calculation formula is: , where F∈{R,G,B}, σ R is the contrast of high brightness area, σ G is the contrast in the medium brightness area, σ B Contrast in low-brightness areas; determining a first Gaussian kernel based on the resolution of the first grayscale image; determining a target standard deviation based on the noise intensity of the first grayscale image and a noise suppression coefficient; adjusting the first Gaussian kernel to obtain a target Gaussian kernel based on the target standard deviation; performing denoising processing on the first grayscale image based on the target Gaussian kernel to obtain a target shell image; preprocessing the point cloud data to obtain first point cloud data; calculating the curvature information of the first point cloud data and adding the curvature information to the first point cloud data to obtain second point cloud data; For each point in the second point cloud data, a curvature gradient thereof is determined, and a type of a point cloud region is determined according to a curvature value of each point of the point cloud and the curvature gradient, the type of the point cloud region including a smooth mutation region, a gentle transition region, and a noise disturbance region, and the curvature gradient is calculated according to a formula wherein, N is a number of points in a neighborhood, i is a point of the second point cloud data; C i is a curvature value of the first i point. determining a size adjustment coefficient according to the point cloud region; registering the second point cloud data according to a normal distribution transformation algorithm and the size adjustment coefficient to obtain target point cloud data; obtaining the shell defect detection result of the wind power control cabinet according to the target shell image and the target point cloud data.

2. The wind power control cabinet shell defect detection method according to claim 1, wherein Further comprising: in response to the noise intensity of the first grayscale image being greater than or equal to a first noise intensity, increasing the noise suppression coefficient by a first step size; in response to the noise intensity of the first grayscale image being less than the first noise intensity, decreasing the noise suppression coefficient by a second step size.

3. The method of claim 1, wherein the method further comprises: According to the normal distribution transformation algorithm, the second point cloud data is registered to obtain target point cloud data, comprising: determining a voxel size according to the point cloud density of the target point cloud data; determining a size adjustment coefficient of the voxel size according to the curvature information; adjusting the voxel size based on the size adjustment coefficient to obtain a target voxel size; registering the second point cloud data based on the normal distribution transformation algorithm and the target voxel size to obtain target point cloud data.

4. The wind power control cabinet shell defect detection method according to claim 3, characterized in that, Further comprising: in response to the curvature value in the curvature information being less than a first curvature threshold, increasing the size adjustment coefficient by a third step size; in response to the curvature value in the curvature information being greater than a second curvature threshold, decreasing the size adjustment coefficient by a fourth step size.

5. The method of claim 1, wherein the method further comprises: The method of obtaining the shell defect detection result of the wind power control cabinet according to the target shell image and the target point cloud data comprises: constructing a fusion matrix according to the target shell image and the target point cloud data, the matrix dimensions of the fusion matrix including texture anomaly degree, curvature mutation coefficient and three-dimensional deformation parameter; inputting the fusion matrix into a target classification model to obtain the shell defect detection result of the wind power control cabinet; wherein the target classification model is obtained by training a random forest model based on first training data; the first training data is the fusion matrix data of historical defect samples.

6. The wind power control cabinet shell defect detection method according to claim 5, wherein Further comprising: calculating a first ratio according to the number of matrix dimensions of the fusion matrix and a ratio formula, and determining a compensation value according to the first ratio, the compensation value being used to compensate the base depth of the decision tree in the target classification model; performing weighted calculation on the compensation value and the base depth to obtain the target depth of the decision tree in the target classification model.

7. A wind power control cabinet shell defect detection system, characterized in that, The method comprises the following steps: a target data determination module for determining target data corresponding to the wind power control cabinet shell based on detection requirements; the target data including image data and point cloud data; The target data processing module is configured to perform gray-scale processing on the image data based on an adaptive gray-scale algorithm, perform gray-scale processing on an original image in the image data according to a brightness partition weighting algorithm, and generate a first gray-scale image, wherein the gray-scale calculation formula is wherein, is a first dynamic weight coefficient, is a second dynamic weight coefficient, is a third dynamic weight coefficient, determines the weight of each brightness partition based on the contrast of the R, G and B channels of the brightness partition, and the weight calculation formula is wherein, F ∈ {R, G, B}, σ R is the contrast of the high-brightness area, σ G is the contrast of the medium-brightness area, and σ B is the contrast of the low-brightness area. determining a first Gaussian kernel based on a resolution of the first grayscale image; determining a target standard deviation based on a noise intensity and a noise suppression coefficient of the first grayscale image; adjusting the first Gaussian kernel based on the target standard deviation to obtain a target Gaussian kernel; performing denoising processing on the first grayscale image based on the target Gaussian kernel to obtain a target shell image; performing preprocessing on the point cloud data to obtain first point cloud data; calculating curvature information of the first point cloud data and adding the curvature information to the first point cloud data to obtain second point cloud data; For each point in the second point cloud data, a curvature gradient thereof is determined, and a type of a point cloud region is determined according to a curvature value of each point of the point cloud and the curvature gradient, the type of the point cloud region including a smooth mutation region, a gentle transition region, and a noise disturbance region, and the curvature gradient is calculated according to a formula wherein, N is a number of points in a neighborhood, i is a point of the second point cloud data; C i is a curvature value of the first i point. determining a size adjustment coefficient according to the point cloud region; performing registration on the second point cloud data according to a normal distribution transformation algorithm and the size adjustment coefficient to obtain target point cloud data; a monitoring result output module configured to obtain a shell defect detection result of the wind power control cabinet according to the target shell image and the target point cloud data.

Images