Transmission conductor anomaly detection method, device and equipment

Abstract

The invention discloses a transmission conductor anomaly detection method, device and equipment, and the method comprises the steps: inputting a multi-modal fusion image obtained through image fusion through employing a DIVFusion technology into a preset YOLOv11 model, carrying out the anomaly detection of a transmission conductor, and obtaining an anomaly detection result which comprises an anomaly type and an anomaly positioning frame; extracting a transmission conductor from the multi-modal fusion image based on an MCMLSD algorithm to obtain a target conductor area graph; performing cross check in a mode of calculating an intersection-to-union ratio of an abnormal region according to the abnormal positioning frame and the target lead region graph to obtain a check result; and if the verification is passed, fusing the abnormity positioning frame and the target lead area graph, and generating a target detection result in combination with the abnormity type. The method can solve the technical problems that the prior art has application limitation and cannot adapt to the influence of a complex and changeable environment and the detection analysis of different wires, so that the wire anomaly detection result is lack of accuracy and reliability.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular to a method, device, and equipment for detecting abnormalities in transmission lines. Background Art

[0002] With the rapid development of intelligent and automated technologies, transmission line inspections are gradually shifting from manual inspections to automated inspections based on drones and image recognition. Detecting abnormalities in transmission line conductors is crucial during power facility inspections. Conductor abnormalities, such as cracks, wear, looseness, or corrosion, can cause serious power outages and even lead to safety accidents. Therefore, timely and accurate detection and location of these defects to ensure the safe operation of the power system has become a core task of power line inspections.

[0003] Existing technologies either perform anomaly detection on conductors made of special materials or rely on historical data or magnetic flux leakage data for anomaly detection. However, these detection technologies have obvious limitations. For example, the detection methods are not universal or cannot adapt to complex and changing environmental influences. As a result, conductor anomaly detection lacks accuracy and reliability, making it difficult to meet actual detection needs. Summary of the Invention

[0004] The present application provides a method, device and equipment for detecting abnormalities in power transmission lines, which are used to solve the technical problems that the existing technology has application limitations, cannot adapt to the influence of complex and changing environments and the detection and analysis of different lines, resulting in the lack of accuracy and reliability of the line abnormality detection results.

[0005] In view of this, the first aspect of the present application provides a method for detecting abnormalities in a transmission line, comprising:

[0006] The multimodal fusion image obtained by image fusion using DIVFusion technology is input into the preset YOLOv11 model to perform transmission line anomaly detection to obtain anomaly detection results, which include anomaly type and anomaly location frame;

[0007] Extracting the transmission lines from the multimodal fusion image based on the MCMLSD algorithm to obtain a target line area map;

[0008] A cross-check is performed by calculating an intersection-over-union ratio of the abnormal area based on the abnormal positioning frame and the target wire area map to obtain a check result;

[0009] If the verification result is passed, the abnormality positioning frame and the target wire area map are merged, and the target detection result is generated in combination with the abnormality type.

[0010] Preferably, the multimodal fusion image obtained by image fusion using the DIVFusion technology is input into a preset YOLOv11 model to perform transmission line anomaly detection, and the anomaly detection result is obtained, including:

[0011] DIVFusion technology is used to fuse the target infrared image and the target RGB image to obtain a multimodal fusion image;

[0012] The multimodal fusion image is input into a preset YOLOv11 model to perform transmission line anomaly detection to obtain an anomaly detection result. The preset YOLOv11 model includes a PKINet structure and an LSCD structure.

[0013] Preferably, the multimodal fusion image obtained by image fusion using the DIVFusion technology is input into a preset YOLOv11 model to perform transmission line anomaly detection to obtain anomaly detection results, and the method further includes:

[0014] Image data along the transmission line is collected by a drone equipped with an infrared camera and an RGB camera to obtain an initial infrared image and an initial RGB image.

[0015] The initial infrared image and the initial RGB image are registered based on the SIFT algorithm to generate a target infrared image and a target RGB image.

[0016] Preferably, the cross-checking is performed by calculating the intersection-over-union ratio of the abnormal area according to the abnormal positioning frame and the target wire area map to obtain the check result, including:

[0017] Determining a target abnormal area in the target wire area map based on the abnormality locating frame;

[0018] Calculating the intersection-over-union ratio between the abnormal positioning frame and the target abnormal area;

[0019] If the intersection-over-union ratio is greater than a preset value, the cross-check is passed; otherwise, the abnormality detection is determined to be incorrect and a check result is generated.

[0020] Preferably, if the verification result is passed, the abnormality positioning frame and the target wire area map are merged, and a target detection result is generated in combination with the abnormality type, and then the method further includes:

[0021] The target detection result triggers the alarm mechanism of the alarm system to achieve real-time abnormality monitoring.

[0022] A second aspect of the present application provides a transmission line abnormality detection device, comprising:

[0023] An anomaly detection unit is used to input the multimodal fusion image obtained by image fusion using DIVFusion technology into a preset YOLOv11 model to perform transmission line anomaly detection and obtain an anomaly detection result, wherein the anomaly detection result includes an anomaly type and an anomaly location frame;

[0024] A conductor extraction unit, configured to extract transmission conductors from the multimodal fusion image based on an MCMLSD algorithm to obtain a target conductor area map;

[0025] A cross-check unit, configured to perform a cross-check by calculating an intersection-over-union ratio of an abnormal region based on the abnormal location frame and the target conductor region map, to obtain a check result;

[0026] The detection fusion unit is used to fuse the abnormality positioning frame and the target wire area map if the verification result is passed, and generate a target detection result in combination with the abnormality type.

[0027] Preferably, the anomaly detection unit is specifically used to:

[0028] DIVFusion technology is used to fuse the target infrared image and the target RGB image to obtain a multimodal fusion image;

[0029] The multimodal fusion image is input into a preset YOLOv11 model to perform transmission line anomaly detection to obtain an anomaly detection result. The preset YOLOv11 model includes a PKINet structure and an LSCD structure.

[0030] Preferably, it also includes:

[0031] An image acquisition unit is used to collect image data along the transmission line using a drone equipped with an infrared camera and an RGB camera to obtain an initial infrared image and an initial RGB image;

[0032] The image calibration unit is used to align the initial infrared image and the initial RGB image based on the SIFT algorithm to generate a target infrared image and a target RGB image.

[0033] Preferably, the cross-check unit is specifically configured to:

[0034] Determining a target abnormal area in the target wire area map based on the abnormality locating frame;

[0035] Calculating the intersection-over-union ratio between the abnormal positioning frame and the target abnormal area;

[0036] If the intersection-over-union ratio is greater than a preset value, the cross-check is passed; otherwise, the abnormality detection is determined to be incorrect and a check result is generated.

[0037] A third aspect of the present application provides a power transmission line anomaly detection device, the device comprising a processor and a memory;

[0038] The memory is used to store program code and transmit the program code to the processor;

[0039] The processor is configured to execute the power transmission line anomaly detection method according to the instructions in the program code.

[0040] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0041] The present application provides a method for detecting anomalies in power transmission lines, comprising: inputting a multimodal fusion image obtained by image fusion using DIVFusion technology into a preset YOLOv11 model to perform power transmission line anomaly detection, thereby obtaining an anomaly detection result, wherein the anomaly detection result includes an anomaly type and an anomaly location frame; extracting the power transmission line from the multimodal fusion image based on the MCMLSD algorithm to obtain a target conductor area map; performing a cross-check by calculating an intersection-over-union ratio of the anomaly area based on the anomaly location frame and the target conductor area map to obtain a verification result; and if the verification result passes the verification, fusing the anomaly location frame and the target conductor area map to generate a target detection result in combination with the anomaly type.

[0042] The power transmission line anomaly detection method provided by the present application improves the performance of power transmission line anomaly detection from multiple different angles, such as multimodal image data, improved detection model and cross-checking. The fused multimodal fusion image can provide richer wire information, ensuring the reliability of detection and analysis from the data perspective. The preset YOLOv11 model can not only accurately locate the abnormal position of the wire, but also detect the abnormal type of the wire. In addition, the target wire extracted by the MCMLSD algorithm is cross-checked against the model detection results. If the check passes, information fusion is performed to ensure that the target detection results contain more and more accurate wire anomaly information, providing reliable information support for subsequent work. This process has strong robustness to the influence of wire differences and environmental changes, has few application limitations, and can ensure detection accuracy and reliability. Therefore, the present application can solve the technical problems that the existing technology has application limitations, cannot adapt to the influence of complex and changing environments and the detection and analysis of different wires, resulting in the lack of accuracy and reliability of wire anomaly detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of a flow chart of a method for detecting abnormalities in a power transmission line provided in an embodiment of the present application;

[0044] Figure 2 A schematic structural diagram of a transmission line anomaly detection device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] In order to help those skilled in the art better understand the present invention, 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 those skilled in the art without creative work are within the scope of protection of this application.

[0046] For easier understanding, see Figure 1 , an embodiment of a method for detecting abnormalities in a power transmission line provided by the present application includes:

[0047] Step 101: Input the multimodal fusion image obtained by image fusion using the DIVFusion technology into a preset YOLOv11 model to perform transmission line anomaly detection to obtain anomaly detection results. The anomaly detection results include anomaly types and anomaly positioning frames.

[0048] Furthermore, step 101 includes:

[0049] DIVFusion technology is used to fuse the target infrared image and the target RGB image to obtain a multimodal fusion image;

[0050] The multimodal fusion image is input into the preset YOLOv11 model to perform transmission line anomaly detection to obtain anomaly detection results. The preset YOLOv11 model includes a PKINet structure and a LSCD structure.

[0051] Furthermore, before step 101, the following steps are also included:

[0052] Image data along the transmission line is collected by a drone equipped with an infrared camera and an RGB camera to obtain an initial infrared image and an initial RGB image.

[0053] The initial infrared image and the initial RGB image are registered based on the SIFT algorithm to generate the target infrared image and the target RGB image.

[0054] It should be noted that the images used for the fusion operation in this embodiment are infrared images and RGB color images. These images can be collected along the transmission lines using drones equipped with infrared and RGB cameras; specifically, the drones hover over the conductors at a vertical downward angle to capture the images. There are no specific weather or time requirements for the images, as the dataset used in the model training phase already covers a wide range of complex environmental conditions, such as weather, lighting, and viewing angles, ensuring robustness to these conditions.

[0055] Since the initial infrared image and the initial RGB image obtained in this application need to be fused, in order to ensure the fusion effect and avoid fusion distortion and information loss, in addition to selecting DIVFusion (Darkness-free Infrared and Visible Image Fusion) technology to solve this problem, this embodiment also performs image registration processing before image fusion.

[0056] The registration uses the SIFT algorithm. First, the Difference of Gaussians (DoG) is used in spaces of different scales to detect key points. These key points are usually local extreme points, that is, in a Gaussian blurred image, the intensity value of a point is higher than the intensity values ​​of all points in its neighborhood. Then, a main direction is assigned to each key point, which is determined by the direction histogram of the local gradient of the image to ensure that the feature descriptor is invariant to rotation. Next, the histogram of the local image gradient centered on the key point is calculated to generate a feature descriptor. The descriptor consists of a series of vectors, each vector containing gradient direction and gradient amplitude information. The image is usually divided into 4×4 blocks, and each block is divided into 8 directions, generating a total of 128-dimensional feature vectors. Finally, these feature descriptors are used to perform feature matching between different images, and the best match is usually found by calculating the Euclidean distance between feature descriptors. The mathematical principle of the SIFT algorithm can be summarized as follows: for each key point, calculate its position σ in the scale space and its direction θ in the gradient space, then construct a 4×4 local area grid, and calculate the gradient histogram of 8 directions in each grid cell.

[0057] The DIVFusion technology used in this embodiment consists of two core components: the Scene Illumination Decoupling Network (SIDNet) and the Texture Contrast Enhancement Fusion Network (TCEFNet). This technology addresses the problem of weak texture detail and poor visual perception caused by severe degradation of visible light images in extreme low-light environments. The SIDNet consists of an encoder, three parallel Squeeze-and-Excitation (SE) blocks, and three parallel decoders. This network removes degraded illumination information from images while preserving the information characteristics of the source image, thereby eliminating the problem of illumination degradation in nighttime visible light images. In this embodiment, the unique features of multimodal images are enhanced. TCEFNet, consisting of a gradient-preserving module and a contrast-enhancing module, fuses complementary information to enhance the texture information and contrast of the fused features. Furthermore, DIVFusion technology uses a color consistency loss function to reduce color distortion, injecting more visible light domain information into the fused image, improving the fusion effect.

[0058] The preset YOLOv11 model in this embodiment is a pre-built and trained anomaly detection model. The training data is obtained by first acquiring a large number of infrared and RGB images of power transmission lines in complex environments and scenes. Data augmentation can then be performed before fusion to expand the image data, increase the diversity of the dataset, reduce the model's reliance on large-scale annotated data, and ensure the accuracy and reliability of model training. Enhancement processing includes, but is not limited to, horizontal and vertical flipping, image blurring, and random transformations of hue and saturation. Next, a labeling system can be used to manually label the weeds within the capacitor bank fence, using a minimum bounding rectangle to label the weeds in the image. The labeled dataset is stored in the YOLO dataset's annotation format as a txt file, containing the category and coordinate information of the labeled box in each image. Categories include loose and broken wire strands, recorded as the center point coordinates (x, y) and the width and height (w, h) of the labeled box to determine the relative position of the weeds in the image. Furthermore, the ratio between the dataset used for training the model and the validation set in this embodiment is 8:2. This ratio can be adjusted based on actual conditions and is provided for illustrative purposes only.

[0059] The YOLOv11 model in this embodiment is a lightweight and improved detection model. It primarily consists of a backbone network with a PKINet (Poly Kernel Inception Network) structure and a detection head with an LSCD (Lightweight Shared Convolutional Detection) architecture. The basic mechanism is that the backbone extracts features from the image, outputting a low-resolution, high-dimensional feature map. This high-dimensional feature map is then passed to the Neck for multi-scale feature fusion to accommodate the detection of objects of varying sizes. The multi-scale feature maps output by the Neck are then passed to the LSCD detection head, which performs bounding box regression, class prediction, and confidence calculation based on these feature maps. The LSCD detection head ultimately outputs the meter class, confidence, and relative position within the image. The LSCD detection head's shared convolution mechanism significantly reduces the number of model parameters and computational complexity, making the model more lightweight and reducing computational resource consumption. Therefore, this architecture overcomes the high computational resource requirements of existing models, which limit data processing efficiency and can also introduce processing delays or computational bottlenecks.

[0060] Specifically, PKINet uses multiple deep convolution kernels of varying sizes to perform multi-level feature extraction on images. This allows for the extraction of multi-scale texture features across different receptive fields, particularly the fusion of local features and global contextual information. This avoids the introduction of complex environmental noise and sparse feature representation. Furthermore, the context-anchored attention (CAA) module in PKINet captures long-range contextual information and works in conjunction with multi-scale deep convolution kernels to facilitate adaptive feature extraction of local and global contextual information, improving target detection performance. This module derives attention weights through global average pooling, 1×1 convolution, two serially connected depthwise separable convolutions, and then 1×1 convolution and a sigmoid activation function. The CAA module shares weights across all convolution modules, significantly reducing computational complexity and improving detection efficiency. Therefore, the PKINet of this embodiment can more efficiently and accurately identify and locate defective areas, particularly for detecting conductor anomalies of varying sizes, such as microcracks and large-scale corrosion. This approach is well-suited for the diverse forms of transmission line anomalies.

[0061] During model training, the initial batch size was set to 300, the training data size for each batch was 32, the learning rate was 0.0001, and the learning decay factor was 0.8. These parameters can be configured during the specific model training process and are provided here for illustrative purposes only and are not intended to be limiting. The trained, pre-set YOLOv11 model can perform transmission line anomaly detection and analysis on real-time multimodal fusion images, generating anomaly detection results. These results include anomaly types and anomaly location frames. The anomaly location frames are the locations of anomalies on the conductors, including specific coordinates, to quickly help operators locate the anomaly.

[0062] Step 102: Extract the transmission lines from the multimodal fusion image based on the MCMLSD algorithm to obtain a target line area map.

[0063] The MCMLSD algorithm simplifies the problem into a Markov chain model labeling on 1D line segments and combines it with dynamic programming to find the optimal line segment sequence. This method utilizes the global Hough transform for line segment recognition while restricting the search space. It then optimizes the extracted line segments using the Markov chain model and dynamic programming, ensuring both efficiency and accuracy. Furthermore, the algorithm considers the recognition of multiple line segments, making it particularly suitable for scenes with repetitive structures and multiple line segments.

[0064] The MCMLSD algorithm operates as follows: First, a global probabilistic Hough transform is used to identify the globally optimal line in the image. The problem is simplified by restricting the line segment search to the 1D lines detected by the Hough transform. Next, it is assumed that the distribution of line segments in these 1D point sequences can be modeled using a Markov chain. The state of each line segment depends only on its previous state, which allows the problem to be transformed into labeling hidden states on the point sequence. The output states in the Markov chain represent the possible positions of each point in the sequence. Next, a dynamic programming solution is used to accurately calculate the optimal probabilistic labeling of the line segment sequence given the observed data, outputting the optimal line segment sequence, which is represented as the optimal labeling on the point sequence. Finally, the local marginal posterior probability is used to estimate the expected number of correctly labeled points on the line segment. The detected line segments are ranked and the line segments are output, ranked by the expected number of labeled points. This generates a target wire region map.

[0065] Step 103 : performing a cross check by calculating the intersection-over-union ratio of the abnormal area according to the abnormal location frame and the target wire area map to obtain a check result.

[0066] Furthermore, step 103 includes:

[0067] Determine a target abnormal area in the target wire area map based on the abnormal location frame;

[0068] Calculate the intersection-over-union ratio between the anomaly localization box and the target anomaly area;

[0069] If the intersection-over-union ratio is greater than the preset value, the cross-check is passed; otherwise, the anomaly detection is judged to be incorrect and a check result is generated.

[0070] It should be noted that the intersection-and-union ratio is not calculated directly based on the target wire area map, but the area corresponding to the abnormal positioning frame in the target wire area map is found, that is, the target abnormal area; then the intersection-and-union ratio is calculated based on the abnormal positioning frame and the target abnormal area. :

[0071]

[0072] in, Represents the abnormal positioning box predicted by the model, represents the target abnormal area obtained by the wire extraction operation; Represents the area intersection of the anomaly positioning box and the target anomaly area; Represents the area union of the anomaly localization box and the target anomaly area.

[0073] Assume that the default value is recorded as ,like , it indicates that the abnormal positioning frame has a high degree of overlap with the abnormal wire part of the target abnormal area, verifying that the abnormal detection is effective, that is, it passes the verification; otherwise, it means that the detected abnormal positioning frame is not accurate enough, or the wire extraction is incorrect, and re-detection and evaluation are required, and it fails the verification.

[0074] Step 104: If the verification result is passed, the abnormality location frame and the target wire area map are merged, and the target detection result is generated in combination with the abnormality type.

[0075] If the verification is passed, the abnormal location frame can be integrated into the target wire area map, and the target detection result can be generated in combination with the abnormal type. That is, the abnormal position and abnormal type of the wire are marked on the wire map, which makes it easier for operators to quickly and accurately understand the situation of the abnormal wire.

[0076] Furthermore, step 104 further includes:

[0077] The alarm mechanism of the alarm system is triggered based on the target detection results to achieve real-time anomaly monitoring.

[0078] The purpose of the alarm is to quickly deliver target detection results to operators, enabling them to quickly identify conductor anomalies based on the anomaly type and location on the conductor map, enabling them to respond and address them promptly, thus preventing further grid losses caused by conductor anomalies. The alarm mechanism is not limited to specific alarm types and can be designed to include audible, visual, or text messages, depending on actual needs.

[0079] The power transmission line anomaly detection method provided by the embodiment of the present application improves the performance of power transmission line anomaly detection from multiple different angles, such as multimodal image data, improved detection model and cross-checking. The fused multimodal fusion image can provide richer wire information, ensuring the reliability of detection and analysis from the data perspective. The preset YOLOv11 model can not only accurately locate the abnormal position of the wire, but also detect the abnormal type of the wire. In addition, the target wire extracted by the MCMLSD algorithm is cross-checked against the model detection results. If the check passes, information fusion is performed to ensure that the target detection results contain more and more accurate wire anomaly information, providing reliable information support for subsequent work. This process has strong robustness to the influence of wire differences and environmental changes, has few application limitations, and can ensure detection accuracy and reliability. Therefore, the embodiment of the present application can solve the technical problems that the existing technology has application limitations, cannot adapt to the influence of complex and changing environments and the detection and analysis of different wires, resulting in the lack of accuracy and reliability of wire anomaly detection results.

[0080] For easier understanding, see Figure 2 The present application provides an embodiment of a transmission line abnormality detection device, comprising:

[0081] An anomaly detection unit 201 is configured to input a multimodal fusion image obtained by image fusion using the DIVFusion technology into a preset YOLOv11 model to perform transmission line anomaly detection and obtain an anomaly detection result, wherein the anomaly detection result includes an anomaly type and an anomaly location frame;

[0082] The wire extraction unit 202 is used to extract the power transmission wires from the multimodal fusion image based on the MCMLSD algorithm to obtain a target wire area map;

[0083] A cross-check unit 203 is configured to perform a cross-check by calculating an intersection-over-union ratio of the abnormal area based on the abnormal location frame and the target wire area map to obtain a check result;

[0084] The detection fusion unit 204 is configured to fuse the abnormality location frame and the target wire area map if the verification result is passed, and generate a target detection result in combination with the abnormality type.

[0085] Preferably, the anomaly detection unit 201 is specifically configured to:

[0086] DIVFusion technology is used to fuse the target infrared image and the target RGB image to obtain a multimodal fusion image;

[0087] The multimodal fusion image is input into the preset YOLOv11 model to perform transmission line anomaly detection to obtain anomaly detection results. The preset YOLOv11 model includes a PKINet structure and a LSCD structure.

[0088] Preferably, it also includes:

[0089] The image acquisition unit 205 is used to collect image data along the transmission line using a drone equipped with an infrared camera and an RGB camera to obtain an initial infrared image and an initial RGB image;

[0090] The image calibration unit 206 is configured to register the initial infrared image and the initial RGB image based on the SIFT algorithm to generate a target infrared image and a target RGB image.

[0091] Preferably, the cross-check unit 203 is specifically configured to:

[0092] Determine a target abnormal area in the target wire area map based on the abnormal location frame;

[0093] Calculate the intersection-over-union ratio between the anomaly localization box and the target anomaly area;

[0094] If the intersection-over-union ratio is greater than the preset value, the cross-check is passed; otherwise, the anomaly detection is judged to be incorrect and a check result is generated.

[0095] The present application also provides a power transmission line anomaly detection device, the device including a processor and a memory;

[0096] The memory is used to store program codes and transmit the program codes to the processor;

[0097] The processor is configured to execute the power transmission line anomaly detection method in the above method embodiment according to instructions in the program code.

[0098] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0099] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0100] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0101] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the method described in each embodiment of the present application through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name: Read-Only Memory, English abbreviation: ROM), random access memory (full name: Random Access Memory, English abbreviation: RAM), disk or optical disk, and other media that can store program code.

[0102] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above 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 method for detecting abnormality of a transmission line, characterized in that: include: The multimodal fusion image obtained by image fusion using DIVFusion technology is input into the preset YOLOv11 model to perform transmission line anomaly detection to obtain anomaly detection results, which include anomaly type and anomaly location frame; Extracting the transmission lines from the multimodal fusion image based on the MCMLSD algorithm to obtain a target line area map; A cross-check is performed by calculating an intersection-over-union ratio of the abnormal area based on the abnormal positioning frame and the target wire area map to obtain a check result; If the verification result is passed, the abnormality positioning frame and the target wire area map are merged, and the target detection result is generated in combination with the abnormality type.

2. The method for detecting abnormality of a transmission line according to claim 1, wherein: The multimodal fusion image obtained by image fusion using the DIVFusion technology is input into a preset YOLOv11 model to perform transmission line anomaly detection, and the anomaly detection results obtained include: DIVFusion technology is used to fuse the target infrared image and the target RGB image to obtain a multimodal fusion image; The multimodal fusion image is input into a preset YOLOv11 model to perform transmission line anomaly detection to obtain an anomaly detection result. The preset YOLOv11 model includes a PKINet structure and an LSCD structure.

3. The method for detecting abnormality of a transmission line according to claim 2, wherein: The multimodal fusion image obtained by image fusion using the DIVFusion technology is input into a preset YOLOv11 model to perform transmission line anomaly detection to obtain anomaly detection results, which also includes: Image data along the transmission line is collected by a drone equipped with an infrared camera and an RGB camera to obtain an initial infrared image and an initial RGB image. The initial infrared image and the initial RGB image are registered based on the SIFT algorithm to generate a target infrared image and a target RGB image.

4. The method for detecting abnormality of a transmission line according to claim 1, wherein: The cross-check is performed by calculating the intersection-over-union ratio of the abnormal area according to the abnormal positioning frame and the target wire area map to obtain a check result, including: Determining a target abnormal area in the target wire area map based on the abnormality locating frame; Calculating the intersection-over-union ratio between the abnormal positioning frame and the target abnormal area; If the intersection-over-union ratio is greater than a preset value, the cross-check is passed; otherwise, the abnormality detection is determined to be incorrect and a check result is generated.

5. The method for detecting abnormality of a transmission line according to claim 1, wherein: If the verification result is passed, the abnormality positioning frame and the target wire area map are merged, and a target detection result is generated in combination with the abnormality type, and then the following steps are further included: The target detection result triggers the alarm mechanism of the alarm system to achieve real-time abnormality monitoring.

6. A transmission line abnormality detection device, characterized in that: include: An anomaly detection unit is used to input the multimodal fusion image obtained by image fusion using DIVFusion technology into a preset YOLOv11 model to perform transmission line anomaly detection and obtain an anomaly detection result, wherein the anomaly detection result includes an anomaly type and an anomaly location frame; A conductor extraction unit, configured to extract transmission conductors from the multimodal fusion image based on an MCMLSD algorithm to obtain a target conductor area map; A cross-check unit, configured to perform a cross-check by calculating an intersection-over-union ratio of an abnormal region based on the abnormal location frame and the target conductor region map, to obtain a check result; The detection fusion unit is used to fuse the abnormality positioning frame and the target wire area map if the verification result is passed, and generate a target detection result in combination with the abnormality type.

7. The power transmission line abnormality detection device according to claim 6, characterized in that: The anomaly detection unit is specifically configured to: DIVFusion technology is used to fuse the target infrared image and the target RGB image to obtain a multimodal fusion image; The multimodal fusion image is input into a preset YOLOv11 model to perform transmission line anomaly detection to obtain an anomaly detection result. The preset YOLOv11 model includes a PKINet structure and an LSCD structure.

8. The power transmission line abnormality detection device according to claim 7, characterized in that: Also includes: An image acquisition unit is used to collect image data along the transmission line using a drone equipped with an infrared camera and an RGB camera to obtain an initial infrared image and an initial RGB image; The image calibration unit is used to align the initial infrared image and the initial RGB image based on the SIFT algorithm to generate a target infrared image and a target RGB image.

9. The power transmission line abnormality detection device according to claim 6, characterized in that: The cross-check unit is specifically used to: Determining a target abnormal area in the target wire area map based on the abnormality locating frame; Calculating the intersection-over-union ratio between the abnormal positioning frame and the target abnormal area; If the intersection-over-union ratio is greater than a preset value, the cross-check is passed; otherwise, the anomaly detection is determined to be incorrect and a check result is generated.

10. A power transmission line abnormality detection device, characterized in that: The device includes a processor and a memory; The memory is used to store program code and transmit the program code to the processor; The processor is configured to execute the power transmission line anomaly detection method according to any one of claims 1 to 5 according to instructions in the program code.

Images