A flash discharge monitoring method for high voltage transmission network

Abstract

The invention discloses a method for monitoring lightning discharge of a high-voltage transmission network, relates to the field of electric power, and solves the problem of inaccurate monitoring of lightning discharge of the existing high-voltage transmission network. The method is specifically as follows: step S1, based on the monitored highlight phenomenon, using a monitoring camera to acquire video data of the high-voltage transmission network; step S2, analyzing the monitoring video stream saved by the monitoring camera, and analyzing to obtain a monitoring image data set of the high-voltage transmission network; step S3, deriving the monitoring image data set, dividing the sky and ground images in the monitoring image data set into a sky image and a ground image, and adding a black background to the divided sky image to obtain a candidate frame; step S4, comparing the candidate frame with the cache frame in the cache frame data set for histogram similarity, filtering the abnormal situation and outputting the lightning frame, the invention realizes accurate monitoring of the lightning discharge phenomenon of the high-voltage transmission network.

Description

Technical Field

[0001] The invention belongs to the technical field of flash discharge monitoring, and in particular to a flash discharge monitoring method for a high voltage transmission network. Background Art

[0002] High-voltage transmission network flashover monitoring is used to timely detect and record electrical discharge phenomena occurring in high-voltage transmission networks. Flashover events may cause equipment damage, system failures, and endanger the safety of operators. By monitoring the characteristics of local discharge in high-voltage transmission networks, potential failure risks can be warned in a timely manner, and necessary maintenance and protection measures can be taken. This not only helps to extend the service life of equipment, improve system stability and reliability, but also optimize equipment operation management strategies, reduce system downtime and maintenance costs, and reduce environmental and personnel safety risks.

[0003] Existing methods for monitoring flash discharges in high-voltage transmission networks include electric field sensor technology, current sensor technology, infrared imaging technology, and image processing based on deep learning. However, when strong lightning weather occurs, the lightning features in the captured image may be covered by bright electric light, making it impossible to distinguish between strong lightning scenes and flash discharge scenes in high-voltage transmission networks, which in turn affects the accuracy of flash discharge monitoring in high-voltage transmission networks, resulting in false alarms or missed alarms.

[0004] To this end, we propose a flash discharge monitoring method for high-voltage transmission networks. Summary of the invention

[0005] In view of the deficiencies in the prior art, an object of the present invention is to provide a flash discharge monitoring method for a high voltage transmission network.

[0006] The technical problems to be solved by the present invention are:

[0007] How to achieve accurate monitoring of flashover phenomena in high-voltage transmission networks.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] In a first aspect, a flash discharge monitoring method for a high voltage transmission network is provided, the method being specifically as follows:

[0010] Step S1, based on the monitored highlight phenomenon, using a surveillance camera to acquire video data of the high-voltage transmission network;

[0011] Step S2, analyzing the monitoring video stream saved by the monitoring camera to obtain a monitoring image data set of the high-voltage transmission network;

[0012] Step S3, exporting a surveillance image data set, segmenting the sky-ground image in the surveillance image data set into a sky image and a ground image, and adding a black background to the segmented sky image to obtain a candidate frame;

[0013] Step S4, comparing the candidate frame with the cached frames in the cached frame data set by histogram similarity, filtering out abnormalities and outputting a lightning frame;

[0014] Step S5, sending the output lightning frame to the designated display terminal for result display.

[0015] Furthermore, the step S1 includes the following sub-steps:

[0016] Step S11, using a high-sensitivity photosensor and synchronizing the time of the photosensor with that of the surveillance camera so that the time point in the video stream corresponds to the time point of the photosensor;

[0017] Step S12, setting an alarm threshold of the photosensor, when the photosensor receives a highlight reading exceeding the alarm threshold at a certain time, it is determined that a highlight phenomenon occurs and the current time is recorded as the end time, and the second before the end time is recorded as the start time;

[0018] Step S13, extracting a section from the video stream as the monitoring video stream of the monitoring camera according to the start time and the end time.

[0019] Furthermore, the step S2 includes the following sub-steps:

[0020] Step S21, respectively obtaining monitoring image frames corresponding to the start time and the end time of the monitoring video stream, wherein the monitoring image frame corresponding to the end time is defined as the current monitoring image frame, and the monitoring image frame corresponding to the start time is defined as the previous monitoring image frame;

[0021] Step S22, adjusting the resolution of the decoded monitoring image frame to 256×256, and comparing the difference values ​​of adjacent monitoring image frames;

[0022] Step S23, establishing a mapping relationship between the key frame and the previous monitoring image frame and summarizing them into a monitoring image data set.

[0023] Furthermore, the calculation steps of the difference value are as follows:

[0024] Calculate the pixel color information difference between the current monitoring image frame and the previous monitoring image frame. The analysis process of the pixel color information difference is as follows:

[0025] A coordinate system is established with the upper left corner vertices of the two surveillance image frames as the origin, and the two sides of the surveillance image frames are used as the x-axis and y-axis respectively. The value range of x and y is 0 to 255. The RGB value of each coordinate point in the coordinate system is obtained. The RGB value consists of R value, G value and B value;

[0026] Calculate the difference between the RGB values ​​of the two surveillance image frames at the same coordinate point to obtain the RGB difference at the same coordinate point;

[0027] If the RGB difference of the same coordinate point in the two surveillance image frames is greater than or equal to the preset RGB difference, the corresponding coordinate point is recorded as an abnormal coordinate point; if the RGB difference of the same coordinate point in the two surveillance image frames is less than the preset RGB difference, no operation is performed;

[0028] The number of abnormal coordinate points is counted and recorded as the number of abnormal coordinate points. When the number of abnormal coordinate points is greater than or equal to the preset number of abnormal coordinate points, the difference value is recorded as 1; when the number of abnormal coordinate points is greater than or equal to the preset number of abnormal coordinate points, the difference value is recorded as 0.

[0029] Furthermore, the step of calculating the difference value further includes:

[0030] The gray-level co-occurrence matrix is ​​used to calculate the probability distribution difference between adjacent surveillance image frames. The analysis process of the probability distribution difference is as follows:

[0031] Select any pixel in the current monitoring image frame and record the pixel as the initial pixel i (xi, yi), where i is the number of the initial pixel;

[0032] Get the offset of the initial pixel (±a, ±b), get another pixel in the current monitoring image frame according to the offset, record the other pixel as the offset pixel (xi±a, yi±b), the initial pixel and its corresponding offset pixel form a plurality of grayscale pairs Miu(g1i, g2i), u represents the number of the grayscale pair corresponding to the initial pixel i, where g1i is the grayscale value of the initial pixel (xi, yi), and g2i is the grayscale value of the offset pixel (xi+a, yi+b);

[0033] Count the number of occurrences of each grayscale pair Miu (g1i, g2i), divide the number of occurrences of each grayscale pair Miu (g1i, g2i) by the total number of all grayscale pairs, and obtain the normalized probability matrix P1 (g1i, g2i) of the current monitoring image frame;

[0034] Similarly, the normalized probability matrix P2(g1i, g2i) of the previous surveillance image frame is obtained according to the above method, and the Euclidean distance is used Calculate the difference measure between the probability matrices P1 and P2, where (P1(g1i, g2i)-P2(g1i, g2i))² represents the square of the difference in the probabilities of the corresponding positions in the two matrices;

[0035] When the Euclidean distance is greater than or equal to the preset distance, the difference value is recorded as 1, and when the Euclidean distance is less than the preset distance, the difference value is recorded as 0.

[0036] Furthermore, the step of calculating the difference value further includes:

[0037] Calculate the local grayscale difference between the current surveillance image frame and the previous surveillance image frame. The specific steps for calculating the local grayscale difference are as follows:

[0038] The surveillance image frame is evenly divided into 16×16 small areas according to the coordinate system, and each small area contains 256 pixels;

[0039] Use the local median as the local median threshold. For each pixel (x, y), if its gray value is greater than or equal to the local median threshold, the pixel value is set to 1. If its gray value is less than the local median threshold, the pixel value is set to 0. Repeat the pixel binarization process. After each 16×16 small area is processed, a new 16×16 local binary image block is formed, in which each pixel value is 0 or 1.

[0040] The local binary image block is copied to the corresponding position in the coordinate system corresponding to the current monitoring image frame to form a local binary image. Similarly, the local binary image of the previous monitoring image frame is obtained according to the above method;

[0041] The difference degree of local information between two monitoring image frames is calculated by Hamming distance. If the pixel values ​​of the two local binary images at the pixel position are different, the count of Hamming distance increases by 1. If the pixel values ​​of the two local binary images at the pixel position are the same, the count of Hamming distance remains unchanged. The difference counts of all different pixels are summed to obtain the Hamming distance.

[0042] When the Hamming distance is greater than or equal to the preset difference distance, the difference value is recorded as 1, and when the Hamming distance is less than the preset difference distance, the difference value is recorded as 0;

[0043] The difference values ​​corresponding to the feature differences are added together to obtain the total difference value. If the total difference value is greater than or equal to the difference threshold, the current monitoring image frame is recorded as a key frame. If the total difference value is less than the difference threshold, no operation is performed.

[0044] Furthermore, the step S3 includes the following sub-steps:

[0045] Step S31, collecting publicly available sky-ground image datasets, and labeling pictures in the publicly available sky-ground image datasets;

[0046] Step S32, preprocessing the labeled sky-ground image dataset, adjusting the image resolution in the sky-ground image dataset to 256×256, and normalizing the pixel value of each pixel in the sky-ground image to 0-1;

[0047] Step S33, using the NestedUNET model as an image segmentation algorithm, and training the NestedUNET model;

[0048] Step S34, use the trained NestedUNET model to segment the key frames in the summarized monitoring data set and the previous monitoring image frame into sky images and ground images, retain the sky image and add a completely black background, create a completely black background image, the resolution of the completely black background image is 256×256, traverse the pixels of each pixel in the sky image after the segmentation is completed, copy the pixels of all pixels of all sky images to the corresponding positions of the completely black background image, and mark the sky image corresponding to the key frame as a candidate frame after adding the completely black background.

[0049] Furthermore, the step S4 includes the following sub-steps:

[0050] Step S41, the candidate frame is converted into a grayscale image, and a grayscale histogram is constructed. The grayscale histogram construction process is as follows:

[0051] Step S411, creating an array H with a length of 256 to store the pixel count of each gray value, initialized to 0;

[0052] Step S412, traverse the grayscale value of each pixel in the candidate frame, and update the array H according to the grayscale value;

[0053] Step S413, establish a rectangular coordinate system, where the x-axis is the grayscale value, ranging from 0 to 255, and the y-axis is the number of pixels of each grayscale value, and construct a grayscale histogram based on the updated array H. Similarly, obtain the grayscale histogram of the cache frame;

[0054] Step S42, comparing the candidate frame with the cached frame by histogram similarity. When the Euclidean distance is less than the similarity threshold, it indicates that there is a flash discharge phenomenon in the candidate frame. When the Euclidean distance is greater than the similarity threshold, it indicates that there is no flash discharge phenomenon in the candidate frame.

[0055] Step S43, convert the candidate frame and the atmospheric flash discharge image into local binary images, and use the Hamming distance to calculate the number of different pixel values ​​at the same position of the two local binary images. The larger the Hamming distance, the more different pixel values ​​of the two local binary images at corresponding positions. If the number of different pixel values ​​is less than or equal to a preset threshold, it means that there is atmospheric flash discharge in the candidate frame, which is recorded as an abnormal phenomenon. If the number of different pixel values ​​is greater than the preset threshold, it means that there is no atmospheric flash discharge in the candidate frame, which is recorded as a lightning frame and saved to the cache frame data set.

[0056] In a second aspect, a computer device is provided, the computer device comprising:

[0057] A memory storing a computer program;

[0058] A processor is communicatively connected to the memory, and when the computer program is executed by the processor, a flash discharge monitoring method for a high-voltage transmission network is implemented.

[0059] In a third aspect, a computer-readable storage medium stores a computer program, which, when executed by a processor, implements a flash discharge monitoring method for a high-voltage transmission network.

[0060] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0061] The present invention combines the monitored highlight phenomenon, uses a monitoring camera to acquire video data of a high-voltage transmission network, analyzes the monitoring video stream saved by the monitoring camera, obtains a monitoring image data set of the high-voltage transmission network through analysis, and then exports the monitoring image data set, divides the sky and ground images in the monitoring image data set into sky images and ground images, adds a black background to the divided sky images to obtain candidate frames, compares the candidate frames with cache frames in the cache frame data set for histogram similarity, outputs lightning frames after filtering abnormal conditions, and sends the output lightning frames to a designated display terminal for result display. The present invention realizes accurate monitoring of lightning discharge phenomena in high-voltage transmission networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0063] Figure 1 It is the overall system block diagram of the present invention.

[0064] Figure 2 It is a schematic diagram of the local median in the present invention.

[0065] Figure 3 It is a schematic diagram of the structure of the computer device in the present invention. DETAILED DESCRIPTION

[0066] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than 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.

[0067] Example 1: Please refer to Figure 1-Figure 2 As shown, the technical solution provided by the present invention is: a flash discharge monitoring method for a high-voltage transmission network, the method is specifically as follows:

[0068] Step S1, based on the monitored highlight phenomenon, using a surveillance camera to acquire video data of the high-voltage transmission network;

[0069] In this embodiment, step S1 includes the following sub-steps:

[0070] Step S11, using a high-sensitivity photosensor and synchronizing the time of the photosensor with that of the surveillance camera so that the time point in the video stream corresponds to the time point of the photosensor;

[0071] Step S12, setting an alarm threshold of the photosensor, when the photosensor receives a highlight reading exceeding the alarm threshold at a certain time, it is determined that a highlight phenomenon occurs and the current time is recorded as the end time, and the second before the end time is recorded as the start time;

[0072] Step S13, intercepting a section from the video stream as a surveillance video stream of the surveillance camera according to the start time and the end time;

[0073] Step S2, analyzing the monitoring video stream saved by the monitoring camera to obtain a monitoring image data set of the high-voltage transmission network;

[0074] The step S2 includes the following sub-steps:

[0075] Step S21, obtaining monitoring image frames corresponding to the start time and the end time of the monitoring video stream respectively through decoding technology, wherein the monitoring image frame corresponding to the end time is defined as the current monitoring image frame, and the monitoring image frame corresponding to the start time is defined as the previous monitoring image frame;

[0076] It should be explained that this embodiment only analyzes the monitoring image frame corresponding to the start time and the monitoring image frame corresponding to the end time; the existing decoding technologies include OpenCV technology, FFmpeg technology, etc. In this embodiment, OpenCV technology is preferably used to decode the real-time video stream. OpenCV is an open source computer vision library that focuses on real-time image processing and computer vision tasks;

[0077] Step S22, adjusting the resolution of the decoded monitoring image frame to 256×256, and comparing the difference values ​​of adjacent monitoring image frames;

[0078] The difference value is expressed as the difference between the current monitoring image frame and the previous monitoring image frame. In this embodiment, the difference is calculated by three feature differences: pixel color information difference, probability distribution difference, and local grayscale difference. The pixel color information difference refers to the pixel difference between the pixels corresponding to the same coordinate point in the two monitoring image frames. The probability distribution difference refers to the difference between the normalized probability distribution matrices of the two monitoring image frames, and determines the difference between the two probability matrices in the probability space. The local grayscale difference refers to comparing the number of different pixel values ​​in the corresponding areas of the two monitoring image frames in a group of image areas.

[0079] The calculation steps of the above difference value are as follows:

[0080] Step S221, calculating the difference in pixel color information between the current monitoring image frame and the previous monitoring image frame. The specific analysis process of the pixel color information difference is as follows:

[0081] Step S2211, establish a coordinate system with the upper left corner vertices of the two monitoring image frames as the origin, and the two sides of the monitoring image frames as the x-axis and y-axis respectively, the value range of x and y is 0 to 255, and obtain the RGB value of each coordinate point in the coordinate system, and the RGB value is composed of R value, G value and B value;

[0082] Step S2212, calculating the difference between the RGB values ​​corresponding to the same coordinate point of the two monitoring image frames to obtain the RGB difference at the same coordinate point;

[0083] Step S2213, if the RGB difference of the same coordinate point in the two monitoring image frames is greater than or equal to the preset RGB difference, the corresponding coordinate point is recorded as an abnormal coordinate point; if the RGB difference of the same coordinate point in the two monitoring image frames is less than the preset RGB difference, no operation is performed;

[0084] Step S2214, counting the number of abnormal coordinate points and recording it as the number of abnormal coordinate points, when the number of abnormal coordinate points is greater than or equal to the preset number of abnormal coordinate points, the difference value is recorded as 1, and when the number of abnormal coordinate points is greater than or equal to the preset number of abnormal coordinate points, the difference value is recorded as 0;

[0085] Step S222, using the gray level co-occurrence matrix to calculate the probability distribution difference between adjacent monitoring image frames, the analysis process of the probability distribution difference is as follows:

[0086] Select a pixel point in the current monitoring image frame and record it as the initial pixel point i (xi, yi), where i is the number of the initial pixel point, obtain the offset of the initial pixel point (±a, ±b), obtain another pixel point in the current monitoring image frame according to the offset, and record the other pixel point as the offset pixel point (xi±a, yi±b), the initial pixel point and its corresponding offset pixel point respectively form a plurality of groups of grayscale pairs Miu (g1i, g2i), u represents the number of the grayscale pair corresponding to the initial pixel point i, wherein g1i is the grayscale value of the initial pixel point (xi, yi), g2i is the grayscale value of the offset pixel point (xi+a, yi+b), count the number of occurrences of each group of grayscale pairs Miu (g1i, g2i), divide the number of occurrences of each group of grayscale pairs Miu (g1i, g2i) by the total number of all grayscale pairs, and obtain the normalized probability matrix P1 (g1i, g2i) of the current monitoring image frame;

[0087] Similarly, the normalized probability matrix P2(g1i, g2i) of the previous surveillance image frame is obtained according to the above method, and the Euclidean distance is used Calculate the difference measure between the probability matrices P1 and P2, where (P1(g1i, g2i)-P2(g1i, g2i))² represents the square of the probability difference between the corresponding positions of the two matrices. The larger the Euclidean distance between the probability matrices P1 and P2, the greater the difference between the two matrices.

[0088] When the Euclidean distance is greater than or equal to the preset distance, the difference value is recorded as 1, and when the Euclidean distance is less than the preset distance, the difference value is recorded as 0;

[0089] It should be explained that obtaining the grayscale value usually requires weighted averaging of the values ​​of the three RGB channels to obtain a single grayscale value, which is usually calculated using the floating point method. The formula is: grayscale value = 0.299×R+0.587×G+0.114×B;

[0090] Step S223, calculating the local grayscale difference between the current monitoring image frame and the previous monitoring image frame, the steps for calculating the local grayscale difference are as follows:

[0091] According to the coordinate system, the monitoring image frame is evenly divided into 16×16 small areas, each of which contains 256 pixels. The local median is used as the local median threshold. For each pixel point (x, y), if its grayscale value is greater than or equal to the local median threshold, the pixel value is set to 1 (white). If its grayscale value is less than the local median threshold, the pixel value is set to 0 (black). Repeat the above pixel binarization process. After each 16×16 small area is processed above, a new 16×16 local binary image block is formed, in which each pixel value is 0 or 1. The local binary image block is copied to the corresponding position in the corresponding coordinate system of the current monitoring image frame to form a The final local binary image. Similarly, the final local binary image of the previous monitoring image frame is obtained according to the above method. The difference degree of local information between the two monitoring image frames is calculated by the Hamming distance. For each pixel point (x, y), if the pixel values ​​of the two local binary images at the pixel point position are different, the count of the Hamming distance is increased by 1. If the pixel values ​​of the two local binary images at the pixel point position are the same, the count of the Hamming distance remains unchanged. The difference counts of all different pixels are summed to obtain the Hamming distance. When the Hamming distance is greater than or equal to the preset difference distance, the difference value is recorded as 1. When the Hamming distance is less than the preset difference distance, the difference value is recorded as 0.

[0092] It should be explained that the local median diagram is shown in Figure 2 , the local median means that for each small area, the grayscale values ​​of all pixels are calculated, and the median is taken after arranging them in order from small to large or from large to small;

[0093] The Hamming distance is used to compare the difference between two local binary images. Given two local binary images A and B, the Hamming distance represents the number of different pixels at the same position of the two local binary images. The Hamming distance formula is: , where m represents the total number of pixels in the image, Ai and Bi represent the pixel values ​​of images A and B at the i-th pixel position, respectively, and δ(Ai, Bi) represents the indicator function, which is 1 when Ai≠Bi, otherwise it is 0;

[0094] Step S224, summing up the difference values ​​corresponding to the feature differences to obtain a total difference value. If the total difference value is greater than or equal to the difference threshold, the current monitoring image frame is recorded as a key frame. If the total difference value is less than the difference threshold, no operation is performed.

[0095] Step S23, establishing a mapping relationship between the key frame and the previous monitoring image frame and summarizing them into a monitoring image data set;

[0096] Step S3, exporting a surveillance image data set, segmenting the sky-ground image in the surveillance image data set into a sky image and a ground image, and adding a black background to the segmented sky image to obtain a candidate frame;

[0097] In step S3, the public sky-ground image dataset is annotated using the via annotation tool, the annotated sky-ground image dataset is preprocessed, the NestedUNET model is trained and tuned using the preprocessed sky-ground image dataset, and the sky and ground of the monitoring image frames in the monitoring image dataset are segmented using the trained NestedUNET model;

[0098] The step S3 includes the following sub-steps:

[0099] Step S31, collecting a publicly available sky-ground image dataset, and using a via annotation tool to annotate pictures in the publicly available sky-ground image dataset, wherein the annotated objects include: the outline of the sky area and the outline of the ground area in the picture;

[0100] It should be explained that the existing sky-ground image datasets include the MIT Places dataset, the SUN dataset, the DeepGlobe Challenge Dataset dataset, etc. In this embodiment, the DeepGlobe Challenge Dataset dataset is preferred, which contains a large number of sky-ground images and is suitable for the sky-ground image segmentation task;

[0101] Step S32, preprocessing the labeled sky-ground image dataset, adjusting the image resolution in the sky-ground image dataset to 256×256, and normalizing the pixel value of each pixel in the sky-ground image to 0-1;

[0102] It should be explained that normalization refers to scaling the data so that the scaled data falls within a preset range, which in this embodiment is between 0 and 1;

[0103] Step S33, using the NestedUNET model as the image segmentation algorithm, and training the NestedUNET model in combination with a binary cross entropy loss function and a stochastic gradient descent algorithm;

[0104] The binary cross entropy loss function is used to evaluate the prediction performance of the model on the binary classification task. At the same time, the model uses the stochastic gradient descent algorithm to save memory and computing resources;

[0105] Among them, the training process of the NestedUNET model is as follows:

[0106] Step S331, select the binary cross entropy loss function as the loss function L of the NestedUNET model. The specific formula is as follows:

[0107] , where z represents the category of the pixel in the sky-ground image. If z=1, it means the pixel is in the sky category. If z=0, it means the pixel is in the ground category. Represents the predicted probability of the pixel category by the NestedUNET model;

[0108] Step S332, using the stochastic gradient descent algorithm to optimize the NestedUNET model. The stochastic gradient descent algorithm uses only one randomly selected mini-batch to calculate the gradient in each iteration, and uses the gradient and learning rate to update the model parameters.

[0109] It should be explained that the parameter update rule is: θ←θ-η×∂L / ∂θ, where θ is a parameter, representing the weight of the NestedUNET model, η is the learning rate, representing the amplitude of parameter adjustment in each iteration of the NestedUNET model, ∂L / ∂θ is the gradient of the loss function L with respect to the parameter θ, representing the rate of change of the loss function L along the parameter θ at the current weight;

[0110] Mini-batch refers to a small set of data samples used in a training in deep learning. The entire training data set is divided into several small batches, each of which contains a fixed number of image sample data;

[0111] Step S333, dividing the processed sky-ground image data set into a training set, a test set, and a validation set according to a ratio of 7:2:1 and inputting the data into a NestedUNET model for training; training the NestedUNET model;

[0112] Step S34, using the trained NestedUNET model to segment the key frame in the summarized monitoring data set and the previous monitoring image frame into sky images and ground images, retaining the sky image and adding a completely black background, creating a completely black background image with a resolution of 256×256, traversing the pixels of each pixel in the sky image after the segmentation, copying the pixels of all pixels in the sky image to the corresponding positions of the completely black background image, and marking the sky image corresponding to the key frame as a candidate frame after adding the completely black background;

[0113] Step S4, comparing the candidate frame with the cached frames in the cached frame data set by histogram similarity, filtering out abnormalities and outputting a lightning frame;

[0114] It should be explained that the cached frame dataset is a collection of lightning frames that have been determined to be flash discharges in the high-voltage transmission network;

[0115] The lightning frame determination process in step S4 includes the following sub-steps:

[0116] Step S41, the candidate frame is converted into a grayscale image, and a grayscale histogram is constructed. The grayscale histogram construction process is as follows:

[0117] Step S411, creating an array H with a length of 256 to store the pixel count of each gray value, initialized to 0;

[0118] Step S412, traverse the grayscale value of each pixel in the candidate frame, and update the array H according to the grayscale value. For example, for a grayscale value of 50, increase the value of H

[50] ; for a grayscale value of 100, increase the value of H

[100] to obtain the number of occurrences of the grayscale value 50 and the number of occurrences of the grayscale value 100.

[0119] Step S413, establish a rectangular coordinate system, where the x-axis is the grayscale value, ranging from 0 to 255, and the y-axis is the number of pixels of each grayscale value, and construct a grayscale histogram based on the updated array H. Similarly, obtain the grayscale histogram of the cache frame;

[0120] Step S42, comparing the candidate frame with the cached frame by histogram similarity. When the Euclidean distance is less than the similarity threshold, it indicates that there is a flash discharge phenomenon in the candidate frame. When the Euclidean distance is greater than the similarity threshold, it indicates that there is no flash discharge phenomenon in the candidate frame.

[0121] Step S43, converting the candidate frame and the lightning discharge image in the atmosphere into local binary images, using the Hamming distance to calculate the number of different pixel values ​​at the same position of the two local binary images, the larger the Hamming distance, the more different pixel values ​​at the corresponding positions of the two local binary images, if the number of different pixel values ​​is less than or equal to the preset number threshold, it means that there is lightning discharge in the atmosphere in the candidate frame, which is recorded as an abnormal phenomenon, if the number of different pixel values ​​is greater than the preset number threshold, it means that there is no lightning discharge in the atmosphere in the candidate frame, which is recorded as a lightning frame and saved to the cache frame data set;

[0122] Step S5, sending the output lightning frame to a designated display terminal for result display;

[0123] In this application, if corresponding calculation formulas appear, the above calculation formulas are all dimensionless and take their numerical calculations. The weight coefficients, proportional coefficients and other coefficients in the formulas are set to a result value obtained by quantifying each parameter. The size of the weight coefficient and the proportional coefficient can be determined as long as it does not affect the proportional relationship between the parameter and the result value.

[0124] Embodiment 2: Figure 3 The present invention is a structural diagram of a computer device, which may include: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. The processor may call the logic instructions in the memory to execute a flash discharge monitoring method for a high-voltage transmission network, which includes: based on the monitored highlight phenomenon, using a monitoring camera to obtain video data of the high-voltage transmission network; analyzing the monitoring video stream saved by the monitoring camera to obtain a monitoring image data set of the high-voltage transmission network; exporting the monitoring image data set, segmenting the sky and ground images in the monitoring image data set into sky images and ground images, and adding a black background to the segmented sky images to obtain candidate frames; comparing the candidate frames with the cached frames in the cached frame data set for histogram similarity, and outputting lightning frames after filtering abnormal conditions; and sending the output lightning frames to a designated display terminal for result display.

[0125] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0126] Embodiment 3: The present application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute a flash discharge monitoring method for a high-voltage transmission network provided by the above-mentioned methods, the method including: based on the monitored highlight phenomenon, using a monitoring camera to acquire video data of the high-voltage transmission network; analyzing the monitoring video stream saved by the monitoring camera to obtain a monitoring image data set of the high-voltage transmission network; exporting the monitoring image data set, dividing the sky and ground images in the monitoring image data set into sky images and ground images, and adding a black background to the divided sky images to obtain candidate frames; comparing the candidate frames with the cached frames in the cached frame data set for histogram similarity, and outputting lightning frames after filtering abnormal conditions; sending the output lightning frames to a designated display terminal for result display.

[0127] Embodiment 4: The present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute a method for monitoring flash discharges in a high-voltage transmission network provided above, the method comprising: based on the monitored highlight phenomenon, using a surveillance camera to acquire video data of the high-voltage transmission network; analyzing the surveillance video stream saved by the surveillance camera to obtain a surveillance image data set of the high-voltage transmission network; exporting the surveillance image data set, segmenting the sky and ground images in the surveillance image data set into sky images and ground images, and adding a black background to the segmented sky images to obtain candidate frames; performing a histogram similarity comparison between the candidate frames and the cached frames in the cached frame data set, and outputting lightning frames after filtering abnormal conditions; and sending the output lightning frames to a designated display terminal for result display.

[0128] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0129] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A flash discharge monitoring method for a high voltage transmission network, characterized in that: The specific method is as follows: Step S1, based on the monitored highlight phenomenon, using a surveillance camera to acquire video data of the high-voltage transmission network; Step S2, analyzing the monitoring video stream saved by the monitoring camera to obtain a monitoring image data set of the high-voltage transmission network; Step S3, exporting a surveillance image data set, segmenting the sky-ground image in the surveillance image data set into a sky image and a ground image, and adding a black background to the segmented sky image to obtain a candidate frame; Step S4, comparing the candidate frame with the cached frames in the cached frame data set by histogram similarity, filtering out abnormalities and outputting a lightning frame; The step S4 includes the following sub-steps: Step S41, the candidate frame is converted into a grayscale image, and a grayscale histogram is constructed. The grayscale histogram construction process is as follows: Step S411, creating an array H with a length of 256 to store the pixel count of each gray value, initialized to 0; Step S412, traverse the grayscale value of each pixel in the candidate frame, and update the array H according to the grayscale value; Step S413, establish a rectangular coordinate system, where the x-axis is the grayscale value, ranging from 0 to 255, and the y-axis is the number of pixels of each grayscale value, and construct a grayscale histogram based on the updated array H. Similarly, obtain the grayscale histogram of the cache frame; Step S42, comparing the candidate frame with the cached frame by histogram similarity. When the Euclidean distance is less than the similarity threshold, it indicates that there is a flash discharge phenomenon in the candidate frame. When the Euclidean distance is greater than the similarity threshold, it indicates that there is no flash discharge phenomenon in the candidate frame. Step S43, converting the candidate frame and the lightning discharge image in the atmosphere into local binary images, using the Hamming distance to calculate the number of different pixel values ​​at the same position of the two local binary images, the larger the Hamming distance, the more different pixel values ​​at the corresponding positions of the two local binary images, if the number of different pixel values ​​is less than or equal to the preset number threshold, it means that there is lightning discharge in the atmosphere in the candidate frame, which is recorded as an abnormal phenomenon, if the number of different pixel values ​​is greater than the preset number threshold, it means that there is no lightning discharge in the atmosphere in the candidate frame, which is recorded as a lightning frame and saved to the cache frame data set; Step S5, sending the output lightning frame to the designated display terminal for result display.

2. A flash discharge monitoring method for a high voltage transmission network according to claim 1, characterized in that: The step S1 includes the following sub-steps: Step S11, using a high-sensitivity photosensor and synchronizing the time of the photosensor with that of the surveillance camera so that the time point in the video stream corresponds to the time point of the photosensor; Step S12, setting an alarm threshold of the photosensor, when the photosensor receives a highlight reading exceeding the alarm threshold at a certain time, it is determined that a highlight phenomenon occurs and the current time is recorded as the end time, and the second before the end time is recorded as the start time; Step S13, extracting a section from the video stream as the monitoring video stream of the monitoring camera according to the start time and the end time.

3. A flash discharge monitoring method for a high voltage transmission network according to claim 1, characterized in that: The step S2 includes the following sub-steps: Step S21, respectively obtaining monitoring image frames corresponding to the start time and the end time of the monitoring video stream, wherein the monitoring image frame corresponding to the end time is defined as the current monitoring image frame, and the monitoring image frame corresponding to the start time is defined as the previous monitoring image frame; Step S22, adjusting the resolution of the decoded monitoring image frame to 256×256, and comparing the difference values ​​of adjacent monitoring image frames; Step S23, establishing a mapping relationship between the key frame and the previous monitoring image frame and summarizing them into a monitoring image data set.

4. A flash discharge monitoring method for a high voltage transmission network according to claim 3, characterized in that: The steps for calculating the difference value are as follows: Calculate the pixel color information difference between the current monitoring image frame and the previous monitoring image frame. The analysis process of the pixel color information difference is as follows: A coordinate system is established with the upper left corner vertices of the two surveillance image frames as the origin, and the two sides of the surveillance image frames are used as the x-axis and y-axis respectively. The value range of x and y is 0 to 255. The RGB value of each coordinate point in the coordinate system is obtained. The RGB value consists of R value, G value and B value; Calculate the difference between the RGB values ​​of the two surveillance image frames at the same coordinate point to obtain the RGB difference at the same coordinate point; If the RGB difference of the same coordinate point in the two surveillance image frames is greater than or equal to the preset RGB difference, the corresponding coordinate point is recorded as an abnormal coordinate point; if the RGB difference of the same coordinate point in the two surveillance image frames is less than the preset RGB difference, no operation is performed; The number of abnormal coordinate points is counted and recorded as the number of abnormal coordinate points. When the number of abnormal coordinate points is greater than or equal to the preset number of abnormal coordinate points, the difference value is recorded as 1; when the number of abnormal coordinate points is greater than or equal to the preset number of abnormal coordinate points, the difference value is recorded as 0.

5. A flash discharge monitoring method for a high voltage transmission network according to claim 4, characterized in that: The step of calculating the difference value also includes: The gray-level co-occurrence matrix is ​​used to calculate the probability distribution difference between adjacent surveillance image frames. The analysis process of the probability distribution difference is as follows: Select any pixel in the current monitoring image frame and record the pixel as the initial pixel i (xi, yi), where i is the number of the initial pixel; Get the offset of the initial pixel (±a, ±b), get another pixel in the current monitoring image frame according to the offset, record the other pixel as the offset pixel (xi±a, yi±b), the initial pixel and its corresponding offset pixel form a plurality of grayscale pairs Miu(g1i, g2i), u represents the number of the grayscale pair corresponding to the initial pixel i, where g1i is the grayscale value of the initial pixel (xi, yi), and g2i is the grayscale value of the offset pixel (xi+a, yi+b); Count the number of occurrences of each grayscale pair Miu (g1i, g2i), divide the number of occurrences of each grayscale pair Miu (g1i, g2i) by the total number of all grayscale pairs, and obtain the normalized probability matrix P1 (g1i, g2i) of the current monitoring image frame; Similarly, the normalized probability matrix P2(g1i, g2i) of the previous surveillance image frame is obtained according to the above method, and the Euclidean distance is used Calculate the difference measure between the probability matrices P1 and P2, where (P1(g1i, g2i)-P2(g1i, g2i))² represents the square of the difference in the probabilities of the corresponding positions in the two matrices; When the Euclidean distance is greater than or equal to the preset distance, the difference value is recorded as 1, and when the Euclidean distance is less than the preset distance, the difference value is recorded as 0.

6. A flash discharge monitoring method for a high voltage transmission network according to claim 5, characterized in that: The step of calculating the difference value also includes: Calculate the local grayscale difference between the current surveillance image frame and the previous surveillance image frame. The specific steps for calculating the local grayscale difference are as follows: The surveillance image frame is evenly divided into 16×16 small areas according to the coordinate system, and each small area contains 256 pixels; Use the local median as the local median threshold. For each pixel (x, y), if its gray value is greater than or equal to the local median threshold, the pixel value is set to 1. If its gray value is less than the local median threshold, the pixel value is set to 0. Repeat the pixel binarization process. After each 16×16 small area is processed, a new 16×16 local binary image block is formed, in which each pixel value is 0 or 1. The local binary image block is copied to the corresponding position in the coordinate system corresponding to the current monitoring image frame to form a local binary image. Similarly, the local binary image of the previous monitoring image frame is obtained according to the above method; The difference degree of local information between two monitoring image frames is calculated by Hamming distance. If the pixel values ​​of the two local binary images at the pixel position are different, the count of Hamming distance increases by 1. If the pixel values ​​of the two local binary images at the pixel position are the same, the count of Hamming distance remains unchanged. The difference counts of all different pixels are summed to obtain the Hamming distance. When the Hamming distance is greater than or equal to the preset difference distance, the difference value is recorded as 1, and when the Hamming distance is less than the preset difference distance, the difference value is recorded as 0; The difference values ​​corresponding to the feature differences are added together to obtain the total difference value. If the total difference value is greater than or equal to the difference threshold, the current monitoring image frame is recorded as a key frame. If the total difference value is less than the difference threshold, no operation is performed.

7. A flash discharge monitoring method for a high voltage transmission network according to claim 1, characterized in that: The step S3 includes the following sub-steps: Step S31, collecting publicly available sky-ground image datasets, and labeling pictures in the publicly available sky-ground image datasets; Step S32, preprocessing the labeled sky-ground image dataset, adjusting the image resolution in the sky-ground image dataset to 256×256, and normalizing the pixel value of each pixel in the sky-ground image to 0-1; Step S33, using the NestedUNET model as an image segmentation algorithm, and training the NestedUNET model; Step S34, use the trained NestedUNET model to segment the key frames in the summarized monitoring data set and the previous monitoring image frame into sky images and ground images, retain the sky image and add a completely black background, create a completely black background image, the resolution of the completely black background image is 256×256, traverse the pixels of each pixel in the sky image after the segmentation is completed, copy the pixels of all pixels of all sky images to the corresponding positions of the completely black background image, and mark the sky image corresponding to the key frame as a candidate frame after adding the completely black background.

8. A computer device, characterized in that: The computer device comprises: A memory storing a computer program; A processor is communicatively connected to the memory, and when the computer program is executed by the processor, the method described in any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images