Power transmission line external force damage hidden danger threat analysis method and system

Abstract

The invention provides a power transmission line external force damage hidden danger threat analysis method and system, and relates to the technical field of power transmission line detection. Power transmission channel images are collected through a tower camera, an external damage hidden danger detection and power line segmentation data set is constructed, images are inferred in parallel based on a double-branch detection algorithm, pixel coordinates of an external damage hidden danger target and a power line are obtained respectively, the threat level is judged in a space intersection detection mode, and an alarm is triggered only for a high-threat image. According to the method, high-precision hidden danger identification and accurate early warning based on the pixel-level spatial relationship are realized, and the intelligent identification and response capability of the external damage hidden danger of the power transmission line is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power transmission line detection, and in particular to a method and system for analyzing the threat of hidden dangers of external force damage to a power transmission line. Background Art

[0002] With rapid economic development and increasing electricity demand, most countries are focusing on the construction of high-voltage transmission lines, resulting in long mileage and wide coverage. Transmission lines often pass through densely populated urban areas and complex wilderness environments, creating the potential for damage from illegal construction machinery, aerial work vehicles, and other external forces, leading to safety accidents.

[0003] Most existing research focuses on the application and improvement of target detection algorithms, generating alarms for all potential damage targets within the image. This technical approach has significant limitations: the transmission channel's safety protection zone is not precisely located, the threat level of potential damage is not assessed, and there is a lack of classification and screening of potential damage. This results in a large number of alarm images requiring manual review and screening. If a detected target in the image clearly does not cause external damage to the transmission channel, but is still mistakenly reported as a potential damage, it wastes human and material resources. Therefore, simply detecting potential damage targets is insufficient. Summary of the Invention

[0004] In order to overcome the shortcomings of the existing technology, the purpose of the present invention is to provide a method and system for analyzing the threat of hidden dangers of external force damage to transmission lines. In view of the problem that the existing technology lacks the classification and screening of hidden dangers according to threat, which easily produces system false alarms, resulting in a large number of alarm images that need to be manually reviewed and screened again, the present invention realizes the identification and image screening of hidden dangers of transmission lines based on threat analysis, and can alarm for images or videos with high-threat targets, effectively reducing the workload of subsequent manual review, which is conducive to improving the intelligence level of transmission line inspection.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for analyzing the threat of hidden dangers of external force damage to a transmission line, comprising:

[0007] Capture transmission channel images through cameras installed on transmission line towers;

[0008] Performing external damage hidden danger target anchor frame annotation and transmission line segmentation annotation on the transmission channel image to generate a transmission line external damage hidden danger dataset and a transmission line power line segmentation dataset respectively;

[0009] Based on the double-branch transmission line external failure hidden danger detection algorithm, training is performed according to the transmission line external failure hidden danger dataset and the transmission line power line segmentation dataset, and the weights obtained from the training are used for image parallel reasoning to obtain the category and pixel coordinates of the external failure hidden danger target and the pixel coordinates of the power line;

[0010] The pixel coordinates of the external hidden danger target and the pixel coordinates of the power line are spatially intersected by a pixel-level positioning and discrimination module. If there is an intersection, the corresponding hidden danger target is determined to be a high threat, otherwise it is determined to be a low threat;

[0011] An alarm is triggered for an image containing the high-threat hidden danger target, and the image and the label name and location information of the high-threat hidden danger in the image are output.

[0012] Preferably, based on the double-branch transmission line external failure hidden danger detection algorithm, training is performed based on the transmission line external failure hidden danger dataset and the transmission line power line segmentation dataset, and the weights obtained from the training are used for image parallel reasoning to obtain the category and pixel coordinates of the external failure hidden danger target and the pixel coordinates of the power line, including:

[0013] Inputting the transmission line external damage hidden danger data set into the hidden danger detection model to obtain the category and pixel coordinates of the external damage hidden danger target;

[0014] The power line segmentation dataset of the transmission line is input into a power line segmentation model to obtain the pixel coordinates of the power lines.

[0015] Preferably, the transmission line external damage hidden danger dataset is input into a hidden danger detection model to obtain the category and pixel coordinates of the external damage hidden danger target, including:

[0016] Anchor frames are annotated on the transmission line external damage hidden danger dataset, and the dataset is divided into a training set and a validation set in a ratio of 8:2, and input into the hidden danger detection model to start training;

[0017] During the training phase, the transmission line external damage hidden danger dataset is fed into a feature extraction network including a multi-scale multi-head convolutional gated attention mechanism (M2CGA) to obtain a multi-scale, fine-grained feature map.

[0018] The feature map is input into the neck network containing the multi-scale adaptive spatial attention gate MASAG for multi-scale fusion, spatial selection, spatial interaction, cross-regulation and recalibration to enhance the key area expression and suppress irrelevant background information;

[0019] The fused feature map output by the multi-scale adaptive spatial attention gate MASAG is fed into the FOD-DETR detection head;

[0020] In the inference stage, the detection model weights are loaded, and forward inference is performed on the transmission line external damage hidden danger data set to output the category of the external damage hidden danger target and the corresponding pixel coordinates.

[0021] Preferably, inputting the power line segmentation dataset into a power line segmentation model to obtain pixel coordinates of the power lines comprises:

[0022] Embedding the contextual anchor attention mechanism into the power line segmentation model to form a CAA-TransUNet network to enhance the contextual modeling capability of slender power lines;

[0023] Using the divided power line segmentation dataset to perform offline training on the CAA-TransUNet network to obtain segmentation model weights;

[0024] In the inference stage, the segmentation model weights are loaded, and forward inference is performed on the power line segmentation dataset of the transmission line to output the pixel coordinates of the power line.

[0025] Preferably, the categories of targets with potential external damage hazards include: pump trucks, forklifts, cranes, excavators, tower cranes, pile drivers and pole towers.

[0026] Preferably, the spatial intersection detection adopts a pixel-level intersection detection method of target frame by target frame and line segment by line segment. After the first intersection is detected, the target is determined to be a high threat and subsequent detection of the current target frame is terminated.

[0027] Preferably, it also includes:

[0028] Images containing only low-threat or no-threat potential targets are counted without triggering an alarm.

[0029] Preferably, the hidden danger detection model and the power line segmentation model obtain weights through offline training, and the inference stage runs in parallel on the same processor.

[0030] A system for analyzing the hidden dangers and threats of external force damage to power transmission lines, comprising:

[0031] A data acquisition unit, configured to acquire images of a transmission channel through a camera installed on a transmission line tower;

[0032] a data annotation unit, configured to annotate the transmission channel image with an anchor frame of an external damage hidden danger target and a transmission line segmentation annotation, thereby generating a transmission line external damage hidden danger dataset and a transmission line power line segmentation dataset respectively;

[0033] An image training and inference unit is configured to perform training based on a dual-branch transmission line external failure hazard detection algorithm according to the transmission line external failure hazard dataset and the transmission line power line segmentation dataset, and use the weights obtained from the training for image parallel inference to obtain the category and pixel coordinates of the external failure hazard target and the pixel coordinates of the power line;

[0034] a hidden danger identification unit, configured to perform spatial intersection detection on the pixel coordinates of the external breach hidden danger target and the pixel coordinates of the power line through a pixel-level positioning identification module; if there is an intersection, the corresponding hidden danger target is determined to be a high threat; otherwise, it is determined to be a low threat;

[0035] The alarm unit is used to trigger an alarm for an image containing the high-threat hidden danger target, and output the image and the label name and location information of the high-threat hidden danger in the image.

[0036] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0037] By constructing a dual-branch transmission line external failure hazard detection algorithm, this invention achieves simultaneous identification of external failure hazard targets and power lines, and determines their spatial relationship. This algorithm accurately determines, with pixel-level precision, whether a hazard target poses an actual threat to the power line, and uses this information to classify and filter images captured by the data acquisition unit. Compared to existing methods, this invention not only improves the accuracy of hazard identification and the targeted nature of alarms, effectively reducing the workload of subsequent manual review, but also possesses good engineering deployability and can be widely applied to intelligent inspection and risk warning systems for transmission lines, possessing significant practical value and promising prospects for promotion. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 A flow chart of a method provided by an embodiment of the present invention;

[0040] Figure 2 A block diagram of the technical principles provided by an embodiment of the present invention;

[0041] Figure 3 A structural diagram of the FOD-DETR model provided in an embodiment of the present invention;

[0042] Figure 4 A structural diagram of a pixel-level positioning and discrimination module provided in an embodiment of the present invention;

[0043] Figure 5 A module structure diagram provided for an embodiment of the present invention;

[0044] Figure 6 This is a visualization result diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] The purpose of the present invention is to provide a method and system for analyzing the threat of hidden dangers of external force damage to transmission lines, which realizes high-precision threat identification and accurate alarm based on pixel-level spatial intersection, and effectively improves the intelligent identification and response capabilities of hidden dangers of external force damage to transmission lines.

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] Figure 1 A flow chart of the method provided in the embodiment of the present invention is shown in FIG. Figure 1 As shown, the present invention provides a method for analyzing the threat of hidden dangers of external force damage to power transmission lines, comprising:

[0049] Step 100: Acquire a transmission channel image using a camera installed on a transmission line tower;

[0050] Step 200: Annotate the transmission channel image with external damage hidden danger target anchor frames and segment the transmission lines to generate a transmission line external damage hidden danger dataset and a transmission line power line segmentation dataset respectively;

[0051] Step 300: Based on the dual-branch transmission line external failure hazard detection algorithm, training is performed based on the transmission line external failure hazard dataset and the transmission line power line segmentation dataset. The weights obtained from the training are used for image parallel reasoning to obtain the category and pixel coordinates of the external failure hazard target and the pixel coordinates of the power line.

[0052] Step 400: The pixel coordinates of the external hidden danger target and the pixel coordinates of the power line are spatially intersected by the pixel-level positioning and discrimination module. If there is an intersection, the corresponding hidden danger target is determined to be a high threat, otherwise it is determined to be a low threat.

[0053] Step 500: trigger an alarm for an image containing a high-threat hidden danger target, and output the image and the label name and location information of the high-threat hidden danger in the image.

[0054] This embodiment provides a method for analyzing the threat of external force damage hazards to power transmission lines. It examines the fact that failure to consider the threat of hazards during external force damage detection can lead to a large number of false positive images. By determining the threat of hazards based on their relative position to the safe area of ​​the transmission corridor, and defining and screening hazards based on their threat level, the system can significantly reduce the number of false positive images and the workload for subsequent manual review.

[0055] Please see the attached Figure 2 As shown, the overall process of this embodiment includes the following steps:

[0056] S1. Image data acquisition: The transmission channel images are collected regularly through the video stream composed of cameras deployed on the transmission line towers.

[0057] Specifically, in this embodiment, a video stream composed of cameras deployed on the transmission line towers is used to collect a transmission channel image every hour, and images with hidden dangers of external damage are retained for subsequent training of the double-branch transmission line external damage hazard detection algorithm.

[0058] S2. Image annotation and dataset construction: Use image annotation tools to annotate anchor frames of external damage hazards and segment and annotate transmission lines.

[0059] Specifically, in this embodiment, the collected images were annotated to construct two datasets. The power transmission line damage hazard dataset annotates the objects in the images with anchor boxes, including seven labels: pump truck, forklift, crane, excavator, tower crane, pile driver, and tower. The power transmission line segmentation dataset segments and annotates the power lines in the images. Finally, the annotated dataset is divided into a training set and a validation set with a ratio of 8:2.

[0060] S3, a double-branch transmission line external damage hidden danger detection algorithm, uses the new hidden danger detection model FOD-DETR and the new power line segmentation model CAA-TransUNet to detect images in parallel. FOD-DETR outputs the name and coordinate information of the external damage hidden danger in the image, and CAA-TransUNet outputs the power line coordinate information in the image.

[0061] Specifically, in this embodiment, the network structure of FOD-DETR is as follows: Figure 3As shown, first, based on the RT-DETR model, improvements are made to address the multi-scale and widely varying appearance characteristics of transmission line scenarios and external damage targets. Specifically, in the feature extraction stage, transmission line image data is fed into the model, where initial features are extracted through convolution and block modules. Three layers of M2CGA are introduced after each of the three block modules. The multi-level feature maps from the block modules are fed into a multi-scale self-attention module, which extracts fine-grained local features within the neighborhood of the query location through 3×3 depthwise separable dilated convolutions. Adaptive average pooling is also used for multi-scale downsampling to enhance context modeling capabilities and reduce computational overhead. M2CGA integrates multi-scale mechanisms, channel attention, and multi-head self-attention. Each token has a unique gating signal based on its nearest fine-grained features, addressing the issues of excessive coarse-grained variance and lack of positional information in global average pooling. The output of the multi-scale self-attention module is then batch normalized and fed into a convolutional gated linear unit. The convolutional gated linear unit first expands the channel dimension through 1×1 convolution to preserve the original semantics, then reduces the dimension through another set of 1×1 convolutions, and combines 3×3 depthwise separable convolutions to extract local neighborhood information. Dynamic channel attention weights are generated through a gating mechanism to achieve fine-grained feature enhancement. Finally, the three-layer M2CGA jump connection is connected to the three-layer MASAG for feature fusion.

[0062] Secondly, during the feature fusion stage, MASAG is introduced to enhance the model's ability to fuse multi-scale features. The output of M2CGA is first fed into MASAG, which combines features from the encoder and decoder. Through four mechanisms: multi-scale fusion, spatial selection, spatial interaction, and cross-regulation and recalibration, MASAG dynamically adjusts the spatial receptive field, significantly enhances the representation of key regions, and suppresses irrelevant background information. Finally, the output of MASAG is fed into the detection head of FOD-DETR for object detection. Specifically, the encoder and decoder outputs are X and G, respectively. Multi-scale fusion aggregates X and G in a semantically similar manner, preparing the fused feature map for subsequent processing. Spatial selection dynamically adjusts the receptive field to prioritize essential features, while spatial interaction and cross-regulation enrich the feature map with both local details and global context. During the recalibration stage, dynamically generated selective attention maps derived from multi-scale adaptive spatial attention gates are used to refine the initial encoder input, ensuring spatial accuracy during segmentation. This approach enables the multi-scale adaptive spatial attention gate to effectively cope with various detection challenges in the complex environment of transmission line external force damage, thereby ensuring precise and reliable detection results to obtain accurate external damage hidden danger categories and coordinate information.

[0063] In addition, power lines often appear long and thin in images. In order to enable the model to obtain a wider range of contextual information and thus obtain more accurate coordinates of power lines, the CAA-TransUNet model is built on the basis of TransUNet. By integrating the contextual anchor attention mechanism into the convolutional neural network, the feature extraction network can better grasp the contextual interdependence between distant pixels and enhance the central features, which facilitates better extraction of feature information to obtain accurate coordinate information of power lines.

[0064] S4, pixel-level positioning and discrimination module, is used to analyze the threat of external damage hazards in the image.

[0065] Specifically, in this embodiment, the pixel-level positioning discrimination module structure is designed as follows: Figure 4 As shown in the figure, during the inference process of the double-branch transmission line external damage hazard detection algorithm, the pixel-level positioning and discrimination module receives input from the pixel coordinates of the external damage hazard target anchor boxes identified by the FOD-DETR model and the pixel coordinates of the power lines segmented by the CAA-TransUNet model. First, the input set of hidden danger target anchor box coordinates is traversed, and each target box is spatially intersected with all line segments in the transmission line pixel set. A two-layer nested loop structure is used: the outer layer traverses the anchor boxes, and the inner layer traverses the transmission line segments. When a pixel intersection is detected between the current anchor box and any transmission line segment, the current segment traversal is immediately terminated (collision risk). The target is then determined to be a high-threat hazard, the anchor box it is located in is marked red, an alarm is issued, its label and coordinate information are output, and the image in which it is located is retained. If the current anchor box does not intersect with any transmission line segment (safe), the image is considered normal, no alarm is issued, and only the category and number of targets in the image are counted, and the image is not retained.

[0066] like Figure 5 As shown, the present invention discloses a system for analyzing the threat of external force damage to power transmission lines, including the following program modules:

[0067] The module for obtaining a dataset of potential external damage hazards for power lines is used to obtain a dataset. The dataset includes images and labels for potential external damage hazards. The labels are classified into seven categories: pump truck, forklift, crane, excavator, tower crane, pile driver, and tower. The dataset for power line segmentation includes images and labels for power lines.

[0068] The proposed model was trained and tested using an Ubuntu 20.04 operating system, an NVIDIA RTX3090 24GB GPU, an Intel Xeon Silver 4310 @ 2.10GHz CPU, Python 3.8.20, CUDA 11.8, and the PyTorch 2.4.1 deep learning framework. Experiments showed that the RT-DETR model achieved accuracy rates of 38.9% for pump trucks, 76.3% for forklifts, 65.6% for cranes, 80.4% for excavators, 82.2% for tower cranes, 78.9% for pile drivers, and 95.0% for pole towers, with an average accuracy of 73.9%. The TransUNet model achieved an Intersection-Over-Union (IoU) ratio of 77.95% for power line segmentation. After using the proposed model, the accuracy rates for each category were 62.2% for pump trucks, 82.8% for forklifts, 67.9% for cranes, 81.7% for excavators, 81.9% for tower cranes, 83.8% for pile drivers, and 95.6% for tower cranes. The average precision was 79.4%, and the IOR for power line segmentation was 79.11%. The average accuracy for identifying hidden danger targets increased by 5.5%, and the IOR for power line segmentation increased by 1.16%, demonstrating the effectiveness of the proposed model.

[0069] The threat analysis module performs spatial intersection detection on the relative position information of the power lines and the hidden dangers output by the dual-branch transmission line external damage hidden danger detection module. This module determines the threat level of the external damage hidden danger to the transmission channel and filters images based on the threat level of the hidden danger. The output includes the hidden danger's label information, coordinate information, hidden danger threat level, and images containing high-threat hidden danger targets.

[0070] The proposed system can perform threat analysis on images collected from video streams in batches. In this embodiment, an image is shown as an example. Figure 5 This is the visualization result diagram of Example 2. There are two targets with hidden dangers of external force damage in the figure, namely X1 and X2. Among them, X1 is a crane. After spatial intersection detection, it is detected that there is a risk of collision with the power line. The target is determined to be a high-threat hidden danger. The anchor box where it is located is marked red and an alarm is issued. Its label and coordinate information are output, and the image where it is located is retained; X2 is an excavator. After spatial intersection detection, it is determined to be a normal target. The anchor box where it is located is marked green. Since there is at least one high-threat target in the image, the image is retained.

[0071] This embodiment can identify external damage hazards and conduct threat analysis on images captured periodically by video streams during the task of protecting transmission lines from external force damage. Compared with target detection algorithms without threat analysis functions, the proposed method can filter images based on whether there are high-threat hazards, greatly reducing the number of invalid alarm images generated by the system, thereby significantly reducing the workload of subsequent manual review, and is conducive to improving the intelligent level of transmission line inspection.

[0072] Corresponding to the above technical solution, this embodiment further provides a system for analyzing the threat of hidden dangers of external force damage to power transmission lines, which is characterized by comprising:

[0073] A data acquisition unit, configured to acquire images of a transmission channel through a camera installed on a transmission line tower;

[0074] a data annotation unit, configured to annotate the transmission channel image with an anchor frame of an external damage hidden danger target and a transmission line segmentation annotation, thereby generating a transmission line external damage hidden danger dataset and a transmission line power line segmentation dataset respectively;

[0075] An image training and inference unit is configured to perform training based on a dual-branch transmission line external failure hazard detection algorithm according to the transmission line external failure hazard dataset and the transmission line power line segmentation dataset, and use the weights obtained from the training for image parallel inference to obtain the category and pixel coordinates of the external failure hazard target and the pixel coordinates of the power line;

[0076] a hidden danger identification unit, configured to perform spatial intersection detection on the pixel coordinates of the external breach hidden danger target and the pixel coordinates of the power line through a pixel-level positioning identification module; if there is an intersection, the corresponding hidden danger target is determined to be a high threat; otherwise, it is determined to be a low threat;

[0077] The alarm unit is used to trigger an alarm for an image containing the high-threat hidden danger target, and output the image and the label name and location information of the high-threat hidden danger in the image.

[0078] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0079] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for analyzing the threat of hidden dangers of external force damage to power transmission lines, characterized in that: include: Capture transmission channel images through cameras installed on transmission line towers; Performing external damage hidden danger target anchor frame annotation and transmission line segmentation annotation on the transmission channel image to generate a transmission line external damage hidden danger dataset and a transmission line power line segmentation dataset respectively; Based on the double-branch transmission line external failure hidden danger detection algorithm, training is performed according to the transmission line external failure hidden danger dataset and the transmission line power line segmentation dataset, and the weights obtained from the training are used for image parallel reasoning to obtain the category and pixel coordinates of the external failure hidden danger target and the pixel coordinates of the power line; The pixel coordinates of the external hidden danger target and the pixel coordinates of the power line are spatially intersected by a pixel-level positioning and discrimination module. If there is an intersection, the corresponding hidden danger target is determined to be a high threat, otherwise it is determined to be a low threat; An alarm is triggered for an image containing the high-threat hidden danger target, and the image and the label name and location information of the high-threat hidden danger in the image are output.

2. The method for analyzing the threat of hidden dangers of external force damage to power transmission lines according to claim 1, characterized in that: Based on the double-branch transmission line external damage hidden danger detection algorithm, training is performed based on the transmission line external damage hidden danger dataset and the transmission line power line segmentation dataset, and the weights obtained from the training are used for image parallel reasoning to obtain the category and pixel coordinates of the external damage hidden danger target and the pixel coordinates of the power line, including: Inputting the transmission line external damage hidden danger data set into the hidden danger detection model to obtain the category and pixel coordinates of the external damage hidden danger target; The power line segmentation dataset of the transmission line is input into a power line segmentation model to obtain the pixel coordinates of the power lines.

3. The method for analyzing the threat of hidden dangers of external force damage to power transmission lines according to claim 2, characterized in that: The transmission line external damage hidden danger dataset is input into the hidden danger detection model to obtain the category and pixel coordinates of the external damage hidden danger target, including: Anchor frames are annotated on the transmission line external damage hidden danger dataset, and the dataset is divided into a training set and a validation set in a ratio of 8:2, and input into the hidden danger detection model to start training; During the training phase, the transmission line damage risk dataset is fed into a feature extraction network containing a multi-scale multi-head convolutional gated attention mechanism to obtain a multi-scale, fine-grained feature map. Inputting the feature map into a multi-scale adaptive spatial attention gate in a neck network comprising a multi-scale adaptive spatial attention gate to perform multi-scale fusion, spatial selection, spatial interaction, cross-adjustment and recalibration to enhance key area expression and suppress irrelevant background information; The fused feature map output by the multi-scale adaptive spatial attention gate MASAG is fed into the FOD-DETR detection head; In the inference stage, the detection model weights are loaded, and forward inference is performed on the transmission line external damage hidden danger data set to output the category of the external damage hidden danger target and the corresponding pixel coordinates.

4. The method for analyzing the threat of hidden dangers of external force damage to power transmission lines according to claim 2, characterized in that: Inputting the power line segmentation dataset into a power line segmentation model to obtain pixel coordinates of the power lines includes: Embedding the contextual anchor attention mechanism into the power line segmentation model to form a CAA-TransUNet network to enhance the contextual modeling capability of slender power lines; Using the divided power line segmentation dataset to perform offline training on the CAA-TransUNet network to obtain segmentation model weights; In the inference stage, the segmentation model weights are loaded, and forward inference is performed on the power line segmentation dataset of the transmission line to output the pixel coordinates of the power line.

5. The method for analyzing the threat of hidden dangers of external force damage to power transmission lines according to claim 1, characterized in that: The categories of targets with potential external damage hazards include: pump trucks, forklifts, cranes, excavators, tower cranes, pile drivers and pole towers.

6. The method for analyzing the threat of hidden dangers of external force damage to power transmission lines according to claim 1, characterized in that: The spatial intersection detection adopts a pixel-level intersection detection method of target frame by target frame and line segment by line segment. After the first intersection is detected, the target is determined to be a high threat and subsequent detection of the current target frame is terminated.

7. The method for analyzing the threat of hidden dangers of external force damage to power transmission lines according to claim 1, characterized in that: Also includes: Images containing only low-threat or no-threat potential targets are counted without triggering an alarm.

8. The method for analyzing the threat of hidden dangers of external force damage to power transmission lines according to claim 2, characterized in that: The hidden danger detection model and the power line segmentation model obtain weights through offline training, and the reasoning stage runs in parallel on the same processor.

9. A system for analyzing the threat of external force damage to power transmission lines, characterized in that: include: A data acquisition unit, configured to acquire images of a transmission channel through a camera installed on a transmission line tower; a data annotation unit, configured to annotate the transmission channel image with an anchor frame of an external damage hidden danger target and a transmission line segmentation annotation, thereby generating a transmission line external damage hidden danger dataset and a transmission line power line segmentation dataset respectively; An image training and inference unit is configured to perform training based on a dual-branch transmission line external failure hazard detection algorithm according to the transmission line external failure hazard dataset and the transmission line power line segmentation dataset, and use the weights obtained from the training for image parallel inference to obtain the category and pixel coordinates of the external failure hazard target and the pixel coordinates of the power line; a hidden danger identification unit, configured to perform spatial intersection detection on the pixel coordinates of the external breach hidden danger target and the pixel coordinates of the power line through a pixel-level positioning identification module; if there is an intersection, the corresponding hidden danger target is determined to be a high threat; otherwise, it is determined to be a low threat; The alarm unit is used to trigger an alarm for an image containing the high-threat hidden danger target, and output the image and the label name and location information of the high-threat hidden danger in the image.

Images