Intelligent classification method and system for defects of tension clamp of power transmission line based on two-dimensional image

Abstract

The application discloses a kind of based on two-dimensional image transmission line strain clamp defect intelligent classification method and system, belong to the field of strain clamp defect classification, including: the two-dimensional image of strain clamp is preprocessed, relevant features are extracted and candidate set of defect area is constructed.First, edge information is obtained using edge detection algorithm, and curvature and roughness are calculated to generate morphological feature vector.Then, support vector machine model is trained based on defect sample set, and the first candidate set is classified to obtain the first defect classification result.Next, the density feature correction is carried out on the clamp image of different materials, the gray gradient feature is extracted, the second candidate set is obtained, and the trained SVM model is input for the second defect classification.Finally, the first and second classification results are integrated to generate the final defect classification result.This method effectively combines morphological features and material features, improving the accuracy and robustness of defect detection.

Description

Technical Field

[0001] This invention belongs to the field of tension clamp defect classification, and particularly relates to an intelligent classification method and system for transmission line tension clamp defects based on two-dimensional images. Background Technology

[0002] Intelligent classification of defects in transmission line tension clamps relies on two-dimensional image analysis. However, in practical applications, a unique technical challenge exists: how to distinguish defects with similar density but significantly different morphologies. Some defects, such as early corrosion and certain types of wear, may exhibit similar pixel value distributions and gray-level gradient characteristics in two-dimensional images, meaning they have similar image densities. However, their actual morphologies may differ significantly. Corrosion may manifest as rough, patchy surfaces, while wear may present as smooth depressions or grooves.

[0003] Due to the complex environment in which tension clamps operate, factors such as lighting, rain, and dirt can further interfere with image quality, making the density characteristics of defective areas more blurred. For example, a thin layer of dust can alter the grayscale value of a corroded area, making it even more difficult to distinguish from a worn area. Furthermore, the material and surface treatment process of the tension clamp also affect the image representation of defects. Different materials have different reflectivities, causing even the same defect to exhibit different density characteristics on clamps made of different materials. This further increases the difficulty of fine-grained defect classification based on image density. Therefore, establishing a more effective differentiation mechanism between defects with similar density characteristics is a key technological bottleneck for achieving intelligent classification of defects in transmission line tension clamps. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides an intelligent classification method for defects in transmission line tension clamps based on two-dimensional images, comprising:

[0005] A two-dimensional image of the tension clamp is acquired, and the two-dimensional image of the tension clamp is preprocessed to obtain a first image;

[0006] Extract relevant features from the first image, and construct a first candidate set of defect regions based on the relevant features;

[0007] Edge information of the first candidate set is extracted using an edge detection algorithm, and the curvature and roughness of the edges are calculated to obtain morphological feature vectors.

[0008] Obtain a defect sample set, and train the support vector machine model based on the defect sample set to obtain the trained support vector machine model;

[0009] The first candidate set is input into the trained support vector machine model for calculation to obtain the first defect classification result;

[0010] Density feature correction is performed on wire clip images of different materials to obtain a second image. The pixel value distribution and gray-level gradient features of the second image are extracted to obtain a second candidate set.

[0011] The second candidate set and the morphological feature vector are input into the trained support vector machine model for calculation to obtain the second defect classification result;

[0012] The first defect classification result and the second defect classification result are integrated to generate the final classification result.

[0013] Preferably, the process of obtaining the first image includes: acquiring a two-dimensional image of the tension clamp to obtain an original image, and performing image grayscale conversion, normalization, and denoising operations on the original image to obtain the first image.

[0014] Preferably, the process of constructing a first candidate set of defect regions based on the relevant features includes:

[0015] For each pixel in the first image, the frequency of its pixel value is counted to obtain a pixel value distribution histogram.

[0016] The Sobel operator is used to calculate the horizontal and vertical gradients of each pixel in the first image to obtain a gradient image.

[0017] The pixel value distribution histogram and the gradient image are fused. If the pixel gradient value is greater than a preset threshold and the pixel value is located in the abnormal range of the pixel value distribution histogram, then the pixel is determined to be a candidate pixel.

[0018] A depth-first search algorithm is used to connect adjacent candidate pixels into connected regions, resulting in a set of multiple connected regions.

[0019] Calculate the pixel density of each connected region. If the pixel density of a connected region is greater than a preset density threshold, then the region is determined to be a defect candidate region and constitutes the first candidate set.

[0020] Preferably, the process of obtaining the morphological feature vector includes:

[0021] The edge roughness is calculated by calculating the fluctuation of the gray values ​​of the edge pixels in the first candidate set, and the curvature is calculated to obtain the curvature value.

[0022] The obtained edge curvature and edge roughness values ​​are combined into a morphological feature vector.

[0023] Preferably, the process of obtaining the trained support vector machine model includes:

[0024] Construct a sample library containing various defect types such as early corrosion and wear, where each sample contains density feature and morphological feature vector;

[0025] Sample feature vectors are obtained from the sample library and input into the support vector machine model for training. During the training process, if the density feature vectors of two samples are similar, the difference between the morphological feature vectors is trained to obtain the trained support vector machine model.

[0026] Preferably, the process of obtaining the first defect classification result includes:

[0027] The first candidate set is input into the trained support vector machine model for calculation, and the candidate region is the probability value of corrosion or wear;

[0028] Based on the probability value, the category of the candidate region is determined. If the corrosion probability is greater than a preset threshold, the candidate region is marked as a corrosion defect. If the wear probability is greater than a preset threshold and the corrosion probability is less than the wear probability, the candidate region is marked as a wear defect, and the first defect classification result is generated.

[0029] Preferably, the process of obtaining the second candidate set includes:

[0030] Retrieve the material properties and surface treatment parameters for each type of wire clamp from the wire clamp database;

[0031] A dataset of wire clamp images was obtained by photographing wire clamps made of different materials and with different surface treatments using an industrial camera.

[0032] The wire clip image is preprocessed using grayscale conversion and histogram equalization to obtain the preprocessed wire clip image.

[0033] The density features of the preprocessed wire clip image are extracted using the gray-level co-occurrence matrix algorithm to obtain the texture information of the wire clip image;

[0034] Based on the material and surface treatment process information of the wire clamp, a mapping relationship model between the material reflectivity and density characteristics is established.

[0035] Based on the reflectivity compensation model, the density characteristics of the wire clip image are corrected to compensate for the reflectivity differences caused by the material and surface treatment process, and the corrected density characteristics are obtained.

[0036] Based on the corrected density features, the wire clip image is reconstructed to obtain the second candidate set.

[0037] Preferably, the process of obtaining the second defect classification result includes:

[0038] Obtain the pixel value distribution of the second image: Calculate the occurrence frequency of each pixel value in the second image, construct a pixel value distribution histogram, and obtain the pixel value distribution features;

[0039] Obtain the gray-level gradient features of the second image: The Sobel operator is used to calculate the gradients in the horizontal and vertical directions of the second image to obtain the gradient magnitude and direction information, which constitute the gray-level gradient features;

[0040] The second defect classification result is obtained based on the pixel value distribution features and the grayscale gradient features.

[0041] To address the aforementioned technical problems, this invention also provides an intelligent classification system for defects in transmission line tension clamps based on two-dimensional images, comprising:

[0042] The first image processing module is used to acquire a two-dimensional image of the tension clamp, and to preprocess the two-dimensional image of the tension clamp to obtain a first image;

[0043] The first candidate processing module is used to extract relevant features from the first image and construct a first candidate set of defect regions based on the relevant features.

[0044] The morphological feature vector construction module is used to extract edge information of the first candidate set through an edge detection algorithm, and calculate the curvature and roughness of the edges to obtain morphological feature vectors.

[0045] The model building module is used to acquire a defect sample set, train the support vector machine model based on the defect sample set, and obtain the trained support vector machine model.

[0046] The first defect classification result generation module is used to input the first candidate set into the trained support vector machine model for calculation to obtain the first defect classification result.

[0047] The second candidate processing module is used to perform density feature correction on wire clip images of different materials to obtain a second image, extract the pixel value distribution and gray-level gradient features of the second image, and obtain a second candidate set.

[0048] The second image processing module is used to input the second candidate set and the morphological feature vector into the trained support vector machine model for calculation to obtain the second defect classification result.

[0049] The fusion module is used to integrate the first defect classification result and the second defect classification result to generate the final classification result.

[0050] Compared with the prior art, the present invention has the following advantages and technical effects:

[0051] This invention first preprocesses the acquired two-dimensional images of tension clamps to eliminate the effects of light and rainwater contamination, and extracts density features such as pixel value distribution and grayscale gradient to construct a preliminary defect candidate set. Since corrosion and wear defects may have similar density features, this invention further extracts morphological features such as edge curvature and roughness of the candidate regions. Combined with a pre-built database containing corrosion and wear samples, a support vector machine model is trained to achieve defect differentiation based on morphological features. Considering the differences in density features between clamps made of different materials, this invention further corrects the density features based on the clamp material information, extracts density features again, and combines them with morphological features. The trained support vector machine model is then used for more accurate defect classification. This invention significantly improves the accuracy of identifying corrosion and wear defects in tension clamps by fusing density and morphological features and correcting for material differences. Attached Figure Description

[0052] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0053] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of the detection process according to an embodiment of the present invention;

[0055] Figure 3 This is another detection schematic diagram according to an embodiment of the present invention;

[0056] Figure 4 This is a flowchart illustrating the construction process of the first candidate set in an embodiment of the present invention.

[0057] Figure 5 This is a flowchart illustrating the construction process of the second candidate set in an embodiment of the present invention.

[0058] Figure 6 This is a schematic diagram of the system modules in an embodiment of the present invention. Detailed Implementation

[0059] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0060] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0061] Example 1

[0062] like Figure 1 As shown, this embodiment provides an intelligent classification method for defects in transmission line tension clamps based on two-dimensional images, including:

[0063] Step S101: Obtain a two-dimensional image of the tension clamp, and preprocess the image to eliminate the effects of changes in lighting conditions and interference from rainwater and dirt, including image grayscale conversion, normalization and denoising, to obtain the first image.

[0064] A two-dimensional image of the tension clamp is obtained to obtain the original image. The original image is converted to a grayscale image to obtain the first image. If the pixel values ​​of the image are not uniformly distributed, histogram equalization is performed to improve the image contrast, resulting in the second image. The second image is normalized to scale the pixel values ​​to a specific range, such as between 0 and 1, resulting in the third image. If the image contains salt-and-pepper noise, median filtering is used to remove the noise, resulting in the fourth image. If the image contains Gaussian noise, Gaussian filtering is used to remove the noise, resulting in the fifth image. Based on the characteristics of rainwater and dirt, morphological operations, such as opening or closing operations, are used to remove rainwater and dirt interference from the image, resulting in the sixth image. An edge detection algorithm, such as the Canny operator, is used to extract edge information from the sixth image, resulting in the edge image. Based on the edge image and a pre-established tension clamp template, a template matching algorithm, such as a normalized cross-correlation algorithm, is used to locate the position of the tension clamp in the edge image, obtaining the matching result. The matching result is used to determine whether the tension clamp has defects. If the matching degree is lower than the preset threshold, the tension clamp is judged to have a defect and the defect information is output; if the matching degree is higher than the preset threshold, the tension clamp is judged to be in normal condition and normal information is output.

[0065] Further, a two-dimensional image of the tension clamp is obtained, which is essentially a photograph of the tension clamp taken with a camera. The original image may be in color or grayscale. For example, a color photograph of a tension clamp on an outdoor pole is the original image. This photograph contains background information such as the tension clamp itself, the pole, the conductor, and the sky. Converting the original image to grayscale simplifies the image data and reduces computation. Color images typically contain three color channels: red, green, and blue, while grayscale images have only one channel, where the value of each pixel represents the brightness of that pixel. For example, in the original color image, the tension clamp might be red, the pole gray, and the sky blue. After conversion to grayscale, these parts become different grayscale values. If the pixel values ​​of the grayscale image are unevenly distributed, such as the image being too dark or too bright overall, it will lead to a loss of detail. In this case, histogram equalization is needed. Histogram equalization can stretch the range of pixel values ​​in the image, making the contrast between light and dark areas clearer. Suppose that in the original grayscale image, most pixel values ​​are concentrated in the darker areas, making the details of the tension clamp unclear. After histogram equalization, the pixel values ​​will be redistributed, making the details of the tension clamp more apparent. Normalizing a grayscale image scales the pixel values ​​to a specific range, such as between 0 and 1. This facilitates subsequent image processing algorithms, such as those that require a specific range of input data. For example, if the pixel value range of the grayscale image after histogram equalization is 0 to 255, normalizing the image to between 0 and 1 will divide each pixel value by 255. Images may contain salt-and-pepper noise, which appears as randomly appearing black and white dots. For example, some pixels in the image may become pure black or pure white due to errors in the camera sensor or transmission process. This noise affects subsequent image processing and needs to be removed using median filtering. Median filtering replaces the pixel value with the median value of the surrounding pixels, effectively removing salt-and-pepper noise. Images may also contain Gaussian noise, which causes image blurring. For example, due to insufficient light or camera shake, each pixel value in an image is subject to random interference, causing the image to appear blurry. Gaussian filtering, which involves convolving the image with a Gaussian kernel, can effectively remove Gaussian noise. Due to outdoor environmental factors, tension clamp images may contain interference from rainwater and dirt. For example, rainwater can form droplets on the tension clamp surface, and dirt can form spots, all of which can affect subsequent defect detection. Morphological operations, such as opening operations, can remove small objects from the image, such as raindrops or dirt. Opening operations are erosion followed by dilation, which can remove objects smaller than the structuring element while preserving the overall shape of the image. Edge detection can extract the contour information of objects in an image. For example, the Canny operator can be used to detect the edges of tension clamps. The Canny operator first applies Gaussian filtering to the image, then calculates the image gradient, and finally obtains the edge image through non-maximum suppression and double threshold detection.Template matching can be used to locate the position of a specific object in an image. For example, a template image of a tension clamp can be pre-created, and then a normalized cross-correlation algorithm can be used to search for the position in the edge image that is most similar to the template image.

[0066] Step S102: Based on the first image, extract density-related features such as pixel value distribution and gray-level gradient features to construct a preliminary candidate set of defect regions, thus obtaining the first candidate set.

[0067] Step 1: Acquire the first image. Step 2: Calculate the pixel value distribution of the first image. For each pixel in the first image, count the frequency of its pixel value occurrence to obtain a pixel value distribution histogram. Step 3: Calculate the grayscale gradient features of the first image. Use the Sobel operator to calculate the horizontal and vertical gradients of each pixel in the first image to obtain a gradient image. If the gradient value is greater than a preset threshold, the pixel is considered an edge point. Step 4: Determine candidate pixels based on the pixel value distribution and grayscale gradient features. Combine the pixel value distribution and gradient image obtained in Steps 2 and 3. If the pixel gradient value is greater than a preset threshold and the pixel value is located in an abnormal interval of the pixel value distribution histogram, the pixel is determined to be a candidate pixel. Step 5: Construct connected regions based on candidate pixels.

[0068] The Sobel operator is an edge detection operator used in image processing. It is commonly used to calculate the gradient of an image and detect edges within it. The Sobel operator has two directions: horizontal (X-direction) and vertical (Y-direction), used to detect edge changes in an image.

[0069] Sobel operator (Gx) in the horizontal direction:

[0070]

[0071] Vertical Sobel operator (Gy):

[0072]

[0073] A depth-first search algorithm is used to connect adjacent candidate pixels into connected regions, resulting in multiple sets of connected regions. Step Six: Filter candidate defect regions based on the density features of connected regions. Calculate the pixel density of each connected region. If the pixel density of a connected region is greater than a preset density threshold, the region is determined to be a candidate defect region, forming the first candidate set. Step Seven: Obtain the first candidate set.

[0074] The process of identifying connected components using the depth-first search algorithm includes:

[0075] Initialization: Select a starting pixel as the starting point for the search. This point is usually part of a connected region.

[0076] Marked access: Create a matrix (or array) of the same size as the image to mark whether each pixel has been accessed, in order to avoid duplicate processing.

[0077] Recursive search: Starting from the starting point, recursively check its neighboring pixels (usually in the four directions of up, down, left, and right, or in the eight directions of up, down, left, right, and diagonal). If a neighboring pixel has not been visited and satisfies the connectivity condition (e.g., the pixel values ​​are the same or within a certain range of difference), it is marked as visited and added to the set of connected components.

[0078] Update region information: Each time a pixel is added to a connected region, some statistical information of that region can be updated, such as the total number of pixels in the region and the average pixel value.

[0079] Continue the search: Continue with steps 3 and 4 for newly added pixels in the connected region until no more neighboring pixels satisfy the condition.

[0080] Repeat: Repeat the above process for each unvisited pixel in the image until all pixels have been visited.

[0081] Further, acquiring the first image involves obtaining a grayscale image of the tension clamp. This can be obtained from an image captured by a camera, after grayscale processing. For example, if the pixel values ​​of a color image captured by a camera are RGB(100,150,200), converting it to grayscale values ​​can be done using the formula Gray = 0.299R + 0.587G + 0.114B, resulting in a grayscale value of 142.8. This grayscale image is the first image for subsequent processing. Calculating the pixel value distribution of the first image is to understand the frequency of different grayscale values ​​in the image. The number of times each grayscale value appears can be counted; for example, grayscale value 0 appears 100 times, grayscale value 1 appears 200 times, and so on, resulting in a pixel value distribution histogram. This histogram reflects the overall brightness and contrast information of the image. If the histogram is concentrated around a certain grayscale value, it indicates low image contrast. Histogram equalization can be used to improve image contrast and make image details clearer. Calculating the grayscale gradient features of the first image is to find the edge information of objects in the image. The Sobel operator can be used to calculate the gradient value of each pixel. The Sobel operator calculates the gradient through a convolution operation. For example, if a pixel's surrounding pixel values ​​are 100, 110, 120, 90, 100, 110, 80, 90, and 100, after the Sobel operator calculation, the horizontal gradient is 20 and the vertical gradient is 10. If the gradient value is greater than a preset threshold, such as 30, the pixel is considered an edge point. Extracting edge information helps in subsequent location of tension clamps. Pixel value distribution and gradient images are fused to determine candidate pixels. If a pixel has a large gradient value and its pixel value is in an abnormal range in the pixel value distribution histogram (e.g., very bright or very dark), then this pixel is considered a potential defect candidate. For example, if a pixel has a grayscale value of 250 (very bright) and its gradient value is greater than the threshold 30, while the overall image grayscale values ​​are concentrated around 100, then this pixel is likely a defect candidate. This effectively eliminates interference from background and normal areas, improving the accuracy of defect detection. Candidate pixels are often scattered, requiring the connection of adjacent candidate pixels into connected regions. This can be achieved using a depth-first search algorithm. Starting from a candidate pixel, search its eight neighboring pixels. If a neighboring pixel is also a candidate pixel, add it to the current connected region, and continue searching the surrounding pixels of this neighboring pixel until no new candidate pixels can be added. This results in a connected region. Repeating this process yields multiple sets of connected regions. Connecting scattered candidate pixels into connected regions helps in subsequent analysis of the shape and size of defects. Not all connected regions are defect regions; some may be noise or other interference. To filter out true defect candidate regions, the pixel density of each connected region can be calculated.Pixel density refers to the proportion of candidate pixels in a connected region to the total number of pixels in that region. A high pixel density in a connected region indicates a high density of candidate pixels, making it more likely to be a defective area. For example, if a connected region containing 100 pixels has 80 candidate pixels, its pixel density is 80%. If the density threshold is set to 50%, this connected region is a defective candidate and will be added to the first candidate set. This effectively removes noise and interference, improving the reliability of defect detection. The first candidate set contains all connected regions that meet the density condition; these regions are potential defective areas. Obtaining the first candidate set is a crucial step in defect detection. Further analysis and judgment can then be performed on these candidate regions to ultimately determine whether the tension clamp has a defect. For example, the shape, size, and location of the candidate regions can be analyzed, combined with prior knowledge of the tension clamp, to determine whether these regions are indeed defects.

[0082] Step S103: For each candidate region in the first candidate set, the edge information is extracted using an edge detection algorithm, and the morphological features such as the curvature and roughness of the edge are calculated to obtain a morphological feature vector.

[0083] Acquire candidate region images: Segment the candidate regions of interest from the input image data to generate the first image. Perform edge detection: Process the first image using the Canny edge detection algorithm to obtain the edge information of the candidate regions and generate the second image. If the edges detected by the Canny algorithm are discontinuous, morphological closing operations are used to connect the edges, and the second image is generated again. Calculate edge curvature: Calculate the curvature of each edge segment in the second image using a polynomial fitting method. If the number of edge pixels is less than a preset threshold, linear fitting is used to calculate the curvature value. If the number of edge pixels is greater than the preset threshold, quadratic polynomial fitting is used to calculate the curvature value. Calculate edge roughness: Calculate the edge roughness based on the fluctuation of the gray values ​​of edge pixels in the second image. Obtain the edge roughness value by calculating the standard deviation of the gray values ​​of edge pixels. Construct morphological feature vector: Combine the acquired edge curvature and edge roughness values ​​into a morphological feature vector. Arrange the curvature value and roughness value of each edge segment sequentially to form the feature vector.

[0084] Furthermore, the purpose of acquiring candidate region images is to separate the target region of interest from a complex background for more refined analysis later. For example, in an image of a circuit board containing electronic components, the goal is to extract a specific chip; this chip is the candidate region. There are many segmentation methods, such as color-based thresholding or shape-based contour extraction. The specific method chosen depends on the characteristics of the image and the actual requirements. After obtaining the first image, its features can be further analyzed to determine if it is the true target. Edge detection is used to obtain the contour information of the candidate region. The Canny edge detection algorithm is a commonly used edge detection algorithm that effectively suppresses noise and preserves true edge information. Suppose we are processing a circular candidate region image; after Canny edge detection, a clear circular contour should be obtained. If the detected edges are discontinuous, such as broken edges, morphological closing operations can be used to connect the edges to ensure the integrity of the contour. Closing operations are dilation followed by erosion, which can bridge small breaks and make the edges smoother. The second image processed in this way will have more complete edge information, facilitating subsequent analysis. The purpose of calculating edge curvature is to describe the degree of edge curvature. For each edge segment in the second image, its curvature can be calculated using polynomial fitting. If the number of edge pixels is small, such as only a few pixels, linear fitting can be used, which is equivalent to fitting the edge with a straight line, resulting in a curvature value of zero. If the number of edge pixels is large, such as dozens of pixels, quadratic polynomial fitting can be used, which can more accurately describe the curvature of the edge. For example, for a circular edge, its curvature value should be a constant, while for an elliptical edge, its curvature value will change along the edge. The purpose of calculating edge roughness is to describe the texture information of the edge. Edge roughness can be obtained by calculating the standard deviation of the gray values ​​of edge pixels. For example, a smooth edge has little variation in pixel gray values ​​and a small standard deviation, thus its roughness is low. A jagged edge, on the other hand, has large variation in pixel gray values ​​and a large standard deviation, thus its roughness is high. Edge roughness can be used to distinguish different types of defects, such as cracks and scratches. The purpose of constructing a morphological feature vector is to combine the edge curvature and edge roughness into a vector for subsequent classification. Suppose we calculate the curvature value of an edge segment to be 0.5 and the roughness value to be 0.2. These two values ​​can then be combined into a morphological feature vector [0.5, 0.2]. For multiple edge segments, the curvature and roughness values ​​of each segment can be arranged sequentially to form a longer feature vector. The purpose of training the classifier is to establish a mapping relationship between the morphological feature vector and the candidate region categories.

[0085] The expression for Canny edge detection is:

[0086] I smooth (x, y)=I(x, y)*G(x, y, σ);

[0087]

[0088] G(x,y) is the gradient magnitude, and I(x,y) is the original image.

[0089] Step S104: Construct a sample library containing various defect types such as early corrosion and wear, where each sample contains density feature and morphological feature vector. Train a support vector machine model based on the sample library to distinguish defects with similar density but different morphologies.

[0090] Acquire sample images containing various defect types, including early corrosion and wear. Extract density features from the acquired sample images, calculate the density value for each sample, and construct a density feature vector. Simultaneously, extract morphological features using methods such as Fourier descriptors and wavelet transforms to construct morphological feature vectors. Combine the density and morphological feature vectors into a complete sample feature vector and store it in a sample database. Obtain sample feature vectors from the sample database and input them into a support vector machine (SVM) model for training. During training, if the density feature vectors of two samples are similar, focus on training the differences between their morphological feature vectors. Through training, obtain an SVM model capable of distinguishing defects with similar densities but different morphologies. Acquire the test image, extract its density and morphological features, and construct its feature vector. Input the feature vector of the test image into the trained SVM model for prediction. The SVM model determines the defect type based on the feature vector of the test image. If the density features of the test image are similar to the density features of a certain type of defect, the model makes a more refined judgment based on morphological features, distinguishing defects with similar densities but different morphologies. Obtain the prediction results. If the prediction result indicates that the image to be tested belongs to a specific type of defect, then the defect type and its corresponding confidence level are output. Based on the defect type and confidence level, the severity of the defect is determined. If the confidence level is higher than a preset threshold, the defect type is determined; otherwise, the image to be tested is marked as "cannot be determined".

[0091] Furthermore, such as Figure 2-3As shown, corrosion and wear are common defect types in industrial production, affecting product quality and lifespan. To effectively detect and differentiate these defects, analysis combining density and morphological features is necessary. Density features reflect the degree of mass concentration in the defect area. For example, when corrosion occurs on a metal surface, the metal density in the corroded area decreases. The average grayscale value of the defect area can be calculated using image processing techniques and converted into a density value. Assuming the average grayscale value of a normal metal surface is 200, while the average grayscale value of the corroded area is 100, the density value of the corroded area can be set to 0.5. The density value of the wear area is calculated similarly. By calculating the density value of each sample, a density feature vector can be constructed. Morphological features reflect the geometric information of the defect, such as shape and texture. Fourier descriptors can be used to describe the contour shape of the defect. For example, the Fourier descriptor of a circular corrosion will exhibit different characteristics than that of a square wear. Wavelet transform can be used to analyze the texture information of the defect. For example, a rough wear area will exhibit high-frequency components after wavelet transform, while a smooth corrosion area will exhibit low-frequency components. Combining these morphological features allows the construction of a morphological feature vector. Combining density and morphological feature vectors into a complete sample feature vector and storing it in a sample database provides a data foundation for subsequent defect classification. For example, a sample's feature vector might contain a density value of 0.6, along with a series of Fourier descriptors and / or wavelet coefficients describing its shape and texture. Training the sample feature vectors using a support vector machine (SVM) model establishes a mapping between defect types and feature vectors. During training, if two samples have similar density feature vectors (e.g., their density values ​​are close to 0.5), the model will focus more on the differences between their morphological feature vectors. This allows the model to distinguish defects with similar densities but different shapes, such as differentiating between a circular corrosion and a square wear with the same density value. In actual detection, the density and morphological features of the image to be tested are first extracted to construct its feature vector. For example, if the density value of the image to be tested is 0.55, the Fourier descriptors and / or wavelet coefficients also need to be calculated. This feature vector is then input into the trained SVM model for prediction. The SVM model determines the defect type based on the feature vector of the image to be tested. If the density features of the image under test are similar to the density features of a certain type of defect, for example, both are close to 0.5, the model will make a more refined judgment based on morphological features. For example, if the morphological features of the image under test are closer to those of circular erosion, the model will classify it as circular erosion. The prediction result will provide the defect type and the corresponding confidence level. For example, the model may predict that the image under test is circular erosion with a confidence level of 0.9. If the confidence level is higher than a preset threshold, such as 0.8, the defect type is determined to be circular erosion. Otherwise, the image under test is marked as "undeterminable".The severity of a defect can be determined based on its type and confidence level. For example, if it is identified as circular corrosion with a high confidence level, its severity can be assessed based on indicators such as the area or depth of the corrosion. This effectively distinguishes different types of defects, even those with similar densities, by differentiating them through morphological characteristics, thereby improving the accuracy and reliability of defect detection.

[0092] Step S105: Use the trained support vector machine model to classify each candidate region in the first candidate set. If the morphological features of the candidate region are more similar to those of the corrosion sample, then mark it as a corrosion defect; if the morphological features of the candidate region are more similar to those of the wear sample, then mark it as a wear defect.

[0093] Obtain the first candidate set of image data and a pre-trained Support Vector Machine (SVM) model. The SVM model training data includes corrosion sample image data and wear sample image data, and training is complete. For each candidate region image in the first candidate set, extract its morphological features. The extracted morphological features include: perimeter, area, roundness, rectangularity, eccentricity, Hu invariant moments, etc. Convert the extracted morphological feature data into the input vector of the SVM model. Use the trained SVM model to classify and predict the transformed morphological feature vector. The SVM model outputs the probability value of the candidate region being corrosion or wear. Determine the category of the candidate region based on the probability value output by the SVM model. If the corrosion probability is greater than a preset threshold, the candidate region is marked as a corrosion defect. If the wear probability is greater than a preset threshold and the corrosion probability is less than the wear probability, the candidate region is marked as a wear defect. Use a clustering algorithm to cluster the candidate regions marked as corrosion defects.

[0094] Furthermore, such as Figure 4-5As shown, the first candidate set of image data is acquired, such as image data of the workpiece to be inspected obtained through non-destructive testing equipment (e.g., ultrasound, X-ray). These images may contain various defects such as corrosion and wear. A pre-trained support vector machine (SVM) model, for example, a model trained using a large amount of known corrosion and wear sample image data, can distinguish between corrosion and wear based on the input image features. Since SVM has advantages in handling high-dimensional features and non-linear classification problems, it is chosen as the defect classification model. Morphological features are extracted for each candidate region image in the first candidate set. For example, for a candidate region image, its perimeter is calculated to be 20 pixels, its area to be 150 pixels, its circularity to be 0.8, its rectangularity to be 0.6, its eccentricity to be 0.5, and the values ​​of seven Hu invariant moments. These morphological features describe the shape and geometric characteristics of the candidate region. Extracting morphological features can help distinguish different types of defects; for example, corrosion often presents as an irregular shape, while wear may present as a more regular shape. The extracted morphological feature data is then converted into the input vector of the SVM model. For example, the extracted feature values ​​such as perimeter, area, roundness, rectangularity, eccentricity, and Hu invariant moments are combined into a vector [20, 150, 0.8, 0.6, 0.5, ...]. This vector is the input to the support vector machine (SVM) model. The reason for converting it to vector form is that the input to the SVM model needs to be a fixed-dimensional vector. The trained SVM model is then used to classify and predict the transformed morphological feature vector. For example, inputting the feature vector [20, 150, 0.8, 0.6, 0.5, ...] into the trained SVM model results in an output corrosion probability of 0.9 and a wear probability of 0.1. These probability values ​​represent the likelihood that a candidate region belongs to corrosion or wear. Based on the probability values ​​output by the SVM model, the category of the candidate region is determined. The preset thresholds can be adjusted according to the actual situation. For example, the corrosion probability threshold can be set to 0.8 and the wear probability threshold to 0.7. In the above example, since the corrosion probability of 0.9 is greater than 0.8, the candidate region is marked as a corrosion defect. If the corrosion probability is 0.7 and the wear probability is 0.8, the candidate region is marked as a wear defect. Setting a threshold can control the accuracy and sensitivity of defect identification. A clustering algorithm is used to cluster the candidate regions marked as corrosion defects. For example, the K-means algorithm can be used to cluster the candidate regions marked as corrosion. The distance between the center points of the candidate regions is calculated. If the distance between the center points of two candidate regions is less than a preset threshold (e.g., 10 pixels), the two candidate regions are clustered into one cluster. Clustering can group scattered corrosion regions together, thereby better identifying and distinguishing different corrosion regions. For example, multiple small corrosion points can be clustered into a large corrosion region.Based on the clustering results, the number of candidate regions contained in each cluster is counted, the average morphological features of each cluster are obtained, and the area of ​​each cluster is calculated. For example, a cluster contains 10 candidate regions, with an average area of ​​200 pixels and an average circularity of 0.7.

[0095] Step S106: Based on information such as wire clip material and surface treatment process, perform specific density feature correction on wire clip images of different materials, such as compensating for the difference in reflectivity of different materials, to obtain a second image.

[0096] Furthermore, obtaining information on the material and surface treatment process of the wire clamp is to establish a reflectivity compensation model, thereby eliminating the influence of different materials and surface treatment processes on image density features.

[0097] The expression for the reflectivity compensation model is:

[0098] I(x, y) = k d ·L(x,y)·cos(θ)+k s ·L(x,y)·(cos(α)) n ;

[0099] Where: kd is the diffuse reflectance coefficient (usually representing the diffuse reflectivity of an object). ks is the specular reflectance coefficient (representing the specular reflectivity of an object's surface). θ is the angle between the incident light and the surface normal vector. α is the angle between the incident light and the reflected light. n is the gloss index, which controls the sharpness of specular reflection.

[0100] For example, if a wire clamp is made of copper and its surface treatment is nickel plating, then parameters such as the reflectivity of copper and the reflectivity of the nickel plating layer need to be obtained from the wire clamp database. These parameters will be used to subsequently build a reflectivity compensation model. Acquiring wire clamp images requires using an industrial camera. Choosing the appropriate camera model, lens, and light source is crucial for obtaining high-quality images. For example, an industrial camera equipped with a telecentric lens, combined with a uniform lighting environment, can be used to photograph wire clamps of different materials and surface treatments, and the images should be saved in standard formats such as TIFF or PNG. Preprocessing wire clamp images aims to reduce noise and enhance image contrast. Grayscale conversion transforms a color image into a grayscale image, simplifying subsequent image processing. Histogram equalization can enhance image contrast and make image details clearer. For example, if the pixel values ​​of a wire clamp image are concentrated within a certain grayscale range, resulting in an overall dark image, histogram equalization can redistribute the pixel values, making the image brightness more uniform and highlighting details. If the texture information of the wire clamp image is unclear, histogram equalization is performed again to further enhance the texture details of the image for subsequent density feature extraction. The purpose of extracting density features is to quantify the texture information of wire clip images. The gray-level co-occurrence matrix (GLCM) algorithm is a commonly used texture feature extraction method. It describes the texture characteristics of an image by calculating the spatial relationships between different gray-level pixels. For example, the roughness and texture directionality of the wire clip surface can be described by calculating the energy, contrast, and correlation of the GLCM. Clear density features are crucial for accurately establishing a reflectivity compensation model. The purpose of establishing a reflectivity compensation model is to eliminate the influence of different materials and surface treatment processes on the image density features. Because different materials and surface treatment processes have different reflectivities, even the same texture will exhibit different density features in the image. Therefore, a model is needed to compensate for this difference. Multiple linear regression is a commonly used regression analysis method that can be used to establish the mapping relationship between material reflectivity and density features. For example, the reflectivity of the wire clip material, surface treatment parameters, and extracted density features can be used as input variables to establish a multiple linear regression model to predict the corrected density features. The purpose of density feature correction is to eliminate the influence of material and surface treatment processes on the image density features. Based on the established reflectivity compensation model, the density features of the wire clip image are corrected to obtain density features independent of material and surface treatment process. For example, for a copper wire clip plated with nickel, the corrected density features can be calculated using the reflectivity compensation model based on its reflectivity and surface treatment parameters, thus eliminating the influence of material and surface treatment process on the density features. The purpose of generating a second image is to visualize the corrected density features. By performing an inverse grayscale transformation, the corrected density features are converted into pixel values, allowing the wire clip image to be reconstructed.

[0101] The expression for the gray-level co-occurrence matrix is:

[0102]

[0103] P(i,j) is the gray-level co-occurrence matrix, count(·) is a counting operation that represents the number of pixel pairs that meet the conditions, and I is the number of pixels.

[0104] In step S107, the pixel value distribution and grayscale gradient features of the second image are extracted again, and combined with the morphological feature vector obtained in step 3, the support vector machine model is used again for classification to improve the accuracy of defect classification.

[0105] Obtain the pixel value distribution of the second image: Calculate the frequency of each pixel value in the second image, construct a pixel value distribution histogram, and obtain the pixel value distribution features. Obtain the gray-level gradient features of the second image: Use the Sobel operator to calculate the gradient in the horizontal and vertical directions of the second image, obtain the gradient magnitude and direction information, and constitute the gray-level gradient features. Fuse the feature vector: Concatenate the pixel value distribution features obtained in step 1, the gray-level gradient features obtained in step 2, and the existing morphological feature vector in step 3 to form a new fused feature vector. Split the dataset: Divide the fused feature vector and its corresponding defect category label into a training set and a test set. Train the support vector machine model: Train the support vector machine model using the training set data and optimize the model parameters. If the number of samples in the training set is small, use cross-validation to evaluate the model performance and select the optimal parameters. If the training set samples are imbalanced, use oversampling or undersampling techniques to balance the sample categories. Test the support vector machine model: Use the test set data to evaluate the classification performance of the trained support vector machine model, and calculate metrics such as classification accuracy, precision, and recall. Output classification results: Use the trained support vector machine model to classify the new defect image and obtain the defect category judgment result.

[0106] Furthermore, obtaining the pixel value distribution of the second image is to further extract image features and provide richer information for subsequent defect classification. For example, in a second image of a wire clip, the pixel values ​​are mainly concentrated between grayscale values ​​of 100 and 150, indicating that the overall brightness of the image is moderate. If the pixel value distribution is relatively uniform, it indicates that the image contrast is good. Constructing a pixel value distribution histogram can intuitively show the distribution of pixel values. For example, we can count 1000 pixels with a grayscale value of 0, 1200 pixels with a grayscale value of 1, and so on, and draw a histogram to obtain the pixel value distribution features. Obtaining the grayscale gradient features of the second image is to capture the edge information of objects in the image, as edges are usually the locations where defects occur. For example, the Sobel operator is used to calculate the horizontal and vertical gradients of the image. Assuming that the grayscale value of a certain pixel is 100, the grayscale value of the pixel to its right is 120, and the grayscale value of the pixel below it is 90, then the horizontal gradient is 20, and the vertical gradient is -10. The gradient magnitude is √(20). 2 +(-10) 2The gradient direction can be calculated using arctan(-10 / 20) ≈ 22.36. These gradients constitute the grayscale gradient features, which can be used to describe the texture changes and edge strength of an image. Fusing feature vectors combines image features extracted from different angles to form a more comprehensive feature description. For example, pixel value distribution features, grayscale gradient features, and morphological features extracted in previous steps (e.g., area, perimeter, roundness, etc.) are concatenated into a new feature vector. Assuming the pixel value distribution feature is a 10-dimensional vector, the grayscale gradient feature is a 5-dimensional vector, and the morphological feature is a 3-dimensional vector, then the fused feature vector is an 18-dimensional vector containing richer image information. Splitting the dataset is for evaluating the performance of the classification model. Existing data is divided into training and test sets. For example, 80% of the data is used as the training set to train the model, and the remaining 20% ​​is used as the test set to evaluate the model's generalization ability. This avoids overfitting and ensures that the model performs well on new data. The process of training a support vector machine model involves using the training set data to adjust the model parameters so that the model can distinguish between different categories of defects. If the training set has a small number of samples, such as only 100 images, cross-validation can be used. The training set is divided into several parts, with one part used as the validation set and the rest as the training set, thus making more efficient use of the limited data. If the training set has an imbalanced class distribution, for example, 90 samples of class A defects and only 10 samples of class B defects, oversampling can be used to duplicate or synthesize class B defect samples, or undersampling can be used to reduce the number of class A defect samples, making the class distribution more balanced and preventing the model from favoring the class with the larger number of samples. Testing the support vector machine model is to evaluate its generalization ability. The trained model is tested using test set data, and metrics such as classification accuracy, precision, and recall are calculated. For example, if the test set has 100 images and the model correctly classifies 90, the classification accuracy is 90%. The output classification result is obtained by applying the trained support vector machine model to new defect images to determine the defect category. For example, given a new image of a wire clip, after feature extraction and model classification, if the model outputs "Class A defect," it indicates that the wire clip has a class A defect.

[0107] Example 2

[0108] like Figure 6 As shown, this embodiment provides an intelligent classification method for defects in transmission line tension clamps based on two-dimensional images, including:

[0109] The first image processing module is used to acquire a two-dimensional image of the tension clamp, and to preprocess the two-dimensional image of the tension clamp to obtain a first image;

[0110] The first candidate processing module is used to extract relevant features from the first image and construct a first candidate set of defect regions based on the relevant features.

[0111] The morphological feature vector construction module is used to extract edge information of the first candidate set through an edge detection algorithm, and calculate the curvature and roughness of the edges to obtain morphological feature vectors.

[0112] The model building module is used to acquire a defect sample set, train the support vector machine model based on the defect sample set, and obtain the trained support vector machine model.

[0113] The first defect classification result generation module is used to input the first candidate set into the trained support vector machine model for calculation to obtain the first defect classification result.

[0114] The second candidate processing module is used to perform density feature correction on wire clip images of different materials to obtain a second image, extract the pixel value distribution and gray-level gradient features of the second image, and obtain a second candidate set.

[0115] The second image processing module is used to input the second candidate set and the morphological feature vector into the trained support vector machine model for calculation to obtain the second defect classification result.

[0116] The fusion module is used to integrate the first defect classification result and the second defect classification result to generate the final classification result.

[0117] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A two-dimensional image-based power transmission line strain clamp defect intelligent classification method, characterized in that, include: A two-dimensional image of the tension clamp is acquired, and the two-dimensional image of the tension clamp is preprocessed to obtain a first image; Extract relevant features from the first image, and construct a first candidate set of defect regions based on the relevant features; Edge information of the first candidate set is extracted using an edge detection algorithm, and the curvature and roughness of the edges are calculated to obtain morphological feature vectors. Obtain a defect sample set, and train the support vector machine model based on the defect sample set to obtain the trained support vector machine model; The first candidate set is input into the trained support vector machine model for calculation to obtain the first defect classification result; Density feature correction is performed on wire clip images of different materials to obtain a second image. The pixel value distribution and gray-level gradient features of the second image are extracted to obtain a second candidate set. The second candidate set and the morphological feature vector are input into the trained support vector machine model for calculation to obtain the second defect classification result; The first defect classification result and the second defect classification result are integrated to generate the final classification result; The process of obtaining the trained support vector machine model includes: Construct a sample library containing early corrosion defect types and wear defect types, where each sample contains density feature and morphological feature vector; Sample feature vectors are obtained from the sample library and input into the support vector machine model for training. During the training process, if the density feature vectors of two samples are similar, the difference between the morphological feature vectors is trained to obtain the trained support vector machine model.

2. The method of claim 1, wherein, The process of obtaining the first image includes: acquiring a two-dimensional image of the tension clamp to obtain an original image, and performing image grayscale conversion, normalization, and denoising operations on the original image to obtain the first image.

3. The method of claim 1, wherein, The process of constructing a first candidate set of defect regions based on the aforementioned relevant features includes: For each pixel in the first image, the frequency of its pixel value is counted to obtain a pixel value distribution histogram. The Sobel operator is used to calculate the horizontal and vertical gradients of each pixel in the first image to obtain a gradient image. The pixel value distribution histogram and the gradient image are fused. If the pixel gradient value is greater than a preset threshold and the pixel value is located in the abnormal range of the pixel value distribution histogram, then the pixel is determined to be a candidate pixel. A depth-first search algorithm is used to connect adjacent candidate pixels into connected regions, resulting in a set of multiple connected regions. Calculate the pixel density of each connected region. If the pixel density of a connected region is greater than a preset density threshold, then the region is determined to be a defect candidate region and constitutes the first candidate set.

4. The method of claim 1, wherein, The process of obtaining the morphological feature vector includes: Calculate the fluctuation of grayscale values ​​of edge pixels in the first candidate set, calculate edge roughness, and calculate curvature to obtain curvature values; The obtained edge curvature and edge roughness values ​​are combined into a morphological feature vector.

5. The method of claim 1, wherein, The process of obtaining the first defect classification result includes: The first candidate set is input into the trained support vector machine model for calculation, and the candidate region is the probability value of corrosion or wear; Based on the probability value, the category of the candidate region is determined. If the corrosion probability is greater than a preset threshold, the candidate region is marked as a corrosion defect. If the wear probability is greater than a preset threshold and the corrosion probability is less than the wear probability, the candidate region is marked as a wear defect, and the first defect classification result is generated.

6. The method of claim 1, wherein, The process of obtaining the second candidate set includes: Retrieve the material properties and surface treatment parameters for each type of wire clamp from the wire clamp database; A dataset of wire clamp images was obtained by photographing wire clamps made of different materials and with different surface treatments using an industrial camera. The wire clip image is preprocessed using grayscale conversion and histogram equalization to obtain the preprocessed wire clip image. The density features of the preprocessed wire clip image are extracted using the gray-level co-occurrence matrix algorithm to obtain the texture information of the wire clip image; Based on the material and surface treatment process information of the wire clamp, a mapping relationship model between the material reflectivity and density characteristics is established. Based on the reflectivity compensation model, the density characteristics of the wire clip image are corrected to compensate for the reflectivity differences caused by the material and surface treatment process, and the corrected density characteristics are obtained. Based on the corrected density features, the wire clip image is reconstructed to obtain the second candidate set.

7. The method of claim 1, wherein, The process of obtaining the second defect classification result includes: Obtain the pixel value distribution of the second image: Calculate the occurrence frequency of each pixel value in the second image, construct a pixel value distribution histogram, and obtain the pixel value distribution features; Obtain the grayscale gradient features of the second image: use the Sobel operator to calculate the gradients in the horizontal and vertical directions of the second image, obtain the gradient magnitude and direction information, and form the grayscale gradient features; obtain the second defect classification result based on the pixel value distribution features and the grayscale gradient features.

8. A two-dimensional image-based power transmission line strain clamp defect intelligent classification system, characterized in that, include: The first image processing module is used to acquire a two-dimensional image of the tension clamp, and to preprocess the two-dimensional image of the tension clamp to obtain a first image; The first candidate processing module is used to extract relevant features from the first image and construct a first candidate set of defect regions based on the relevant features. The morphological feature vector construction module is used to extract edge information of the first candidate set through an edge detection algorithm, and calculate the curvature and roughness of the edges to obtain morphological feature vectors. The model building module is used to acquire a defect sample set, train a support vector machine model based on the defect sample set, and obtain a trained support vector machine model. The process includes: Construct a sample library containing early corrosion defect types and wear defect types, where each sample contains density feature and morphological feature vector; Sample feature vectors are obtained from the sample database and input into the support vector machine model for training. During the training process, if the density feature vectors of two samples are similar, the difference between the morphological feature vectors is trained to obtain the trained support vector machine model. The first defect classification result generation module is used to input the first candidate set into the trained support vector machine model for calculation to obtain the first defect classification result. The second candidate processing module is configured to perform density feature correction on the wire clamp images of different materials to obtain second images, extract pixel value distribution and gray gradient features of the second images, and obtain a second candidate set; The second image processing module is configured to input the second candidate set and the shape feature vector into the trained support vector machine model to perform calculation and obtain a second defect classification result; The fusion module is configured to integrate the first defect classification result and the second defect classification result to generate a final classification result.

Images