Transmission line protection zone identification device and cloud-edge collaborative hidden danger monitoring system

Abstract

This application provides a transmission line protection zone identification device and a cloud-edge collaborative hazard monitoring system. The system first inputs the transmission line image into a tower detection model to automatically identify the towers in the image. Then, a conductor trajectory detection module extracts the conductor trajectory based on edge detection methods, and an endpoint coordinate determination module locates the outermost conductor. Finally, combining the identified tower locations and the outermost conductor information, the protection zone determination module automatically calculates the corresponding protection zone range for the transmission line, thus avoiding the tedious process of manually marking protection zones in traditional methods. Secondly, this application identifies hazards near the protection zone and compares the area of ​​the identified hazard area with the area of ​​the protection zone to automatically determine whether the hazard area affects the transmission line, thereby improving the accuracy of hazard alarms. Simultaneously, it eliminates the need to install hazard detection equipment near the transmission line, reducing equipment and maintenance costs.

Description

[0001] The application is a divisional application, the parent application is a Chinese patent application No. 202210870007.3, filed on July 22, 2022, entitled “Power transmission line protection area hidden danger identification method, device and medium” and filed with the China Patent Office. TECHNICAL FIELD

[0002] The present application relates to the technical field of image processing, in particular to a power transmission line protection area identification device and a cloud-edge collaborative hidden danger monitoring system. BACKGROUND

[0003] In the power transmission scenario, the high-voltage line is generally used. Once the safety hidden danger such as construction machinery enters the area and collides with the power transmission line, it will cause unpredictable disaster. Therefore, in order to ensure the stable operation of the power transmission line, it is necessary to identify and alarm the hidden danger of the power transmission line in advance.

[0004] In the prior art, it is usually necessary to manually calibrate the protection area corresponding to the power transmission line, which consumes a large amount of manpower, material resources and time, especially in large-scale power transmission networks, the implementation and maintenance cost is significantly increased; and the calibration accuracy of manual calibration depends on experience, which has poor consistency and affects the accuracy of subsequent hidden danger monitoring. SUMMARY

[0005] The embodiments of the present application provide a power transmission line protection area identification device and a cloud-edge collaborative hidden danger monitoring system, which are used to solve the problems of high cost and large error in identifying the power transmission line protection area and hidden danger in the prior art.

[0006] In a first aspect, the embodiments of the present application provide a power transmission line protection area identification device, comprising:

[0007] A tower detection module is configured to input a power transmission line image into a deep learning tower detection network model based on a feature pyramid, to obtain and output a tower in the power transmission line image for bounding box labeling.

[0008] A conductor track extraction module is configured to perform edge detection on the power transmission line image to determine a plurality of conductor tracks in the power transmission line image.

[0009] An endpoint coordinate determination module is configured to determine first and second endpoint coordinates corresponding to the leftmost and rightmost conductor tracks based on the coordinates of the plurality of conductor tracks, wherein the first endpoint coordinate is an endpoint close to the tower bounding box, and the second endpoint coordinate is an endpoint away from the tower bounding box.

[0010] The protection zone determination module is configured to determine a first projection based on the first end point coordinate and a lower edge of the tower annotation box, determine a second projection based on the second end point coordinate and a lower edge of the power transmission line image, and determine the protection zone corresponding to the power transmission line based on the first projection and the second projection.

[0011] In a possible implementation, the conductor track extraction module includes:

[0012] The image preprocessing unit is configured to perform grayscale processing and histogram equalization processing on the power transmission line image.

[0013] The edge detection unit is configured to perform edge detection processing on the power transmission line image after the histogram equalization processing, and determine a first reference edge region.

[0014] The morphological processing unit is configured to filter the first reference edge region based on an expansion algorithm and a corrosion algorithm, and determine a second reference edge region.

[0015] The conductor recognition unit is configured to recognize a plurality of straight line segments from the second reference edge region according to a Hough transform algorithm, and each straight line segment corresponds to a conductor track.

[0016] In a possible implementation, the protection zone determination module is configured to:

[0017] determine a lower end point of the first projection as a third end point coordinate;

[0018] determine a lower end point of the second projection as a fourth end point coordinate;

[0019] connect the first end point coordinate, the second end point coordinate, the fourth end point coordinate and the third end point coordinate on the same side in sequence to form a reference protection zone corresponding to the power transmission line;

[0020] determine the protection zone corresponding to the power transmission line according to the reference protection zone.

[0021] In a possible implementation, the protection zone determination module is further configured to perform:

[0022] extend the reference protection zone to the left and right sides in the horizontal direction by a preset width or proportion respectively to obtain the final protection zone corresponding to the power transmission line.

[0023] In a possible implementation, the edge detection unit is specifically configured to perform:

[0024] perform denoising processing on the power transmission line image after the histogram equalization processing according to a median filter algorithm;

[0025] The pixel value gradient of the de-noising processed power transmission line image is calculated, and the pixel points are screened out based on a non-maximum suppression algorithm to obtain a third reference edge region of the power transmission line image.

[0026] The third reference edge is screened based on a double threshold algorithm to screen out the pixel points with a gray value less than a preset low threshold to obtain a first reference edge region.

[0027] In a possible implementation, the endpoint coordinate determination module comprises:

[0028] A coordinate point determination unit configured to determine a pixel coordinate set corresponding to each conductor track;

[0029] A first comparison unit configured to compare the horizontal coordinates of each pixel coordinate set to determine a leftmost conductor track with the smallest horizontal coordinate and a rightmost conductor track with the largest horizontal coordinate;

[0030] A second comparison unit configured to compare the vertical coordinates in the pixel coordinate set of the leftmost conductor track and the rightmost conductor track respectively to determine the first endpoint coordinate and the second endpoint coordinate.

[0031] In a second aspect, the embodiments of the present application provide a cloud-edge collaborative hidden danger monitoring system, comprising:

[0032] At least one edge monitoring device deployed at a power transmission tower site, and a cloud service platform in communication connection with the edge monitoring device;

[0033] The edge monitoring device is configured to collect power transmission line images.

[0034] The cloud service platform comprises a power transmission line protection area identification device and a hidden danger identification module as mentioned in the first aspect and / or various possible implementation manners of the first aspect.

[0035] The hidden danger identification module is configured to identify a hidden danger area in the power transmission line image, and trigger a hidden danger alarm based on the positional relationship between the hidden danger area and a protection area corresponding to the power transmission line in the power transmission line image.

[0036] In a possible implementation, the hidden danger identification module comprises:

[0037] A hidden danger identification unit configured to identify hidden dangers in the power transmission line image by using a preset hidden danger identification network model, and output a hidden danger bounding box;

[0038] An area calculation unit configured to calculate the overlapping area of the hidden danger bounding box and the protection area;

[0039] An alarm determination unit configured to issue a hidden danger alarm if the overlapping area is greater than a preset area value.

[0040] In one possible implementation, the cloud service platform further includes a model training module for building and training a deep learning pole detection model based on feature pyramids. The model training module includes:

[0041] The network construction unit is used to build a neural network model consisting of a feature extraction backbone network, a feature image pyramid network, and a transmission line tower classification and localization network.

[0042] The data input unit is used to input the data set of marked transmission line towers into the neural network model, and to extract feature images from the data set of marked transmission line towers through the feature extraction backbone network.

[0043] The feature extraction unit is used to perform top-down feature amplification processing on the feature image through the feature image pyramid network based on upsampling, so as to obtain the tower feature image pyramid corresponding to the transmission line tower dataset with marked towers.

[0044] The feature enhancement unit is used to distinguish the pole region from the background region in the pole feature image pyramid through the pole classification and localization network, and to locate the pole region in the pole feature image pyramid by labeling the pole region.

[0045] The classification and positioning unit is used to calculate the foreground score of the marked tower area and the offset of the marked tower area through the tower classification and positioning network.

[0046] The computation and verification unit is used to complete the training of the deep learning pole detection network model based on feature pyramids, provided that the foreground score and offset both meet the preset model accuracy conditions.

[0047] The transmission line protection zone identification device and cloud-edge collaborative hazard monitoring system provided in the above embodiments first input the transmission line image into the tower detection model to automatically identify the towers in the image. Then, the conductor trajectory detection module extracts the conductor trajectory based on the edge detection method, and the endpoint coordinate determination module locates the outermost conductor. Finally, combining the identified tower positions and the outermost conductor information, the protection zone determination module automatically calculates the protection zone range corresponding to the transmission line, thereby avoiding the tedious process of manually marking protection zones in traditional methods. Secondly, this embodiment identifies hazards near the protection zone and compares the area of ​​the identified hazard area with the area of ​​the protection zone to automatically identify whether the hazard area affects the transmission line, thereby improving the accuracy of hazard alarms and reducing false alarm rates. At the same time, this embodiment eliminates the need to install hazard detection equipment near the transmission line, thus reducing both equipment costs and maintenance costs. Attached Figure Description

[0048] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0049] Figure 1 A schematic diagram of the structure of the transmission line protection zone identification device provided in this application;

[0050] Figure 2 This is a schematic diagram of a transmission line provided in this application;

[0051] Figure 3 A schematic diagram of the cloud-edge collaborative hidden danger monitoring system provided in this application.

[0052] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0053] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0054] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0055] In power transmission scenarios, high-voltage lines are typically used. If construction machinery or other potentially hazardous materials enter the area and come into contact with the power lines, it could cause unpredictable disasters. Therefore, to ensure the stable operation of power transmission lines, it is necessary to identify and warn of potential hazards in advance.

[0056] In existing technologies, the protection zones corresponding to transmission lines typically need to be manually marked, which presents the following problems:

[0057] Inefficient and costly: Transmission lines often span long distances and complex terrains, and manual on-site surveys and calibrations require a lot of manpower, material resources and time. Especially in large-scale transmission networks, implementation and maintenance costs increase significantly.

[0058] The accuracy of calibration depends on experience and has poor consistency: the determination of the boundary of the protected area is affected by the experience and subjective judgment of the operators, which can easily introduce human error, resulting in inconsistent calibration results in different sections or at different times, affecting the accuracy of the protected area.

[0059] Poor adaptability and difficulty in coping with dynamic changes: Transmission lines may require adjustments to the protection zone due to renovation, relocation, or changes in the surrounding environment (such as new buildings or vegetation growth). Manual calibration methods are outdated and cannot adapt to these changes in a real-time and flexible manner, resulting in a mismatch between the protected area and the actual risk area.

[0060] Insufficient scalability: As the scale of transmission lines expands or the number of monitoring points increases, the manual calibration work grows linearly, making it difficult to quickly expand to full network coverage, which restricts the large-scale deployment of intelligent monitoring systems.

[0061] Integration with automated systems is difficult: the results of human calibration usually exist in static forms such as drawings and tables, which are difficult to integrate directly and efficiently with digital image recognition and real-time monitoring systems, and cannot support automatic hazard identification and early warning processes based on visual analysis.

[0062] Safety risks exist: manual on-site marking may involve high-altitude and field operations, posing a potential threat to personnel safety. At the same time, the marking process itself may lead to incomplete delineation of protected areas due to operational errors, leaving safety hazards.

[0063] To address the aforementioned technical issues, this application provides a transmission line protection zone identification device and a cloud-edge collaborative hazard monitoring system. The system first inputs the transmission line image into a tower detection model to automatically identify the towers in the image. Then, a conductor trajectory detection module extracts the conductor trajectory based on edge detection methods, and an endpoint coordinate determination module locates the outermost conductor. Finally, combining the identified tower locations and the outermost conductor information, the protection zone determination module automatically calculates the corresponding protection zone range for the transmission line, thus avoiding the tedious process of manually marking protection zones. Secondly, this application identifies hazards near the protection zone and compares the area of ​​the identified hazard area with the area of ​​the protection zone to automatically determine whether the hazard area affects the transmission line, thereby improving the accuracy of hazard alarms and reducing false alarm rates. Furthermore, this application eliminates the need to install hazard detection equipment near the transmission line, thus reducing both equipment costs and maintenance costs.

[0064] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0065] Figure 1 This is a schematic diagram of the structure of a transmission line protection zone identification device provided in an embodiment of this application. Figure 1 As shown, the transmission line protection zone identification device includes:

[0066] The pole detection module 101 is used to input the transmission line image into the deep learning pole detection network model based on feature pyramids to obtain and output the pole annotation boxes in the transmission line image.

[0067] The conductor trajectory extraction module 102 is used to perform edge detection on the transmission line image to determine the trajectories of multiple conductors in the transmission line image;

[0068] The endpoint coordinate determination module 103 is used to determine the first endpoint coordinates and the second endpoint coordinates corresponding to the leftmost and rightmost traverse tracks based on the coordinates of multiple traverse tracks; wherein, the first endpoint coordinates are the endpoints closer to the tower label frame, and the second endpoint coordinates are the endpoints farther away from the tower label frame;

[0069] The protected area determination module 104 is used to determine a first projection based on the coordinates of the first endpoint and the lower edge of the tower annotation box; determine a second projection based on the coordinates of the second endpoint and the lower edge of the transmission line image; and determine the protected area corresponding to the transmission line based on the first projection and the second projection.

[0070] Figure 2 A cloud-edge collaborative hidden danger monitoring system provided in this application embodiment includes:

[0071] At least one edge monitoring device 201 deployed at the site of the power transmission tower, and a cloud service platform 202 communicating with the edge monitoring device;

[0072] Edge monitoring device 201, used to acquire images of power transmission lines;

[0073] The cloud service platform 202 includes the aforementioned transmission line protection zone identification device and the hazard identification module 105;

[0074] The hazard identification module 105 is used to identify hazard areas in the transmission line image and trigger a hazard alarm based on the positional relationship between the hazard area and the protection zone corresponding to the transmission line in the transmission line image.

[0075] In one embodiment of this application, images of the transmission line can be acquired using an edge monitoring device 201. The edge monitoring device 201 is mounted on a tower with its lens facing the opposite tower. The edge monitoring device 201 can capture images of the transmission line in real time or at regular intervals, obtaining transmission line images, and then send the captured images to a cloud service platform 201.

[0076] Furthermore, the edge monitoring device 201 captures a protected corridor zone within the transmission line scene. This protected corridor zone is defined as the area between two parallel lines extending outwards from the conductor's edge. Transmission lines are typically high-voltage lines; if construction machinery or other safety hazards enter this zone and collide with the transmission line, it could cause unpredictable disasters.

[0077] Figure 3 This is a schematic diagram of a power transmission line provided as an embodiment of this application. Figure 3 As shown, the edge monitoring device 201 installed on the tower will take pictures along the direction of the power transmission line, and the image will show multiple power transmission lines.

[0078] In one embodiment of this application, the system further includes a model training module for training a deep learning pole detection network model based on feature pyramids. The model training module includes:

[0079] The network building unit is used to build a neural network model, which consists of a feature extraction backbone network, a feature image pyramid network, and a transmission line tower classification and localization network.

[0080] The data input unit is used to input the data set of marked transmission line towers into the neural network model, and to extract feature images from the data set of marked transmission line towers through the feature extraction backbone network.

[0081] The feature extraction unit is used to perform top-down feature amplification processing on the feature image through the feature image pyramid network based on upsampling, so as to obtain the tower feature image pyramid corresponding to the transmission line tower dataset with marked towers.

[0082] The feature enhancement unit is used to distinguish between pole regions and background regions in the pole image pyramid through the pole classification and localization network, and to locate the pole regions in the pole feature image pyramid by labeling boxes.

[0083] The classification and positioning unit is used to calculate the foreground score of the marked tower area through the tower classification and positioning network, and to calculate the offset of the marked tower area.

[0084] The computation and verification unit is used to complete the training of the deep learning pole detection network model based on feature pyramids, provided that the foreground score and offset both meet the preset model accuracy conditions.

[0085] Specifically, the tower detection network model involved in this application embodiment is a deep learning model of convolutional neural network, and the training of the model requires a certain amount of labeled data. Therefore, this application embodiment pre-constructs a dataset with labeled towers under transmission lines. For each tower, the minimum bounding rectangle of the tower is labeled for training the deep learning model.

[0086] Furthermore, the convolutional neural network model involved in this application embodiment has two branches: a classification branch and a bounding box regression branch. The input to the model is a color image. The classification branch is used to distinguish the foreground and background of the image, while the bounding box regression branch accurately locates the position of the tower. Since the towers are relatively obvious in this task, and the detection task is relatively simple, a lightweight detection model is selected in this paper to save inference time. During the model training phase, the tower detection model is trained using a pre-constructed dataset with tower markings under transmission lines.

[0087] Specifically, the model consists of three parts: a feature extraction backbone network based on residual modules, a feature image pyramid network that enhances feature information through path aggregation, and a transmission line tower classification and localization network.

[0088] Specifically, the feature extraction backbone network mainly consists of residual modules. These residual modules are composed of residual branches and identity branches. The residual branches primarily consist of convolutional layers, batch normalization layers, and activation layers. The convolutional layers extract feature maps through convolutional operations, the batch normalization layers normalize the convolutional outputs to stabilize model training, and the activation layers activate some features while suppressing less important ones. Furthermore, the residual module contains an identity branch, which directly passes the original input to the module's output. The output of the residual branch is then added to the result of the identity branch to obtain the residual module's output. This design significantly alleviates the gradient vanishing problem caused by excessively deep models. The feature extraction backbone network also uses pooling layers to reduce the size of feature maps and aggregate features, thus reducing computational cost and expanding the receptive field.

[0089] Specifically, the feature image pyramid network performs top-down feature amplification processing on the feature image based on upsampling. This top-down process amplifies the small feature maps at the top level to the features of the previous stage through upsampling. Figure 1The size of the sample is important. This approach leverages both the strong semantic features of the top layer and the high-resolution information of the bottom layer. Upsampling can be implemented using nearest neighbor interpolation. To combine high-level semantic features with the precise localization capabilities of the bottom layer, lateral connections fuse features from the previous layer that have been upsampled and have the same resolution as the current layer through addition, similar to the lateral connection structure of a residual network. Considering the significant loss of feature information during the top-down process, the feature pyramid network used in this embodiment adds a bottom-up process and lateral connections to the traditional feature pyramid network. This significantly improves the network's performance.

[0090] In this neural network, a pole classification and localization network is used to distinguish and classify poles from the background, and to locate the pole detection boxes. The network outputs the scores of the obtained pole detection boxes and the offsets relative to pre-defined anchor boxes. The pole detection network model is trained when both the foreground score and the offset meet the preset model accuracy conditions.

[0091] In one embodiment of this application, the conductor trajectory extraction module 102 includes:

[0092] The image preprocessing unit is used to perform grayscale conversion and histogram equalization on the transmission line image.

[0093] The edge detection unit is used to perform edge detection processing on the transmission line image after histogram equalization processing to determine the first reference conductor region.

[0094] The morphological processing unit is used to filter the first reference edge region based on the dilation algorithm and the erosion algorithm to determine the second reference edge region corresponding to the transmission line image.

[0095] The conductor identification unit is used to identify multiple straight line segments from the second reference edge region according to the Hough transform algorithm, with each straight line segment corresponding to a conductor trajectory.

[0096] Specifically, the conductor recognition unit determines the sine curves corresponding to the pixels in the second reference edge region according to the Hough transform algorithm, and determines the set of pixels corresponding to the transmission line based on the number of sine curves intersecting at the same point. Based on the set of pixels corresponding to the transmission line, it determines the trajectories of multiple conductors in the transmission line image.

[0097] Specifically, the edge detection unit is used to: denoise the histogram-equalized transmission line image using a median filtering algorithm; calculate pixel gradients in the denoised transmission line image and filter out pixels using a non-maximum suppression algorithm to obtain the third reference edge region of the transmission line image; and filter the third reference edge using a dual threshold algorithm to remove pixels with grayscale changes less than a preset low threshold to obtain the first reference conductor region.

[0098] Furthermore, in order to improve the detection speed when performing edge detection on transmission line images, this embodiment abandons the slow convolutional neural network-based conductor trajectory detection algorithm and selects a faster traditional method-based conductor detection algorithm. First, the transmission line images captured by the monitoring device are converted to grayscale. Then, an 8×8 grid is used to perform histogram equalization on the grayscale image. The working principle of histogram equalization is: first, the histogram of image pixel intensity is calculated, and the most frequent pixel values ​​(i.e., the pixel values ​​with the largest counts in the histogram) are evenly distributed. Then, the linear trend of the cumulative distribution function is given, thereby improving the image contrast.

[0099] Further, edge detection is performed on the equalized transmission line image. The specific steps are as follows: (1) Denoising: Noise will affect the accuracy of edge detection, so noise must be filtered out first. In this embodiment, the transmission line image after histogram equalization can be denoised using the median filtering algorithm. (2) Calculate the magnitude and direction of the gradient. (3) Non-maximum suppression: Pixels are screened out, that is, the edges are appropriately "thinned" to obtain the third reference edge region of the transmission line image. (4) Determine the edge: The final edge information is determined using the double threshold algorithm. The third reference edge is screened out to remove pixels with grayscale change values ​​less than the preset low threshold to obtain the first reference conductor region. Then, the complete edge detection algorithm is performed on the image. Next, the obtained binary image is dilated and then eroded. Finally, the endpoints of the conductor can be detected by the statistical probability Hough linear transformation function.

[0100] In one embodiment of this application, the endpoint coordinate determination module 103 includes:

[0101] The coordinate point determination unit is used to determine the set of pixel coordinates corresponding to multiple conductor trajectories.

[0102] The first comparison unit is used to compare the x-coordinates of each pixel coordinate set to determine the leftmost guide trajectory with the smallest x-coordinate and the rightmost guide trajectory with the largest x-coordinate.

[0103] The second comparison unit is used to compare the vertical coordinates in the pixel coordinate sets of the leftmost and rightmost guide trajectories respectively, in order to determine the coordinates of the first endpoint and the second endpoint.

[0104] Specifically, after performing edge detection on the captured images of the power transmission lines, the conductor trajectory can be obtained. Based on the obtained conductor trajectory, the set of pixel coordinates corresponding to the conductor trajectory can be obtained. The power transmission lines erected between the towers are erected in parallel, and the range corresponding to the power transmission line protection zone is related to the outermost power transmission lines on both the left and right sides. Therefore, the set of pixel coordinates can be compared. By comparing the horizontal coordinates, the set of coordinates corresponding to the outermost power transmission lines on the left and right sides can be determined.

[0105] Specifically, the set of pixel coordinates corresponding to the conductor trajectory is grouped according to different ordinates, resulting in multiple pixel coordinates for each row. The x-coordinates of these multiple pixel coordinates for each row are compared to determine the leftmost and rightmost coordinates for each row. A first reference pixel coordinate set is constructed based on the leftmost pixel in each row. A second reference pixel coordinate set is constructed based on the rightmost pixel in each row. The first reference pixel coordinate set is the set of pixels corresponding to the leftmost conductor, and the second reference pixel coordinate set is the set of pixels corresponding to the rightmost conductor.

[0106] Furthermore, the coordinates of the points in the first set of reference pixel coordinates are compared by their ordinates to determine the coordinates of the two endpoints corresponding to the leftmost conductor. Similarly, the coordinates of the points in the second set of reference pixel coordinates are compared by their ordinates to determine the coordinates of the two endpoints corresponding to the rightmost conductor. For example, a coordinate system is established with the straight line containing the left edge of the transmission line image as the Y-axis, the straight line containing the lower edge as the X-axis, and the intersection of the left and lower edges as the origin. In this case, the coordinates with the largest and smallest ordinate values ​​in the first set of reference pixel coordinates are the two endpoints corresponding to the leftmost conductor. Similarly, the coordinates with the largest and smallest ordinate values ​​in the second set of reference pixel coordinates are the two endpoints corresponding to the rightmost conductor.

[0107] In one embodiment of this application, the protected area determination module 104 is specifically used to: determine the coordinates of the first endpoint and the second endpoint of the two outermost conductor trajectories based on the pixel coordinates corresponding to the conductor trajectories in the transmission line image, wherein the first endpoint is the conductor endpoint closer to the tower, and the second endpoint is the conductor endpoint farther from the tower. A first projection is determined based on the first endpoints of the two outermost trajectories and the lower edge of the tower's annotation frame. A second projection is determined based on the coordinates of the second endpoints of the two outermost trajectories and the lower edge of the transmission line image. The protected area corresponding to the transmission line is determined based on the first and second projections.

[0108] Specifically, the edge monitoring device 201 captures images of the transmission line, obtaining an image of the transmission line. This image is then input into a tower detection network. The network model outputs the tower with the highest score, which is the tower location in the transmission line scene. The lower edge of the tower is the ground position. Next, conductor detection is performed on the original image to obtain the conductor trajectory in the image. The outermost conductor trajectories on the left and right sides are obtained through coordinate information, and the endpoints corresponding to the two outermost conductor trajectories on the left and right sides are determined. There are two conductor endpoints: a first endpoint closer to the tower and a second endpoint farther from the tower. The points at both ends of the conductor are projected downwards. The point closer to the tower is projected onto the horizontal line of the ground, obtaining the first projection. The point farther from the tower is projected onto the lower boundary of the image, obtaining the second projection.

[0109] In one embodiment of this application, the protected area determination module 104 is further configured to:

[0110] Determine the lower endpoint of the first projection as the coordinates of the third endpoint;

[0111] Determine the lower endpoint of the second projection as the coordinates of the fourth endpoint;

[0112] Connect the coordinates of the first, second, fourth, and third endpoints on the same side in sequence to form a reference protection zone corresponding to the transmission line.

[0113] Specifically, the coordinates of the third endpoint are obtained based on the intersection of the first projection and the lower edge of the tower annotation frame. The coordinates of the fourth endpoint are obtained based on the intersection of the second projection and the lower edge of the transmission line image. By sequentially connecting the adjacent first, second, fourth, and third endpoints, a cuboid-shaped reference protection zone can be obtained. Next, the reference protection zone is expanded horizontally to the left and right by a predetermined width or proportion to obtain the final protection zone corresponding to the transmission line. For example, the reference protection zone can be expanded outward by 20%, and the area formed in this way is the protection zone corresponding to the transmission line.

[0114] In one embodiment of this application, the hazard identification module 105 includes:

[0115] The hazard identification unit is used to identify hazards in transmission line images through a pre-set hazard identification network model and output hazard annotation boxes.

[0116] The area calculation unit is used to calculate the overlapping area between the hazard marking frame and the protected area.

[0117] The alarm determination unit is used to issue a hazard warning when the overlapping area is greater than a preset area value.

[0118] Specifically, in this embodiment, image samples of transmission lines marked with potential hazard areas are pre-collected. These samples are then used to train a pre-set neural network model to obtain the pre-set hazard identification network model. The current transmission line image captured by the edge monitoring device 201 is input into the pre-set hazard identification network model, which then marks the potential hazard areas in the image. The size of the marked hazard areas is then determined.

[0119] Furthermore, the set of pixel coordinates for the hazardous area and the set of pixel coordinates for the protected area are determined. Comparing these two sets of coordinates yields the overlapping pixels. Based on these overlapping pixels, the area of ​​the overlapping region between the hazardous area and the protected area is determined; that is, the overlapping area of ​​the hazardous equipment entering the protected area. If this overlapping area is greater than 50% of the hazardous area area, a warning is issued.

[0120] It should be noted that, in this embodiment, a warning is preferably issued when the overlapping area is greater than 50% of the area of ​​the potential hazard zone. In practical applications, the preset area value can be adjusted according to the actual situation, and this embodiment does not impose any restrictions on this.

Claims

1. A transmission line protection zone identification device, characterized in that, include: The tower detection module is used to input the transmission line image into a deep learning tower detection network model based on feature pyramids to obtain and output the tower annotation boxes in the transmission line image. The conductor trajectory extraction module is used to perform edge detection on the transmission line image to determine the trajectories of multiple conductors in the transmission line image; The endpoint coordinate determination module is used to determine the first endpoint coordinates and the second endpoint coordinates corresponding to the leftmost and rightmost conductor trajectories, respectively, based on the coordinates of the multiple conductor trajectories; wherein, the first endpoint coordinates are the endpoints closer to the tower label frame, and the second endpoint coordinates are the endpoints farther away from the tower label frame; The protected area determination module is used to determine a first projection based on the coordinates of the first endpoint and the lower edge of the tower marking frame; determine a second projection based on the coordinates of the second endpoint and the lower edge of the transmission line image; and determine the protected area corresponding to the transmission line based on the first projection and the second projection.

2. The apparatus according to claim 1, characterized in that, The conductor trajectory extraction module includes: An image preprocessing unit is used to perform grayscale processing and histogram equalization processing on the transmission line image. The edge detection unit is used to perform edge detection processing on the power line image after histogram equalization processing to determine the first reference edge region. The morphological processing unit is used to filter the first reference edge region based on the dilation algorithm and the erosion algorithm to determine the second reference edge region; The conductor identification unit is used to identify multiple straight line segments from the second reference edge region according to the Hough transform algorithm, and each straight line segment corresponds to a conductor trajectory.

3. The apparatus according to claim 1, characterized in that, The protected area determination module is used for: The lower endpoint of the first projection is determined as the coordinates of the third endpoint; The lower endpoint of the second projection is determined as the coordinates of the fourth endpoint; The coordinates of the first endpoint, the second endpoint, the fourth endpoint, and the third endpoint, which are adjacent on the same side, are connected in sequence to form the reference protection zone corresponding to the transmission line.

4. The apparatus according to claim 3, characterized in that, The protected area determination module is also used to perform: The reference protection zone is expanded horizontally to the left and right by a predetermined width or proportion to obtain the final protection zone corresponding to the transmission line.

5. The apparatus according to claim 1, characterized in that, The endpoint coordinate determination module includes: The coordinate point determination unit is used to determine the set of pixel coordinates corresponding to the multiple conductor trajectories respectively; The first comparison unit is used to compare the abscissas of each set of pixel coordinates to determine the leftmost guide trajectory with the smallest abscissa and the rightmost guide trajectory with the largest abscissa. The second comparison unit is used to compare the vertical coordinates in the pixel coordinate sets of the leftmost and rightmost conductor trajectories respectively, so as to determine the coordinates of the first endpoint and the coordinates of the second endpoint.

6. The apparatus according to claim 2, characterized in that, The edge detection unit is specifically used to perform: Based on the median filtering algorithm, the transmission line image after histogram equalization is denoised. The pixel value gradient is calculated on the denoised transmission line image, and the pixels are filtered out based on the non-maximum suppression algorithm to obtain the third reference edge region of the transmission line image. The third reference edge is filtered based on a dual threshold algorithm to remove pixels with grayscale changes less than a preset low threshold, thus obtaining the first reference edge region.

7. A cloud-edge collaborative hidden danger monitoring system, characterized in that, include: At least one edge monitoring device deployed at the site of a power transmission tower, and a cloud service platform that is communicatively connected to the edge monitoring device; The edge monitoring device is used to acquire images of power transmission lines; The cloud service platform includes the transmission line protection zone identification device as described in any one of claims 1-6, and the hazard identification module; The hazard identification module is used to identify hazard areas in the transmission line image and trigger a hazard alarm based on the positional relationship between the hazard area and the protection zone corresponding to the transmission line.

8. The system according to claim 7, characterized in that, The hazard identification module includes: The hazard identification unit is used to identify hazards in the transmission line image through a pre-set hazard identification network model and output hazard annotation boxes; An area calculation unit is used to calculate the overlapping area between the hazard marking frame and the protected area; An alarm determination unit is used to issue a hazard alarm when the overlapping area is greater than a preset area value.

9. The system according to claim 7, characterized in that, The cloud service platform also includes a model training module, which is used to build and train a deep learning pole detection model based on feature pyramids.

Images