Distribution line unmanned aerial vehicle inspection image fusion method and system and medium

Abstract

The invention relates to a distribution line unmanned aerial vehicle inspection image fusion method and system and a medium, and the method comprises the steps: carrying out the aerial inspection of a distribution line through a high-resolution visible light camera and an infrared thermal imaging camera carried by an unmanned aerial vehicle, and obtaining the image data of a key part; performing denoising processing on the acquired multi-modal image to reduce interference caused by environmental factors and sensor noise, and then enhancing the image contrast and improving the definition of image details through a histogram equalization method to obtain denoised and enhanced intermediate preprocessing data; according to the method, a power distribution line unmanned aerial vehicle image fusion model composed of an encoder and a decoder is constructed, image fusion is performed on source images of different modalities after denoising and enhancement, so that data advantages of different modalities are fully utilized, richer result images are generated, more comprehensive line state information can be provided, and the reliability of the power distribution line unmanned aerial vehicle image fusion is improved. More accurate state evaluation and fault diagnosis are realized, the inspection efficiency and accuracy are improved, the cost is reduced, and the safety is improved.

Description

Technical Field

[0001] This application relates to the field of power system security, and more specifically, to an image fusion method, system, and medium for unmanned aerial vehicle (UAV) inspection of power distribution lines. Background Technology

[0002] Multimodal cross-scale feature fusion technology integrates data from different sensors (such as visible light, infrared, and thermal imaging) to capture detailed information about power distribution lines at multiple scales, thereby improving the accuracy and robustness of fault detection. The application of this technology not only enhances the efficiency and accuracy of inspections but also provides a new direction for the intelligent development of the power industry. Furthermore, multimodal data fusion technology can provide more comprehensive line status information, contributing to more accurate condition assessment and fault diagnosis.

[0003] Traditional image fusion methods for UAV inspections of power distribution lines primarily rely on manually designed feature extraction and target detection algorithms. These methods typically require expert knowledge and experience to select and design effective features, such as edge, texture, and shape features. While traditional methods achieve accuracy close to deep learning methods in classifying power lines and towers, they perform poorly in classifying environmental targets (ground, vegetation, and rooftops). This is mainly due to the significant differences in density and shape of environmental target point clouds, making classification more difficult. Furthermore, traditional methods may be less effective than deep learning-based methods when processing image information containing hierarchical structures, as they struggle to automatically capture and learn complex image features.

[0004] Deep learning-based image fusion methods for UAV inspection of power distribution lines utilize deep neural networks, especially convolutional neural networks (CNNs), to automatically learn features from images. These methods can more effectively identify power equipment and detect defects in aerial images. Compared to traditional methods, deep learning methods can more effectively achieve target detection and feature extraction because they can automatically learn target similarity features at different point cloud densities without manual intervention. For example, the MinkUNet model extracts point cloud features at different resolutions through spatial 3D convolution and a multi-level feature pyramid architecture, and significantly reduces computational burden during training by using sparse matrix operators. Furthermore, deep learning-based methods such as YOLOv5 improve the performance of transmission line fault detection through attention mechanisms and cross-scale feature fusion. Compared to traditional methods, deep learning-based methods have significant advantages in automatic feature extraction, processing complex image information, and adapting to different environmental conditions. Deep learning models can handle large amounts of data and automatically learn hierarchical feature representations, which is particularly important in multimodal cross-scale feature fusion. However, deep learning methods typically require large amounts of labeled data to train the model, while traditional methods rely on expert knowledge and predefined feature sets. In practical applications, deep learning methods are gradually becoming the preferred technology for image fusion in UAV inspections of power distribution lines due to their high efficiency and accuracy. Summary of the Invention

[0005] The purpose of this application is to provide a method, system, and medium for image fusion of UAV inspection of power distribution lines using multimodal cross-scale feature fusion, which can provide more comprehensive line status information, achieve more accurate status assessment and fault diagnosis, improve inspection efficiency and accuracy, reduce costs, and improve safety.

[0006] To achieve the above objectives, this application provides the following technical solution:

[0007] In a first aspect, embodiments of this application provide an image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines, comprising the following steps:

[0008] Step 1: Use the high-resolution visible light camera and infrared thermal imaging camera mounted on the drone to conduct aerial inspection of the power distribution line and obtain image data of key parts, including conductors, insulators and towers.

[0009] Step 2: Denoise the acquired images to reduce interference from environmental factors and sensor noise, and then enhance image contrast and improve the clarity of image details by using histogram equalization.

[0010] Step 3: The features extracted from different scales and modalities are fused to construct a power distribution line UAV image fusion model consisting of an encoder and a decoder. The encoder includes a scale-adaptive residual module, which is composed of a context-aware information exchanger. The core of the context-aware information exchanger is a deep feature distillation module and an inter-scale information integration module. The decoder includes a hierarchical feature fusion network. Finally, the fusion result is used for subsequent fault detection.

[0011] In step 1, the drone flight path is planned to ensure that the drone can cover all key parts of the power distribution line, including but not limited to conductors, insulators, and towers; visible light images and infrared thermal images are collected to ensure that the timestamps of the two types of image data are consistent, which facilitates subsequent image denoising, enhancement, and fusion; the geographic information of the image data, including latitude, longitude, and altitude, is recorded so as to associate the image data with the actual geographic location and improve the accuracy of fault location.

[0012] The denoised and enhanced multimodal images were used to form the dataset for the power distribution line drone project, which was then divided into training and testing sets according to a certain ratio.

[0013] In step 2, the acquired image is denoised to reduce interference caused by environmental factors and sensor noise. Specifically, Gaussian filtering, median filtering, or bilateral filtering are used.

[0014] In step 2, image contrast is enhanced and the clarity of image details is improved using histogram equalization. The formula is as follows:

[0015]

[0016] in, It is the output grayscale level. It is the input grayscale level. It is the total number of gray levels. It is the histogram of the input image.

[0017] In step 3, the scale-adaptive residual module implements the residual connection between the input and output images as follows:

[0018]

[0019] in, This represents the features after processing by the scale-adaptive residual module. Indicates the input image. This represents the convolution operation. Represents a context-aware information exchanger.

[0020] Each context-aware information exchanger consists of three parallel fully convolutional streams, each processing feature maps at different resolutions. Each stream has a specific channel dimension. To improve network training efficiency and avoid the vanishing gradient problem, the context-aware information exchanger uses residual connections, summing the input and output of each convolutional stream to ensure network stability. The input features are downsampled by a factor of 2 and 4, respectively, resulting in features at three different scales. These features are then fed into a deep feature distillation module and an inter-scale information integration module. To obtain more refined features, the results are again fed into the deep feature distillation module and the inter-scale information integration module to obtain the final result. This process can be described as follows:

[0021] .

[0022] The deep feature distillation module enhances feature representation capabilities by extracting and refining features within a convolutional network. This module dynamically focuses on key regions of the image using an attention mechanism while preserving the detail integrity of these regions. First, three feature maps at different scales are input in parallel into the processing unit, and after passing through convolutional layers, the LeakyReLU activation function, and convolution operations, intermediate feature maps are obtained. This process can be represented by the following formula:

[0023]

[0024] in, Represents the input image. This represents the intermediate feature map. Subsequently, the intermediate feature map undergoes channel adjustment through a 1x1 convolutional layer, followed by feature refinement to obtain the enhanced feature map. This process can be represented as:

[0025]

[0026] After feature enhancement, the refined feature map is multiplied by the attention weight matrix to obtain a weighted feature map, which is then input into the feature extraction unit to generate a new feature representation. Finally, the initial feature map and the refined feature map are fused through a residual connection to produce the final output. These operations can be represented by the following formula:

[0027]

[0028] in, This represents the attention weight matrix.

[0029] The inter-scale information integration module receives three feature maps at different scales from the deep feature distillation module as input. First, cross-scale information interaction is performed between the lowest-scale and intermediate-scale feature maps to obtain semantic information from the low-level features. Specifically, the low-scale feature map is upsampled and then element-wise added to the intermediate-scale feature map, followed by average pooling to enhance the generalization ability of the features. Next, channel compression is used to reduce the feature dimensionality, generating a compact feature representation. Subsequently, two sets of key features are generated through two parallel convolution operations. These two feature maps are then element-wise added and fused with the highest-scale feature map in the same way, finally yielding the output image of the encoder.

[0030] The hierarchical feature fusion network extracts two sets of feature maps at four scales from the scale-adaptive residual module. Four features are obtained by splicing features of the same scale in the same region. First, the first-level features... Features through The operation reduces the feature dimension by half while doubling the size, achieving the same result as... The dimensions are consistent. Then... and Then splice and go through a The operation reduces the feature dimension by half while doubling the size, achieving the same result as... The dimensions are consistent. This process of fusing features from low to high hierarchy is repeated to obtain the final fused image.

[0031] Secondly, embodiments of this application provide an image fusion system for unmanned aerial vehicle (UAV) inspection of power distribution lines. The system includes a memory and a processor. The memory includes a program for an image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines. When the program for the image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines is executed by the processor, it implements the steps of the above-described image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines.

[0032] Thirdly, embodiments of this application provide a computer-readable storage medium storing program code, which, when executed by a processor, implements the steps of the image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines as described above.

[0033] Compared with existing technologies, the beneficial effects of this invention are as follows: By performing frequency domain scaling transformation on the preprocessed image using Fourier transform, the detailed features in the image can be effectively enhanced. The application of a high-pass filter further highlights the high-frequency components of the image, making minute hidden dangers that were originally difficult to identify against the background of complex power distribution lines clearly visible, providing a high-quality image foundation for subsequent fine-grained identification, thereby improving the accuracy of identification. For example, for some small cracks, corrosion spots, and other hidden dangers, after frequency domain scaling transformation, their edge and texture features are more obvious, making it easier for the model to accurately distinguish them from the background. In addition, the use of transfer learning enhances the model's generalization ability under different scenarios and conditions, enabling the fine-grained defect identification module to accurately classify and locate defects.

[0034] In terms of image preprocessing, operations such as Gaussian filtering for noise reduction and histogram equalization to enhance contrast effectively improve image quality. The system also features real-time identification and report generation capabilities, enabling rapid output of defect target identification results and the generation of detailed reports, thereby improving the efficiency and response speed of power line inspections. To make defects more intuitive and easily identifiable, the system overlays identified defect targets as graphic markers onto the original image. Even against complex backgrounds, the system can effectively detect typical targets, improving its identification capabilities in complex environments. This allows for the timely detection and identification of hidden dangers and defects in power lines, ensuring the safe and stable operation of power lines. Attached Figure Description

[0035] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is the overall structure diagram of this application;

[0037] Figure 2 This is the scale-adaptive residual module of this application;

[0038] Figure 3 It is the context-aware information exchanger of this application;

[0039] Figure 4 This is the deep feature distillation module of this application;

[0040] Figure 5 This is the inter-scale information integration module of this application;

[0041] Figure 6 This is the hierarchical feature fusion network of this application. Detailed Implementation

[0042] The technical solutions of the embodiments of this application will now be described with reference to the accompanying drawings. It should be noted that similar reference numerals and letters in the following drawings indicate similar items; therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0043] The terms “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0044] The terms “first,” “second,” etc., are used only to distinguish one entity or operation from another, and should not be construed as indicating or implying relative importance, nor as requiring or implying any such actual relationship or order between these entities or operations.

[0045] like Figures 1 to 6 As shown, the present invention provides an image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines, comprising the following steps:

[0046] Step 1: Use the high-resolution visible light camera and infrared thermal imaging camera mounted on the drone to conduct aerial inspection of the power distribution line and obtain image data of key parts, including conductors, insulators and towers.

[0047] Step 2: Denoise the acquired images to reduce interference from environmental factors and sensor noise, and then enhance image contrast and improve the clarity of image details by using histogram equalization.

[0048] Step 3: The features extracted from different scales and modalities are fused to construct a power distribution line UAV image fusion model consisting of an encoder and a decoder. The encoder includes a scale-adaptive residual module, which is composed of a context-aware information exchanger. The core of the context-aware information exchanger is a deep feature distillation module and an inter-scale information integration module. The decoder includes a hierarchical feature fusion network. Finally, the fusion result is used for subsequent fault detection.

[0049] In step 1, the drone flight path is planned to ensure that the drone can cover all key parts of the power distribution line, including but not limited to conductors, insulators, and towers; visible light images and infrared thermal images are collected to ensure that the timestamps of the two types of image data are consistent, which facilitates subsequent image denoising, enhancement, and fusion; the geographic information of the image data, including latitude, longitude, and altitude, is recorded so as to associate the image data with the actual geographic location and improve the accuracy of fault location.

[0050] The denoised and enhanced multimodal images were used to form the dataset for the power distribution line drone project, which was then divided into training and testing sets according to a certain ratio.

[0051] In step 2, the acquired image is denoised to reduce interference caused by environmental factors and sensor noise. Specifically, Gaussian filtering, median filtering, or bilateral filtering are used.

[0052] In step 2, image contrast is enhanced and the clarity of image details is improved using histogram equalization. The formula is as follows:

[0053]

[0054] in, It is the output grayscale level. It is the input grayscale level. It is the total number of gray levels. It is the histogram of the input image.

[0055] In step 3, the scale-adaptive residual module implements the residual connection between the input and output images as follows:

[0056]

[0057] in, This represents the features after processing by the scale-adaptive residual module. Indicates the input image. This represents the convolution operation. Represents a context-aware information exchanger.

[0058] Each context-aware information exchanger consists of three parallel fully convolutional streams, each processing feature maps at different resolutions. Each stream has a specific channel dimension. To improve network training efficiency and avoid the vanishing gradient problem, the context-aware information exchanger uses residual connections, summing the input and output of each convolutional stream to ensure network stability. The input features are downsampled by a factor of 2 and 4, respectively, resulting in features at three different scales. These features are then fed into a deep feature distillation module and an inter-scale information integration module. To obtain more refined features, the results are again fed into the deep feature distillation module and the inter-scale information integration module to obtain the final result. This process can be described as follows:

[0059] .

[0060] The deep feature distillation module enhances feature representation capabilities by extracting and refining features within a convolutional network. This module dynamically focuses on key regions of the image using an attention mechanism while preserving the detail integrity of these regions. First, three feature maps at different scales are input in parallel into the processing unit, and after passing through convolutional layers, the LeakyReLU activation function, and convolution operations, intermediate feature maps are obtained. This process can be represented by the following formula:

[0061]

[0062] in, Represents the input image. This represents the intermediate feature map. Subsequently, the intermediate feature map undergoes channel adjustment through a 1x1 convolutional layer, followed by feature refinement to obtain the enhanced feature map. This process can be represented as:

[0063]

[0064] After feature enhancement, the refined feature map is multiplied by the attention weight matrix to obtain a weighted feature map, which is then input into the feature extraction unit to generate a new feature representation. Finally, the initial feature map and the refined feature map are fused through a residual connection to produce the final output. These operations can be represented by the following formula:

[0065]

[0066] in, This represents the attention weight matrix.

[0067] The inter-scale information integration module receives three feature maps at different scales from the deep feature distillation module as input. First, cross-scale information interaction is performed between the lowest-scale and intermediate-scale feature maps to obtain semantic information from the low-level features. Specifically, the low-scale feature map is upsampled and then element-wise added to the intermediate-scale feature map, followed by average pooling to enhance the generalization ability of the features. Next, channel compression is used to reduce the feature dimensionality, generating a compact feature representation. Subsequently, two sets of key features are generated through two parallel convolution operations. These two feature maps are then element-wise added and fused with the highest-scale feature map in the same way, finally yielding the output image of the encoder.

[0068] The hierarchical feature fusion network extracts two sets of feature maps at four scales from the scale-adaptive residual module. Four features are obtained by splicing features of the same scale in the same region. First, the first-level features... Features through The operation reduces the feature dimension by half while doubling the size, achieving the same result as... The dimensions are consistent. Then... and Then splice and go through a The operation reduces the feature dimension by half while doubling the size, achieving the same result as... The dimensions are consistent. This process of fusing features from low to high hierarchy is repeated to obtain the final fused image.

[0069] This application provides an image fusion system for unmanned aerial vehicle (UAV) inspection of power distribution lines. The system includes a memory and a processor. The memory includes a program for an image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines. When the program for the image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines is executed by the processor, it implements the steps of the above-described image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines.

[0070] This application provides a computer-readable storage medium storing program code. When the program code is executed by a processor, it implements the steps of the image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines as described above.

[0071] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0072] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0073] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0074] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0075] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0076] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0077] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0078] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for image fusion in unmanned aerial vehicle (UAV) inspection of power distribution lines, characterized in that, Includes the following steps: Step 1: Use the high-resolution visible light camera and infrared thermal imaging camera mounted on the drone to conduct aerial inspection of the power distribution line and obtain image data of key parts, including conductors, insulators and towers. Step 2: Denoise the acquired multimodal images to reduce interference from environmental factors and sensor noise. Then, enhance the image contrast and improve the clarity of image details by using histogram equalization to obtain the denoised and enhanced intermediate preprocessed data. Step 3: Construct a UAV image fusion model for power distribution lines consisting of an encoder and a decoder. The source images of different modalities after denoising and enhancement are fused to make full use of the data advantages of different modalities and generate richer result images. The encoder includes a scale-adaptive residual module, which consists of two context-aware information exchangers. The context-aware information exchangers include a deep feature distillation module and an inter-scale information integration module. The decoder includes a hierarchical feature fusion module. The fused result image generated by the image fusion model is used for subsequent fault detection in complex and heavily interfered background environments.

2. The image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines according to claim 1, characterized in that, In step 1, the drone flight path is planned to ensure that the drone can cover all key parts of the power distribution line, including but not limited to conductors, insulators, and towers; visible light images and infrared thermal images are acquired to ensure that the timestamps of the two types of image data are consistent, which facilitates subsequent image denoising, enhancement, and fusion; the geographic information of the image data, including latitude, longitude, and altitude, is recorded so as to associate the image data with the actual geographic location and improve the accuracy of fault location.

3. The image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines according to claim 1, characterized in that, In step 2, the acquired image is denoised to reduce interference caused by environmental factors and sensor noise. Specifically, Gaussian filtering, median filtering, or bilateral filtering are used.

4. The image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines according to claim 1, characterized in that, In step 2, the method of enhancing image contrast and improving the clarity of image details through histogram equalization specifically involves: Histogram equalization enhances image contrast by adjusting the histogram distribution of an image. Its formula is: ， in, It is the output grayscale level. It is the input grayscale level. It is the total number of gray levels. It is the histogram of the input image.

5. The image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines according to claim 1, characterized in that, In step 3, the scale-adaptive residual module implements the residual connection between the input and output images as follows: ， in, This represents the features after processing by the scale-adaptive residual module. Indicates the input image. This represents the convolution operation. Represents a context-aware information exchanger.

6. The image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines according to claim 1, characterized in that, The context-aware information exchanger consists of three parallel fully convolutional streams, each processing feature maps at different resolutions. Each stream has a specific channel dimension. The context-aware information exchanger uses residual connections to sum the input and output of each convolutional stream to ensure network stability. The input features are downsampled by 2x and 4x respectively to obtain features at three different scales. These features are fed into the deep feature distillation module and the inter-scale information integration module. To obtain more refined features, the results are again fed into the deep feature distillation module and the inter-scale information integration module to obtain the final result.

7. The image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines according to claim 1, characterized in that, The deep feature distillation module enhances feature representation by distilling features in the convolutional stream, adaptively focusing on important regions in the image through an attention mechanism while preserving detailed information in these regions.

8. The image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines according to claim 1, characterized in that, The hierarchical feature fusion module gradually restores the image resolution from low to high through multi-scale auxiliary fusion, generating the final multimodal fused image.

9. An image fusion system for unmanned aerial vehicle (UAV) inspection of power distribution lines, characterized in that, The system includes a memory and a processor. The memory includes a program for a method of image fusion for UAV inspection of power distribution lines. When the program for the method of image fusion for UAV inspection of power distribution lines is executed by the processor, it implements the steps of the method of image fusion for UAV inspection of power distribution lines as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program code, which, when executed by a processor, implements the steps of the image fusion method for unmanned aerial vehicle (UAV) inspection of power distribution lines as described in any one of claims 1-8.

Images