Transmission Line Inspection Method Based on Image Processing

Abstract

The present invention relates to the technical field of image data processing, and in particular to a transmission line inspection method based on image processing, comprising: photographing images of conductors of each section of a transmission line, extracting the conductor area in the image, and clustering the pixels in the conductor area of ​​any section using an improved mean shift algorithm, obtaining multiple clusters by the improved mean shift algorithm, and judging whether there is a defect based on the ratio of the number of pixels in each cluster to the total number of pixels in the conductor area; when the ratio exceeds a set threshold, it is judged that there is a defect in the conductor area of ​​the section corresponding to the pixel in the cluster; wherein the improved mean shift algorithm includes adjusting the bandwidth to better adapt to the complexity and diversity of the image. The present invention solves the problem of low accuracy in the process of detecting defects in transmission lines.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and more specifically, to a power transmission line inspection method based on image processing. Background Art

[0002] With the continuous growth of global electricity demand and the gradual aging of power facilities, the safe operation and maintenance management of transmission lines are becoming increasingly important. Traditional transmission line inspection methods mainly rely on manual inspections or regular inspections. Although some obvious defects and problems can be found, this method is not only inefficient, but also easily affected by human factors, with great safety hazards and the risk of false detection and missed detection. In addition, manual inspections are difficult to cover large areas and high-risk areas, and the detection frequency cannot be monitored in real time. Therefore, traditional inspection methods have been unable to meet the needs of modern transmission line safety management.

[0003] With the rapid development of drones, lidar, infrared thermal imaging and high-resolution visible light imaging technologies, the power line inspection method based on image processing has gradually become an efficient solution. It can quickly collect data under different environmental conditions through a variety of sensor devices, automatically identify and locate defects in power lines, and thus significantly improve the inspection efficiency and accuracy. However, the existing technology still faces the challenges of noise interference and external environmental factors during image acquisition, which may affect the image quality and reduce the accuracy of defect detection, especially in infrared thermal imaging and visible light images, where noise may mask real defects. Therefore, how to effectively suppress noise and improve the robustness of image processing algorithms has become the key to improving the performance of power line inspection methods based on image processing.

[0004] The patent application document with application publication number CN111276898A discloses a method for fine inspection of power transmission lines. The patent application document realizes defect detection and information query of power transmission lines by taking images of power transmission lines and using efficient data indexing and custom rendering settings.

[0005] However, although the above technical solution realizes the defect detection of the power transmission line to a certain extent, it does not take into account the noise that may be introduced during the image capture process, thereby resulting in the problem of low accuracy in the process of power transmission line defect detection. Summary of the invention

[0006] In order to solve the problem of low accuracy in the process of power transmission line defect detection raised in the above background technology, the present invention provides the following solution.

[0007] In this solution, the present invention provides a transmission line inspection method based on image processing, comprising:

[0008] The image of each section of the transmission line conductor is obtained, and the conductor area in the image is extracted; the pixel points in the conductor area of ​​any section image are clustered by using the improved mean shift algorithm to obtain multiple clusters, and if the ratio of the number of pixel points in each cluster cluster to the total number of pixel points in the conductor area is greater than a set ratio threshold, it is determined that there is a defect in the conductor area of ​​the section corresponding to the pixel point in the cluster; wherein the improved mean shift algorithm includes a bandwidth, and the bandwidth is the ratio of the initial value to the second attention degree of the corresponding pixel point; the first Section image The second attention of pixels , , where is an exponential function with the natural constant e as base, For the first The left side of the neighborhood with the image as the center is the first image in each image. The average value of the first attention of pixels, Set the right side of the neighborhood range in each image The average value of the first attention of pixels, , Set the first The maximum and minimum differences of the first attention of pixels, The first The similarity between the sequences composed of the first attention degrees of the pixels; the first attention degree represents the abnormality degree of each pixel.

[0009] The above technical solution uses an improved mean shift algorithm to accurately cluster the conductor areas in the transmission line image, which can effectively identify potential defects in the conductor area. It not only improves the accuracy of defect detection, but also can achieve more stable and efficient defect identification in different environments, enhances adaptability to complex scenes, and ensures that real-time monitoring and maintenance of transmission lines are more reliable.

[0010] Furthermore, calculate the Section image The first attention of pixels , , where is the normalization function, For the Section image The reference degree of each pixel, For the first Section image The gray value variance of each pixel in the first set window centered at a pixel point, For the first Section image The gray value variance of each pixel in the second set window centered at a pixel point, For the Section image The grayscale value of each pixel point, and the reference degree represents the similarity of the pixel points.

[0011] The above technical solution calculates the first attention degree of each pixel by comprehensively considering the reference degree of the pixel, the gray value variance and the gray value size in the neighborhood, so as to accurately evaluate the significant area in the image. Through normalization processing, it ensures that the features between different images and pixels can be effectively compared at the same scale, eliminating the influence of light and noise in the image. The reference degree reflects the similarity of the pixels and can effectively identify pixels that are similar or abnormal to the surrounding areas, thereby improving the sensitivity of image analysis. In this way, the key features in the image are captured more accurately and the detection ability of abnormal areas is improved.

[0012] Furthermore, calculate the Section image The first attention of pixels , , where is the normalization function, For the Section image The reference degree of each pixel, For the first Section image The gray value variance of each pixel in the first set window centered at a pixel point, For the first Section image The gray value variance of each pixel in the second set window centered at a pixel point, For the Section image The gray value of each pixel is For the Section image The texture features of the pixels are represented by the reference degree, and the reference degree represents the similarity of the pixels.

[0013] The above technical solution calculates the first attention of each pixel by comprehensively considering the reference degree of the pixel, the grayscale variance in the neighborhood, the grayscale value size and the texture characteristics, so as to effectively evaluate its significance. Through multi-dimensional attention calculation, it not only improves the detection accuracy of abnormal areas, but also enhances the robustness of the overall image processing, ensuring efficient and reliable image analysis capabilities.

[0014] Furthermore, calculate the Section image The reference degree of each pixel , , where is an exponential function with the natural constant e as base, For the Section image The angle between the pixel and the camera, For the Section image The Euclidean distance between the pixel and the camera, For the The actual length of the wire in the node image.

[0015] The above technical solution calculates the reference degree of the pixel points, combines the shooting angle and Euclidean distance factors, and uses an exponential function to weight these parameters, thereby effectively evaluating the relative importance of each pixel point in the image. By considering the angle and distance between the shooting device and the pixel point, it can better adapt to different shooting conditions, reduce the impact caused by angle deviation or distance change, and ensure the accuracy and robustness of image analysis. In addition, the introduction of the actual length of the wire as a standard parameter enables the solution to adapt to images of different sizes, further improving the detection capability of key areas in the image, and providing a more reliable basis for the accurate positioning and subsequent processing of abnormal areas.

[0016] Further, extracting the wire area in the image is specifically as follows: calculating the first The probability that an edge pixel in the image belongs to the wire area , , where is an exponential function with the natural constant e as base, , , , , is an empirical constant, and , For the The gray value of the edge pixel, is the mode of the grayscale values ​​of each edge pixel in the image, For the The gradient direction of edge pixels, is the maximum value of the gradient direction of each edge pixel in the image, For the The variance of the gradient direction of each edge pixel in the edge where the edge pixel is located, is a normalized function; if the probability If it is greater than the set probability threshold, then the edge pixels belong to the wire area.

[0017] The above technical solution calculates the probability that each pixel belongs to the wire area by comprehensively considering the gray value, gradient direction and its changes of the edge pixels, thereby achieving more accurate wire area extraction. The introduction of exponential functions to weight multiple factors can effectively deal with noise and interference in the image, and the influence of each parameter is adjusted by normalization to ensure robustness under different image conditions. By setting the judgment method of probability threshold, it is possible to accurately distinguish between wire areas and non-wire areas, thereby improving the accuracy of wire detection in image processing, especially in complex backgrounds or when the lighting changes greatly, the wire area can still be stably identified, ensuring the reliability of subsequent analysis.

[0018] Furthermore, the wire area in the image is extracted using the Otsu algorithm.

[0019] Furthermore, the method also includes performing denoising and grayscale processing on the images of each section of the conductor of the transmission line.

[0020] Furthermore, the similarity is calculated using a dynamic time warping algorithm.

[0021] Furthermore, the set ratio threshold is 0.8.

[0022] The beneficial effects of the present invention are:

[0023] The present invention can efficiently and accurately identify the conductor area in the transmission line image and effectively detect potential defects by combining image processing, improved mean shift algorithm and multiple image feature extraction methods. By calculating the first and second attention levels of the pixel points, as well as multi-dimensional features such as the reference degree, the accurate positioning of the abnormal area in the image is ensured. The introduction of bandwidth adjustment and probability calculation not only improves the reliability of the clustering results, but also can adaptively adjust to different image backgrounds, increasing the robustness of the method. In addition, the use of advanced technologies such as dynamic time warping algorithm and Otsu algorithm further enhances the accuracy of conductor area extraction, can stably identify and detect defects in complex environments, improves the automation and accuracy of transmission line inspection, and reduces the workload and errors of manual inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0025] Figure 1is a flow chart schematically illustrating a method for inspecting a power transmission line based on image processing according to an embodiment of the present invention;

[0026] Figure 2 FIG. 1 is a grayscale diagram schematically showing a real scene of a power transmission line according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0028] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0029] An embodiment of a power transmission line inspection method based on image processing.

[0030] like Figure 1 As shown, the flow chart of the power transmission line inspection method based on image processing of an embodiment of the present invention comprises the following steps:

[0031] S1: taking images of conductors of each section of a transmission line, and extracting conductor regions in the images.

[0032] In one embodiment, drones can be used to collect high-precision images of each section of the power transmission line. With its mobility and high-resolution camera equipment, drones can capture high-quality images of each section of the power transmission line, especially for high-altitude, remote or dangerous areas that are difficult to reach with traditional manual inspections. The use of drones can significantly improve the coverage and reliability of image acquisition. The images taken by drones can not only be transmitted in real time, but also provide detailed high-definition images to ensure that the status of the power transmission line at all angles is fully captured.

[0033] The acquired images first need to be denoised to remove environmental interference and other non-target noise. Transmission lines are usually in complex environments, where factors such as wind, rain, sunlight, and electromagnetic interference may affect image quality. Therefore, by applying advanced denoising algorithms such as median filtering, bilateral filtering, or convolutional neural network denoising, these noises can be effectively removed, thereby improving image clarity and detail retention. This step not only eliminates random noise in the image, but also ensures that the target features are more prominent in the subsequent image processing stage, further improving the accuracy of defect detection.

[0034] After denoising, the image will be grayscaled. Grayscale not only simplifies the color dimension of the image, but also reduces the computational complexity, making subsequent image processing, feature extraction, and defect analysis more efficient. In the grayscale process, by converting to a grayscale image, the edges, textures, and other important features of the object can be better highlighted. After grayscale, although the color information of the image is lost, most of the important information related to the health status of the transmission line is still retained, and the image processing speed is greatly improved.

[0035] like Figure 2 As shown, a real grayscale image of a power transmission line according to an embodiment of the present invention.

[0036] The wire area in the image is extracted using the Otsu algorithm. The wire area in the image extracted using the Otsu algorithm can automatically and accurately separate the wire from the complex background. The Otsu algorithm automatically determines the optimal threshold by maximizing the inter-class variance, and can efficiently adapt to different lighting conditions and background complexity without manual intervention. This method not only improves the accuracy and robustness of image segmentation, but also reduces the workload of manual participation, providing a clear and reliable foundation for subsequent defect detection and analysis, and greatly improving the overall detection efficiency and accuracy.

[0037] In another embodiment, extracting the wire area in the image is specifically:

[0038] Calculate the The probability that an edge pixel in the image belongs to the wire area , , where is an exponential function with the natural constant e as base, , , , , is an empirical constant, and , For the The gray value of the edge pixel, is the mode of the grayscale values ​​of each edge pixel in the image, For the The gradient direction of edge pixels, is the maximum value of the gradient direction of each edge pixel in the image, For the The variance of the gradient direction of each edge pixel in the edge where the edge pixel is located, is the normalization function;

[0039] If the probability If it is greater than the set probability threshold, then the The edge pixel points belong to the wire area. The probability threshold may be 0.7, and of course, it may be determined according to actual conditions.

[0040] By integrating features such as grayscale value, gradient direction, and gradient direction variance, the probability that the edge pixel belongs to the wire area is calculated, and the wire area is automatically identified by setting a probability threshold. This method uses normalization and exponential functions to effectively balance the impact of different features on the judgment, improving the algorithm's adaptability to complex backgrounds and variable lighting conditions. By introducing a probabilistic model, the possibility of misjudgment and missed detection is reduced, the accuracy and robustness of wire area extraction are improved, and a more accurate basis is provided for subsequent defect detection.

[0041] S2: Clustering the pixels in the wire area using an improved mean shift algorithm to obtain multiple clusters, where the improved mean shift algorithm includes bandwidth.

[0042] In one embodiment, the bandwidth is the ratio of the initial value to the second attention degree of the corresponding pixel point; Section image The second attention of pixels , , where is an exponential function with the natural constant e as base, For the first The left side of the neighborhood with the image as the center is the first image in each image. The average value of the first attention of pixels, Set the right side of the neighborhood range in each image The average value of the first attention of pixels, , Set the first The maximum and minimum differences of the first attention of pixels, The first The similarity between the sequences composed of the first attention degrees of the pixels can be calculated using a dynamic time warping algorithm.

[0043] The above scheme can more accurately evaluate the significance and abnormality of pixel points. The dynamic time warping algorithm is introduced to calculate the similarity, thereby improving the ability to detect feature changes in the image. Through exponential function smoothing, the scheme can effectively reduce noise interference when processing complex images and improve the accuracy and robustness of anomaly detection.

[0044] Calculate the Section image The first attention of pixels , , where is the normalization function, For the Section image The reference degree of each pixel, For the first Section image The gray value variance of each pixel in the first set window centered at a pixel point, For the first Section image The gray value variance of each pixel in the second set window centered at a pixel point, For the Section image The gray value of each pixel.

[0045] The first attention degree of each pixel is calculated by comprehensively considering the reference degree of the pixel, the variance of the grayscale value in the neighborhood, and the grayscale value size. Through normalization processing, the features of different images and pixels can be compared at the same scale, effectively eliminating the influence of illumination differences and noise in the image itself. The combination of grayscale variance and grayscale value size not only enhances the sensitivity to local changes in the image, but also improves the ability to identify abnormal areas. This method can more accurately reflect the key features in the image and improve the accuracy and robustness of overall image analysis and anomaly detection.

[0046] In another embodiment, calculating the Section image The first attention of pixels , , where is the normalization function, For the Section image The reference degree of each pixel, For the first Section image The gray value variance of each pixel in the first set window centered at a pixel point, For the first Section image The gray value variance of each pixel in the second set window centered at a pixel point, For the Section image The gray value of each pixel is For the Section image The texture features of pixels.

[0047] The first attention degree of each pixel is calculated by comprehensively considering the reference degree of the pixel, the neighborhood grayscale variance, the grayscale value size and the texture features. Through normalization processing, it ensures that the features of different images and pixels can be effectively compared at the same scale, eliminating the interference of light and noise. The introduction of texture features further enhances the sensitivity to local details of the image and improves the recognition ability of complex image areas. This method can more comprehensively reflect the significance of pixels, provide more accurate image analysis and anomaly detection, and thus improve the accuracy and robustness of image processing.

[0048] Calculate the Section image The reference degree of each pixel , , where is an exponential function with the natural constant e as base, For the Section image The angle between the pixel and the camera, For the Section image The Euclidean distance between the pixel and the camera, For the The actual length of the wire in the node image.

[0049] By calculating the reference degree of the pixel point, the angle and Euclidean distance between the shooting device and the pixel point, as well as the actual length of the wire are comprehensively considered. Through the processing of the exponential function, the angle deviation and distance factors are effectively mapped to a reasonable reference value range, thereby optimizing the weight of each pixel in the image. This method can effectively reduce the impact of shooting angle and distance on image quality, ensure high-precision extraction and anomaly detection of the wire area, and improve the accuracy and robustness of image analysis, especially when processing images with different shooting angles and distances, it can better adapt to changing shooting conditions.

[0050] S3: If the ratio of the number of pixels in each cluster to the total number of pixels in the wire area is greater than a set ratio threshold, it is determined that the wire area corresponding to the pixel in the cluster has defects.

[0051] The above-mentioned ratio threshold value may be set to 0.8, and of course, it may also be determined according to actual conditions.

[0052] In one embodiment, if the ratio of the number of pixels in a cluster to the total number of pixels in the conductor area exceeds a preset ratio threshold, it can be determined that the conductor area represented by the cluster has defects. The core of this judgment is that the pixel distribution in the conductor area should have a certain uniformity and regularity. If the number of pixels in a cluster is much higher than that in other clusters and exceeds the set threshold, it means that there is an abnormal distribution in the area, which is usually related to physical defects of the conductor, such as cracks, corrosion, and shedding. Through this method, the abnormal area can be accurately located to avoid misjudgment caused by relying solely on local features. In addition, this method makes full use of the globality and locality of clustering information in the image, can effectively identify potential conductor defects, improve the accuracy and reliability of detection, and avoid misjudgment that may be caused by relying solely on a single feature or visual judgment.

[0053] The scheme of the present invention improves the accuracy and efficiency of power transmission line inspection through a multiple analysis method based on image processing. By extracting the conductor area in the image and applying the improved mean shift algorithm to perform pixel clustering, the defects in the conductor area can be effectively identified. At the same time, by calculating the first and second attention levels, the reference level, and the probability of belonging to edge pixels of the pixels, the accuracy of defect detection is further improved. The dynamic time warping algorithm and the Otsu algorithm are used to calculate the similarity and set the threshold, ensuring stability and robustness under different environmental conditions. In addition, by denoising and graying the image, the influence of external interference factors on the detection results is reduced, thereby making the entire power transmission line inspection process more intelligent and automated, and being able to detect potential defects in a timely manner and improve inspection efficiency and accuracy.

[0054] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0055] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A power transmission line inspection method based on image processing, characterized in that: include: Acquire images of each section of the conductor of the transmission line, and extract the conductor area in the image of each section of the conductor; The bandwidth of the improved mean shift algorithm is set to be the ratio of the initial value to the second attention degree of the corresponding pixel point, so as to cluster the pixels in the wire area of ​​the image of any wire section to obtain a plurality of clusters; If the ratio of the number of pixels in the cluster to the total number of pixels in the wire area is greater than the set ratio threshold, it is determined that the wire area of ​​the node corresponding to the pixel in the cluster has a defect; Calculate the In the image of the section wire The second attention of pixels , , In the formula, is the natural exponential function, For the first The image of the node wire is the center of the neighborhood range set to the left. The average value of the first attention of pixels, For the first The image of the node wire is the center of the neighborhood range set on the right side. The average value of the first attention of pixels, , The first The maximum and minimum differences of the first attention of pixels, The first image in each image within the neighborhood range set on the left and the neighborhood range set on the right The similarity between the sequences composed of the first attention degrees of pixels; The first attention level represents the abnormality degree of each pixel point, and the expression is: , In the formula, is the normalization function, For the first In the image of the section wire The gray value variance of each pixel in the first set window centered at a pixel point, For the first In the image of the section wire The gray value variance of each pixel in the second set window centered at a pixel point, For the In the image of the section wire The gray value of each pixel; in, For the In the image of the section wire The degree of reference of each pixel represents the similarity of the pixels. , In the formula, For the In the image of the section wire The angle between the pixel and the camera, For the In the image of the section wire The Euclidean distance between the pixel and the camera, For the The actual length of the wire in the image of the node wire.

2. The power transmission line inspection method based on image processing according to claim 1, characterized in that: Calculate the In the image of the section wire The first attention of pixels The methods also include: , where is the normalization function, For the In the image of the section wire The reference degree of each pixel, For the first In the image of the section wire The gray value variance of each pixel in the first set window centered at a pixel point, For the first In the image of the section wire The gray value variance of each pixel in the second set window centered at a pixel point, For the In the image of the section wire The gray value of each pixel is For the In the image of the section wire The texture features of the pixels are represented by the reference degree, and the reference degree represents the similarity of the pixels.

3. The power transmission line inspection method based on image processing according to claim 1, characterized in that: Extracting the wire area in the image of each section of the wire, specifically: Calculate the The probability that an edge pixel belongs to the wire area in the wire image , , where is an exponential function with the natural constant e as base, , , , , is an empirical constant, and , For the The gray value of the edge pixel, is the mode of the grayscale values ​​of each edge pixel in the wire image, For the The gradient direction of edge pixels, is the maximum value of the gradient direction of each edge pixel in the wire image, For the The variance of the gradient direction of each edge pixel in the edge where the edge pixel is located, is the normalization function; If the probability If it is greater than the set probability threshold, then the edge pixels belong to the wire area.

4. The power transmission line inspection method based on image processing according to claim 1, characterized in that: The wire region in the image of each section of the wire is extracted using the Otsu algorithm.

5. The power transmission line inspection method based on image processing according to claim 1, characterized in that: The method also includes performing denoising and grayscale processing on the images of each section of the conductor of the transmission line.

6. The power transmission line inspection method based on image processing according to claim 1, characterized in that: The similarity between the sequences is calculated using a dynamic time warping algorithm.

7. The power transmission line inspection method based on image processing according to claim 1, characterized in that: The ratio threshold is set to 0.8.

Images