A method and system for detecting galloping of a transmission line
Through the pre-trained semantic segmentation model and wire area tracking method, intelligent monitoring of transmission line dance is realized, the problems of low efficiency and low intelligence in the existing technology are solved, and the accuracy and efficiency of detection are improved.
Patent Information
- Application Number
- CN202411230355.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-09-04
AI Technical Summary
The existing transmission line dance detection methods are inefficient and low in intelligence, especially in severe weather conditions, which are difficult to effectively monitor and judge the line dance, resulting in threatening the safety and stability of the power grid.
The pre-trained semantic segmentation model is used to obtain the wire area of the transmission line image, and the line dancing pixel amplitude is calculated through the matching relationship of adjacent frame images. The wire area is dynamically updated with the transmission line area tracking method to realize intelligent monitoring and discrimination.
It improves the intelligence level of transmission line dance detection, enhances detection efficiency, reduces dependence on pre-trained models, saves computing resources, and improves detection accuracy.
Smart Images

Figure CN118736427B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of smart grids, and in particular to a method and system for detecting power transmission line galloping. Background Art
[0002] With the development of social economy, the demand for electricity consumption continues to rise, and it is increasingly important to ensure the reliability of power supply in the power system. Among them, the galloping of transmission lines has become the main factor that troubles the safe and reliable operation of the power grid in the power system. For example, when encountering severe weather such as strong winds and ice, the three-phase conductors will experience low-frequency galloping. The galloping of transmission lines may cause power grid failures such as wire detachment, collision with other objects, wind-induced discharge, phase-to-phase contact, and damage to insulators and hardware. In severe cases, it may cause accidents such as wire breakage and even tower collapse, which seriously affects the safe and stable operation of the power grid.
[0003] At present, the detection of transmission line galloping is mainly carried out by manual line patrol or sensor monitoring. Among them, manual line patrol is seriously affected by the weather. Transmission line galloping mostly occurs in bad weather, which brings great difficulties to the travel of line patrol personnel. At the same time, manual line patrol detection has the problems of low efficiency, high cost and low intelligence. Sensor monitoring uses acceleration sensor technology, high-precision GPS detection technology, conductor tension judgment technology, etc., which needs to be analyzed based on complex signals, and has the problems of difficult judgment and lack of intuitiveness.
[0004] In view of this, it is necessary to propose a transmission line galloping detection method and system to solve or at least alleviate the above defects. Summary of the invention
[0005] The main purpose of the present invention is to provide a method and system for detecting power transmission line galloping, aiming to solve the technical problems of low efficiency and low intelligence level in the existing manual inspection of power transmission lines.
[0006] To achieve the above-mentioned purpose, the present invention provides a method for detecting the galloping of a power transmission line, comprising the steps of: S10, acquiring multiple frames of power transmission line images arranged in time sequence; S20, using a pre-trained semantic segmentation model to acquire a power transmission line area of a first frame of the power transmission line image; S30, acquiring multiple groups of line galloping pixel amplitudes according to the matching relationship between two adjacent frames of the power transmission line image; specifically comprising: S31, acquiring pixel comparison information when the power transmission line area of the current frame of the power transmission line image migrates to the adjacent next frame of the power transmission line image; S32, judging, according to the pixel comparison information, whether the power transmission line area corresponding to the power transmission cable in the current frame of the power transmission line image is the same as the power transmission line area corresponding to the power transmission cable in the adjacent next frame of the power transmission line image. Whether the transmission conductor area corresponding to the transmission cable in the power line image has a transmission conductor area offset; S331, if the transmission conductor area offset occurs and it is determined that the line is dancing, the transmission conductor area tracking method is used to update the transmission conductor area of the next frame transmission line image; S34, according to the updated transmission conductor area of the next frame transmission line image and the transmission conductor area of the current frame transmission line image, the line dancing pixel amplitude of the adjacent next frame transmission line image and the current frame transmission line image is obtained; S35, the next frame transmission line image is defined as the current frame transmission line image; repeat steps S31 to S35 to obtain multiple groups of line dancing pixel amplitudes.
[0007] Further, step S30 also includes: S36, judging whether the interval frame number between the current frame transmission line image and the first frame transmission line image is greater than a preset interval threshold; S37, if the interval frame number is greater than the preset interval threshold, obtaining the transmission line area offset between the transmission line area corresponding to the transmission cable in the current frame transmission line image and the transmission line area corresponding to the transmission cable in the first frame transmission line image; S38, determining that the current frame transmission line image in which no transmission line area offset occurs for the first time is the target frame transmission line image; S39, determining the line dancing frequency according to the interval time between the target frame transmission line image and the first frame transmission line image.
[0008] Furthermore, the method further includes the step: S332, if no transmission line area deviation occurs and it is determined that the line does not gallop, then the transmission line area in the current frame transmission line image is updated to the transmission line area in the next frame transmission line image, and the process proceeds to step S35.
[0009] Further, step S31 specifically includes: extracting the transmission line area of the current frame transmission line image;
[0010] The transmission line area of the current frame transmission line image is transferred to the adjacent next frame transmission line image, using the formula
[0011]
[0012] Compare the difference in pixels of the transmission line area of two adjacent frames to obtain pixel comparison information, wherein the pixel comparison information includes whether background pixels appear in the migrated transmission line area; wherein DiffCount is used to count the number of different pixel points, Mask represents the extracted transmission line area, x, y represent the pixel distance in the horizontal and vertical directions with the image vertex as the origin, in the present invention, x, y represent the pixel distance in the horizontal and vertical directions with the lower left corner of the picture as the origin, t represents the t-th frame image, t+1 represents the next frame adjacent to the t-th frame image, T represents the pixel difference threshold, 0 represents the absence of background pixels, and 1 represents the presence of background pixels.
[0013] Further, step S32 specifically includes: if no background pixels appear, determining that there is no transmission wire area offset between the transmission wire area corresponding to the transmission cable in the current frame transmission line image and the transmission wire area corresponding to the transmission cable in the adjacent next frame transmission line image; if background pixels appear, determining that there is a transmission wire area offset between the transmission wire area corresponding to the transmission cable in the current frame transmission line image and the transmission wire area corresponding to the transmission cable in the adjacent next frame transmission line image.
[0014] Furthermore, in step S331, "updating the transmission line area of the next frame of the transmission line image using the transmission line area tracking method" specifically includes: segmenting the transmission line area of the current frame of the transmission line image to obtain multiple line segments; gradually pre-offsetting each line segment vertically according to a preset pixel value and using twin network matching to obtain offset information, and determining the target offset of each line segment based on the offset information; integrating each line segment and the corresponding target offset for tracking and then updating the transmission line area of the next frame of the transmission line image of the adjacent frame.
[0015] Further, step S34 specifically includes: obtaining feature points of the next frame of the transmission line image and the current frame of the transmission line image, wherein the feature point is the image midpoint of the transmission line area; and determining the offset pixel distance of the feature point as the line dancing pixel amplitude.
[0016] Furthermore, a three-phase transmission wire semantic segmentation dataset is constructed; a three-phase transmission wire semantic segmentation model is constructed; and the three-phase transmission wire semantic segmentation model is trained using the three-phase transmission wire semantic segmentation dataset to obtain a pre-trained semantic segmentation model.
[0017] Furthermore, before step S10, the method includes the following steps: obtaining current environmental information; normalizing the current environmental information to obtain a current impact value, if the current impact value reaches a preset start threshold; obtaining a captured video with a power transmission cable; frame-processing the captured video and performing image optimization processing on each frame, the image optimization processing including image smoothing, image enhancement and image denoising.
[0018] The present invention also provides a power transmission line galloping detection system, comprising an image acquisition unit and a processor, wherein the image acquisition unit is used to acquire a power transmission line image, and the processor is used to implement the steps of the above-mentioned power transmission line galloping detection method.
[0019] Compared with the prior art, the power transmission line galloping detection method provided by the present invention has the following beneficial effects:
[0020] The present invention provides a method for detecting the galloping of a power transmission line. After acquiring multiple frames of power transmission line images arranged in time sequence, a pre-trained semantic segmentation model is first used to acquire the power transmission wire area of the first frame of the power transmission line image. Then, according to the matching relationship between two adjacent frames of the power transmission line image, multiple groups of line galloping pixel amplitudes are acquired, thereby realizing intelligent monitoring of line galloping through the power transmission line image. When acquiring the line galloping pixel amplitude, pixel comparison information when the power transmission wire area of the current frame of the power transmission line image is migrated to the adjacent next frame of the power transmission line image is first acquired. Then, according to the pixel comparison information, it is determined whether the next frame of the power transmission line image has line galloping relative to the current frame of the power transmission line image. When the line gallops, a power transmission wire area tracking method is used to update the power transmission wire area of the next frame of the power transmission line image. Then, according to the updated power transmission wire area of the next frame of the power transmission line image and the power transmission wire area of the current frame of the power transmission line image, a power transmission wire area tracking method is used to update the power transmission wire area of the next frame of the power transmission line image. Take the line dancing pixel amplitude of the adjacent next frame transmission line image and the current frame transmission line image; finally define the next frame transmission line image as the current frame transmission line image; repeat the steps to obtain multiple groups of line dancing pixel amplitudes; fully consider the changes in line appearance due to changes in light and weather over time, dynamically adjust and update the transmission line area in the next frame transmission line image, and obtain the line dancing pixel amplitude of the adjacent next frame transmission line image according to the updated transmission line area of the next frame transmission line image and the transmission line area of the current frame transmission line image. It is only necessary to use the pre-trained semantic segmentation model when obtaining the transmission line area of the first frame transmission line image. For the subsequent transmission line images of the first frame transmission line image, the method is updated to obtain the transmission line area, avoiding the need to call the pre-trained semantic segmentation model each time, and greatly saving computing resources while having high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.
[0022] Figure 1 Schematic diagram of a flow chart of a method for detecting power transmission line galloping in one embodiment of the present invention;
[0023] Figure 2 The present invention is a flow chart of extracting the transmission line area in an image based on a training line semantic segmentation model.
[0024] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION
[0025] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0026] 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 only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0027] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.
[0028] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.
[0029] Please refer to the attached Figure 1 and Figure 2The present invention provides a method for detecting the galloping of a power transmission line, comprising the steps of: S10, obtaining a plurality of frames of power transmission line images arranged in time sequence; S20, using a pre-trained semantic segmentation model to obtain a power transmission line area of a first frame of the power transmission line image; S30, according to the matching relationship between two adjacent frames of the power transmission line image, obtaining a plurality of groups of line galloping pixel amplitudes; specifically comprising: S31, obtaining pixel comparison information when the power transmission line area of the current frame of the power transmission line image migrates to the adjacent next frame of the power transmission line image; S32, judging the power transmission line area corresponding to the power transmission cable in the current frame of the power transmission line image according to the pixel comparison information;
[0030] Whether there is a transmission conductor area offset between the conductor area and the transmission conductor area corresponding to the transmission cable in the adjacent next frame transmission line image; S331, if the transmission conductor area offset occurs and it is determined that the line is dancing, the transmission conductor area tracking method is used to update the transmission conductor area of the next frame transmission line image; S34, according to the updated transmission conductor area of the next frame transmission line image and the transmission conductor area of the current frame transmission line image, calculation is performed to obtain the line dancing pixel amplitude of the adjacent next frame transmission line image and the current frame transmission line image; S35, define the next frame transmission line image as the current frame transmission line image; repeat steps S31 to S35 to obtain multiple groups of line dancing pixel amplitudes.
[0031] The present invention provides a method for detecting the galloping of a power transmission line. After acquiring multiple frames of power transmission line images arranged in time sequence, a pre-trained semantic segmentation model is first used to acquire the power transmission wire area of the first frame of the power transmission line image. Then, according to the matching relationship between two adjacent frames of the power transmission line image, multiple groups of line galloping pixel amplitudes are acquired, thereby realizing intelligent monitoring of line galloping through the power transmission line image. When acquiring the line galloping pixel amplitude, pixel comparison information when the power transmission wire area of the current frame of the power transmission line image is migrated to the adjacent next frame of the power transmission line image is first acquired. Then, according to the pixel comparison information, it is determined whether the next frame of the power transmission line image has line galloping relative to the current frame of the power transmission line image. When the line gallops, a power transmission wire area tracking method is used to update the power transmission wire area of the next frame of the power transmission line image. Then, according to the updated power transmission wire area of the next frame of the power transmission line image and the power transmission wire area of the current frame of the power transmission line image, a power transmission wire area tracking method is used to update the power transmission wire area of the next frame of the power transmission line image. Take the line dancing pixel amplitude of the adjacent next frame transmission line image and the current frame transmission line image; finally define the next frame transmission line image as the current frame transmission line image; repeat the steps to obtain multiple groups of line dancing pixel amplitudes; fully consider the changes in line appearance due to changes in light and weather over time, dynamically adjust and update the transmission line area in the next frame transmission line image, and obtain the line dancing pixel amplitude of the adjacent next frame transmission line image according to the updated transmission line area of the next frame transmission line image and the transmission line area of the current frame transmission line image. It is only necessary to use the pre-trained semantic segmentation model when obtaining the transmission line area of the first frame transmission line image. For the subsequent transmission line images of the first frame transmission line image, the method is updated to obtain the transmission line area, avoiding the need to call the pre-trained semantic segmentation model each time, and greatly saving computing resources while having high accuracy.
[0032] It can be understood that in the present invention, the transmission wire area is the area corresponding to a single transmission cable; when judging the dancing of a three-phase transmission wire, a pre-trained semantic segmentation model is used to obtain three frames of current images at the same time, and each frame of the current image has a transmission wire area corresponding to a single transmission cable; the input of the pre-trained semantic segmentation model (pre-trained semantic segmentation model) is the first frame of the transmission line image, and the output of the semantic segmentation model is the area where the single transmission cable is located in the first frame of the transmission line image.
[0033] It can be understood that when judging line dancing, the transmission line area of the current frame is migrated to the adjacent frame image, and the pixels of the transmission line areas of the two adjacent frames are compared. If background pixels appear in the part of the transmission line area of the migrated current frame, it can be determined that the transmission line area is offset, and then the line is judged to be dancing.
[0034] Furthermore, step S30 further includes:
[0035] S36, judging whether the interval frame number between the current frame transmission line image and the first frame transmission line image is greater than the preset interval threshold; S37, obtaining the interval frame number greater than the preset interval threshold, judging the transmission line area offset between the transmission line area corresponding to the transmission cable in the current frame transmission line image and the transmission line area corresponding to the transmission cable in the first frame transmission line image; S38, determining the current frame transmission line image in which the transmission line area offset does not occur for the first time as the target frame transmission line image; S39, determining the line dancing frequency according to the interval time between the target frame transmission line image and the first frame transmission line image. It can be understood that by obtaining the subsequent transmission line image outside the pre-examination interval time through the preset interval threshold, the current frame transmission line image in which the transmission line area offset does not occur for the first time is determined as the target frame transmission line image from the subsequent transmission line image, thereby avoiding the error caused by the small amplitude of the adjacent frames, and reducing the error in determining the line dancing frequency. Optionally, the preset interval threshold can be 1, or any value from 2 to 30.
[0036] Specifically, a start-up judgment may be set, and multiple frames of transmission line images are obtained after the start-up judgment is activated, and the first frame of the transmission line image is an image of the line in a static state. During the dancing of the line, the transmission line area corresponding to the transmission cable in the frame transmission line image of different frames updated in step S331 is compared with the transmission line area of the first frame of the transmission line image. If the two coincide, it can be considered that the transmission cable has returned to a static state. By calculating the time difference between the two transmission cables returning to a static state, the frequency of the line dancing can be obtained. The formula is as follows:
[0037]
[0038] Among them, f represents the dancing frequency, T is the dancing period, and the number of pictures between the two returns to the static state of the wire is n. Since the time between adjacent frames of the video stream is fixed at , where the interval duration is equal to the number of intervals multiplied by the time between adjacent frames, so the dance cycle .
[0039] Furthermore, the method further comprises the steps of:
[0040] S332, if no transmission line area deviation occurs to determine that the line does not gallop, the transmission line area in the current frame transmission line image is updated to the transmission line area in the next frame transmission line image, and the process proceeds to step S35. Optionally, if no transmission line area deviation occurs to determine that the line does not gallop, the galloping pixel amplitude of the current frame transmission line image is assigned to 0.
[0041] Furthermore, step S31 specifically includes:
[0042] Extract the transmission line area of the current frame transmission line image; migrate the transmission line area of the current frame transmission line image to the adjacent next frame transmission line image, using the formula
[0043]
[0044] Compare the difference in pixels of the transmission line area of two adjacent frames to obtain pixel comparison information, wherein the pixel comparison information includes whether background pixels appear in the migrated transmission line area; wherein DiffCount is used to count the number of different pixel points, Mask represents the extracted transmission line area, x, y represent the pixel distance in the horizontal and vertical directions with the image vertex as the origin, in the present invention, x, y represent the pixel distance in the horizontal and vertical directions with the lower left corner of the picture as the origin, t represents the t-th frame image, t+1 represents the next frame adjacent to the t-th frame image, T represents the pixel difference threshold, 0 represents the absence of background pixels, and 1 represents the presence of background pixels.
[0045] Specifically, adjacent frames of transmission line area tracking and comparison are adopted. Adjacent frames of transmission line images have the characteristics of environmental consistency, which is manifested as small differences in appearance. Therefore, pixel comparison is used to distinguish the dancing of the wire. According to the transmission line area extracted from the current frame transmission line image, the transmission line area is migrated to the next adjacent frame image. By comparing the pixel difference of the transmission line area of two adjacent frames, if there are background pixels in the migrated transmission line area, it can be judged that the wire is offset. In the present invention, pixel difference mainly refers to color difference. Its main principle is that it is believed that the wire pixels will not change in the case of two adjacent frames, so the color of the pixel point can be judged.
[0046] Further, step S32 specifically includes:
[0047] If no background pixels appear, it is determined that there is no transmission wire area offset between the transmission wire area corresponding to the transmission cable in the current frame transmission line image and the transmission wire area corresponding to the transmission cable in the adjacent next frame transmission line image; if background pixels appear, it is determined that there is a transmission wire area offset between the transmission wire area corresponding to the transmission cable in the current frame transmission line image and the transmission wire area corresponding to the transmission cable in the adjacent next frame transmission line image.
[0048] It can be understood that the transmission wire area iteration in the present invention updates the transmission wire area in the next adjacent frame image to the transmission wire area in the current frame image. If there is no wire offset in the next adjacent frame image, the transmission wire area in the adjacent frame image is directly updated to the transmission wire area in the current frame image; if there is a wire offset, the transmission wire area is extracted according to the wire tracking result of the adjacent frame image obtained by the transmission wire area tracking algorithm and updated to the transmission wire area of the current frame image. The updated current frame transmission wire area is used to perform wire tracking comparison on subsequent image frames, and the input video stream image is continuously processed.
[0049] Further, in step S331, updating the transmission line area of the next frame of the transmission line image by using the transmission line area tracking method specifically includes:
[0050] The transmission line area of the current frame transmission line image is segmented to obtain multiple line segments; each line segment is gradually pre-shifted along the vertical direction according to the preset pixel value and the offset information is obtained by using the twin network matching, and the target offset of each line segment is determined according to the offset information; each line segment and the corresponding target offset are integrated for tracking and then the transmission line area of the next frame transmission line image of the adjacent frame is updated.
[0051] It can be understood that in the specific implementation, when the line is dancing and offset in the next frame of the transmission line image, the transmission line area uses the feature matching method to track the dancing line of the current video frame, and through the line segmentation processing, the twin network is used to match the offset of the line segment of the current frame, and finally the offset of the line segment is integrated to obtain the complete line tracking result. Specifically, let the length of the transmission line area of the current frame be L, segment the line, and intercept the tiny segment. As a line segment, it can be assumed that the line segment will not undergo obvious deformation during the line dancing process, and each line segment will only move in the vertical direction; the twin network uses the VGG model based on the full convolutional network and uses the pre-trained weights to initialize the model parameters; for each line segment, obtain the The area is used as the matching area. On the adjacent frame images, the offset of the line segment is obtained by vertical pixel wandering and using twin similarity matching. The formula is as follows:
[0052]
[0053]
[0054] Where F(k) represents the twin network, k represents the k pixels offset in the vertical direction of the line segment, and the value range is , cos is the feature vector similarity calculated using cosine similarity, and They represent the line segment region representation vector at time t and the line segment region representation vector offset by k pixels in the vertical direction at time t+1, respectively. The network performs vector representation calculations. The twin network is used to obtain the optimal offset position for each line segment, and finally the complete line tracking result is obtained by integrating each line segment, and the updated line area is extracted.
[0055] Furthermore, before step S10, the method includes the following steps:
[0056] Obtain current environmental information; normalize the current environmental information to obtain the current impact value, if the current impact value reaches a preset start threshold; obtain a captured video with a power transmission cable; frame the captured video and perform image optimization processing on each frame, the image optimization processing includes image smoothing, image enhancement and image denoising.
[0057] Optionally, line galloping discrimination factors are obtained through sensors, weather forecasts, line parameters, etc., and the occurrence of line galloping is judged by combining the discrimination factors.
[0058] Collection of line galloping discrimination factors: Use sensors to measure the specific environment of the line, and further obtain line galloping discrimination factors through weather forecasts, line parameters, etc., to build diversified line galloping factor information. Line galloping discrimination factors include: sensor information, such as temperature and humidity information obtained by temperature and humidity sensors; wind speed and wind direction information obtained by wind sensors; weather forecast information, such as rainfall, snowfall, wind direction, wind speed, etc.; line parameter information, such as surrounding terrain, line span, etc. Line galloping factor discrimination: Use the collected line galloping factor information and combine it with the line galloping discrimination standard. If the conditions for the occurrence of transmission line galloping are met, the line galloping intelligent discrimination method is enabled. The discrimination standard needs to be set in the range where galloping is prone to occur: such as wind speed between 4 and 20m / s, ice thickness between 4 and 6 mm, and temperature between -3 and 0℃. In addition, the terrain is generally plain open land, hills, rivers and lakes, etc. Specific parameters and values need to be set according to specific regional conditions.
[0059] It can be understood that when capturing video frames and optimizing images (image data preprocessing), after the activation of the judgment is started, the video stream transmitted by the front-end camera device of the transmission line is framed and the first video frame is obtained as the line extraction image; the initial image transmitted from the front-end camera device contains a lot of noise data, and image preprocessing methods need to be used for image optimization, including: image smoothing, image enhancement and image denoising. Among them,
[0060] Image smoothing: Use a mean filter to smooth the image and reduce noise. The formula is as follows:
[0061]
[0062] in, Indicates that the original image is at position The pixel value at Indicates that the filtered image is at position The pixel value at , N is the radius of the filter, indicating the size of the filter is .
[0063] Image enhancement: Enhance image clarity and quality by adjusting parameters such as contrast, brightness and color balance;
[0064] In the process of image denoising, the wavelet denoising algorithm is used to reduce the noise in the image. Its formula can be expressed as:
[0065]
[0066] Where X is the initial image, DWT represents wavelet transform, and IDWT represents inverse wavelet transform. Represents the denoised image, Denoise represents threshold processing, which can be processed using soft thresholding:
[0067]
[0068] Where w is the size of the wavelet coefficient, is the size of the wavelet coefficient after applying the threshold, is the threshold value.
[0069] Furthermore, a three-phase transmission line semantic segmentation dataset is constructed; a three-phase transmission line semantic segmentation model is constructed; the three-phase transmission line semantic segmentation model is trained using the three-phase transmission line semantic segmentation dataset (training the three-phase transmission line semantic segmentation model) to obtain a pre-trained semantic segmentation model. Furthermore, the pre-trained semantic segmentation model is used to extract the transmission line region from the first frame of the transmission line image.
[0070] It can be understood that to construct a semantic segmentation dataset of three-phase transmission lines, image samples containing three-phase transmission lines at different time points and in different weather conditions are collected from the camera device at the front end of the transmission line; the collected image samples are manually annotated, and each pixel is associated with the corresponding category, that is, the three-phase transmission lines are marked as the target category, and the background is marked as the background category, so as to construct a semantic segmentation dataset of three-phase transmission lines.
[0071] It can be understood that the semantic segmentation model of three-phase transmission lines is constructed based on YOLOv8. YOLOv8 adopts the idea of decoupling head to make the training and reasoning of the network more efficient. At the same time, YOLOv8 has fast reasoning speed and good model performance, which can meet the requirements of accuracy and speed in the process of power transmission video stream information processing. The basic structure of the YOLOv8 model includes four parts: Backbone, Neck, Head, and Loss. Backbone is the backbone network, which adopts the CSPDarkNet network structure and contains a series of convolution and pooling layers to gradually reduce the size of the feature map and extract high-level semantic features; Neck is the intermediate connection layer, which adopts the PANet network structure. The goal is to further extract and integrate features in order to better meet the requirements of the semantic segmentation task; Head is the detection head. In the semantic segmentation model, additional branches are required to generate semantic segmentation prediction results, such as convolution layers, upsampling layers, or transposed convolution layers, which are used to increase the resolution of the feature map to the same size as the input image and output the semantic label of each pixel; In the Loss part, it is used to define a loss function suitable for the semantic segmentation task. The binary BCE (Binary Cross-Entropy) loss function can be used. The formula is as follows:
[0072]
[0073] in, is the pixel classification probability output of the model, and y is the corresponding true label.
[0074] It can be understood that the three-phase transmission line semantic segmentation model is trained using the constructed semantic segmentation dataset. During the training process, the model parameters are adjusted through the back propagation algorithm and the optimizer so that it can accurately identify and segment the three-phase transmission line in the image.
[0075] It can be understood that the three-phase transmission line semantic segmentation model is used for line extraction, and the trained three-phase transmission line semantic segmentation model is applied to the first frame of transmission line image to accurately extract the transmission line area of the first frame of transmission line image.
[0076] Furthermore, the physical amplitude of the line dancing is obtained according to the line dancing pixel amplitude and the image displacement ratio.
[0077] Further, step S34 specifically includes:
[0078] Obtain the feature points of the next frame of the transmission line image and the current frame of the transmission line image, wherein the feature point is the image midpoint of the transmission line area; determine the offset pixel distance of the feature point as the line dancing pixel amplitude. In the specific implementation, for the transmission line area of the next frame of the transmission line image and the transmission line area of the current frame of the transmission line image updated in step S331, select the image midpoint of each transmission line area as the feature point, obtain the feature point offset in different video frames, and record the pixel point with the maximum feature point offset as the line dancing pixel amplitude. More preferably, according to the actual length of the line and the pixel length of the line, the actual distance represented by each unit pixel can be calculated, and then the physical amplitude of the line dancing can be calculated according to the line dancing pixel amplitude.
[0079] The present invention also provides a power transmission line galloping detection system, comprising an image acquisition unit and a processor, wherein the image acquisition unit is used to acquire a power transmission line image, and the processor is used to implement the steps of the above-mentioned power transmission line galloping detection method.
[0080] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A method for detecting the galloping of a power transmission line, characterized in that: Includes steps: S10, acquiring a plurality of frames of transmission line images arranged in time sequence; S20, using a pre-trained semantic segmentation model to obtain a transmission line area of the first frame of the transmission line image; S30, obtaining multiple groups of line dancing pixel amplitudes according to the matching relationship between two adjacent frames of transmission line images; specifically comprising: S31, obtaining pixel comparison information when the power transmission line area of the current frame power transmission line image is migrated to the adjacent next frame power transmission line image; Step S31 specifically includes: Extracting the transmission line area of the current frame transmission line image; The transmission line area of the current frame transmission line image is transferred to the adjacent next frame transmission line image, using the formula Diff(X,y)=|Mask t (X,y)-Mask t+1 (X,y)| Comparing the difference between the pixels of the power transmission line area of two adjacent frames to obtain pixel comparison information, wherein the pixel comparison information includes whether background pixels appear in the part of the power transmission line area that is migrated; Among them, DiffCount is used to count the number of different pixels, Mask represents the extracted transmission line area, x, y represent the pixel distance in the horizontal and vertical directions with the image vertex as the origin, t represents the t-th frame image, t+1 represents the next frame adjacent to the t-th frame image, T represents the pixel difference threshold, 0 represents the absence of background pixels, and 1 represents the presence of background pixels; S32, judging whether a transmission line area corresponding to a transmission cable in a current frame of the transmission line image and a transmission line area corresponding to a transmission cable in an adjacent next frame of the transmission line image have a transmission line area offset according to the pixel comparison information; S331, if the transmission line region is offset and it is determined that the line is galloping, a transmission line region tracking method is used to update the transmission line region of the next frame of the transmission line image; S34, performing calculation according to the updated transmission line area of the next frame transmission line image and the transmission line area of the current frame transmission line image, to obtain the line dancing pixel amplitude of the adjacent next frame transmission line image and the current frame transmission line image; S35, defining the next frame of the power transmission line image as the current frame of the power transmission line image; Repeat steps S31 to S35 to obtain multiple groups of line dancing pixel amplitudes; The method further includes the step: S332, if no transmission line area deviation occurs and it is determined that the line does not gallop, then the transmission line area in the current frame transmission line image is updated to the transmission line area in the next frame transmission line image, and the process proceeds to step S35; The method of updating the transmission line area of the next frame of the transmission line image by the transmission line area tracking method in step S331 specifically includes: segmenting the transmission line area of the current frame of the transmission line image to obtain multiple line segments; gradually pre-shifting each line segment vertically according to a preset pixel value and using twin network matching to obtain offset information, and determining the target offset of each line segment according to the offset information; integrating each line segment and the corresponding target offset for tracking and then updating the transmission line area of the next frame of the transmission line image of the adjacent frame.
2. The method for detecting power transmission line galloping according to claim 1, characterized in that: Step S30 also includes: S36, determining whether the interval number of frames between the current frame of the power transmission line image and the first frame of the power transmission line image is greater than a preset interval threshold; S37, if the interval frame number is greater than the preset interval threshold, obtaining the transmission line area offset between the transmission line area corresponding to the transmission cable in the current frame transmission line image and the transmission line area corresponding to the transmission cable in the first frame transmission line image; S38, determining the current frame transmission line image in which no transmission line area deviation occurs for the first time as the target frame transmission line image; S39, determining the line dancing frequency according to the interval between the target frame power transmission line image and the first frame power transmission line image.
3. The method for detecting power transmission line galloping according to claim 1, characterized in that: Step S32 specifically includes: If no background pixels appear, determining that there is no transmission line area offset between the transmission line area corresponding to the transmission cable in the current frame transmission line image and the transmission line area corresponding to the transmission cable in the next adjacent frame transmission line image; If background pixels appear, it is determined that a transmission line area corresponding to the transmission cable in the current frame transmission line image and a transmission line area corresponding to the transmission cable in the next adjacent frame transmission line image have a transmission line area offset.
4. The method for detecting power transmission line galloping according to any one of claims 1 to 3, characterized in that: Step S34 specifically includes: Acquire feature points of the next frame of the transmission line image and the current frame of the transmission line image, wherein the feature point is the image midpoint of the transmission line area; The offset pixel distance of the feature point is determined as the line dancing pixel amplitude.
5. The method for detecting power transmission line galloping according to any one of claims 1 to 3, characterized in that: Construct a three-phase transmission line semantic segmentation dataset; construct a three-phase transmission line semantic segmentation model; use the three-phase transmission line semantic segmentation dataset to train the three-phase transmission line semantic segmentation model to obtain a pre-trained semantic segmentation model.
6. The method for detecting the galloping of a power transmission line according to any one of claims 1 to 3, characterized in that: Before step S10, the method includes the following steps: Get current environment information; Normalizing the current environmental information to obtain a current impact value, if the current impact value reaches a preset start threshold; Get a video of the shooting with power transmission cables; The captured video is frame-processed and each frame is subjected to image optimization processing, wherein the image optimization processing includes image smoothing, image enhancement and image denoising.
7. A transmission line galloping detection system, characterized in that: comprising an image acquisition unit and a processor, The image acquisition unit is used to acquire a transmission line image, and the processor is used to implement the steps of the transmission line galloping detection method according to any one of claims 1 to 6.
Citation Information
Patent Citations
Power transmission line galloping detection method and device based on gray scale fluctuation field and gray scale accumulation
CN116402770A
Spacer-free high-voltage transmission line galloping early warning method in night vision scene
CN116453034A
Cited By
Omnidirectional sensing detection system and method for galloping of power transmission line
CN121430798A
Power transmission line galloping detection method based on environmental energy capture
CN121521248A
