A Fault Identification Method and System for Power Transmission Lines Based on Deep Learning

Abstract

The present invention belongs to the technical field of power line fault identification, and discloses a power transmission line fault identification method and system based on deep learning. The method includes: acquiring a monitoring image of a transmission line, preprocessing the monitoring image of the transmission line to obtain a target image; solving the light value of the target image based on an enhanced light value estimation model to obtain the atmospheric light value of the target image; performing multi-level feature extraction on the target image based on a dense pyramid network to obtain the transmittance map of the target image; inputting the atmospheric light value of the target image and the transmittance map of the target image into an atmospheric scattering model to obtain a defogged image; and inputting the defogged image into a pre-constructed fault identification model to output the fault type of the power transmission line. By accurately calculating the atmospheric light value and transmittance map of the target image, the present invention can obtain a high-quality defogged image; by performing fault identification on the high-quality defogged image, the accuracy of line fault identification can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power line fault identification, and specifically relates to a method and system for power transmission line fault identification based on deep learning. Background Art

[0002] China's demand for electricity continues to grow with its rapid economic development. China is carrying out power grid construction across the country to meet the needs of industrial and civilian electricity. With the continuous development of hydropower energy in southwest my country and southeast Tibet, a number of ultra-high voltage transmission projects have been put into operation as the "artery" for hydropower transmission. However, due to the existence of a large number of nonlinear devices and uncertain factors in the power system, the system stability is reduced and the accident rate is increased. Power lines that are outdoors for a long time are not only disturbed by natural environmental factors, but also bear the load of power systems and mechanical equipment. Natural factors such as lightning, hurricanes, and heavy rains can cause negative effects such as corrosion and oxidation of power line equipment. For the maintenance of power lines, failure to take preventive measures can easily lead to the emergence of corresponding problems.

[0003] In the past, the inspection method of power transmission lines mainly relied on manual inspection, relying on the naked eye and manual experience to identify defects. Often, only obvious defects such as the self-explosion of some tower base insulators can be detected, and it is difficult to detect faults such as hardware corrosion, loose bolts, and pin shedding. However, this method has a long detection cycle, high labor intensity for workers, and high maintenance costs. At the same time, during the detection process, the degree of fault in the transmission line depends on the judgment of the workers based on their own experience and handheld devices, which makes the professional ability and technical requirements of the workers high. When the workers themselves lack experience and there are errors in the field of view of manual detection, the accuracy of line inspection will be greatly reduced.

[0004] With the development of drone technology, the use of drone inspection has saved more manpower and financial resources and has been widely used. Drones use computer vision systems to capture images of transmission lines. By performing target detection and defect recognition on inspection images, intelligent inspection of transmission lines can be achieved. Since drone inspections are mostly carried out in the field, the image quality is affected by fog and haze. Haze weather will increase the noise of transmission line inspection images, reduce brightness and contrast, and lose detail information, significantly reducing image quality. Due to the serious loss of feature information in foggy images, the accuracy of target detection algorithms decreases, which reduces the accuracy of defect recognition in transmission line images. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for identifying power transmission line faults based on deep learning, so as to solve the problem that the accuracy of target detection algorithm decreases due to serious loss of image feature information in fog, thereby reducing the accuracy of defect recognition in transmission line images.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for identifying faults in a power transmission line based on deep learning, the method comprising:

[0008] Obtain a monitoring image of the transmission line, preprocess the monitoring image of the transmission line to obtain a target image;

[0009] Solve the light value of the target image based on the enhanced light value estimation model to obtain the atmospheric light value of the target image;

[0010] Perform multi-level feature extraction on the target image based on the dense pyramid network to obtain the transmittance map of the target image;

[0011] Input the atmospheric light value of the target image and the transmittance map of the target image into the atmospheric scattering model to obtain a defogged image;

[0012] Input the defogged image into a pre-constructed fault identification model, and output the fault type of the power transmission line.

[0013] Preferably, the dense pyramid network includes: a first encoder and a first decoder, and the interiors of the first encoder and the first decoder are both composed of densely connected dense blocks and multi-level pyramid pooling modules.

[0014] Preferably, the enhanced light value estimation model includes: an enhanced U-net network, an attention network, and a fusion network;

[0015] The enhanced U-net network is used for: extracting features from the target image to obtain a light value feature matrix;

[0016] The attention network is used for: transforming the features of the target image based on the attention mechanism to obtain an initial feature matrix; extracting high-frequency information from the initial feature matrix to obtain a high-frequency feature matrix; fusing the high-frequency feature matrix with the initial feature matrix to obtain an attention feature matrix;

[0017] The fusion network is used for: fusing the attention feature matrix and the light value feature matrix to obtain an atmospheric light value feature map, and using the atmospheric light value feature map as the atmospheric light value of the target image.

[0018] Preferably, extracting high-frequency information from the initial feature matrix to obtain a high-frequency feature matrix includes:

[0019] Construct a plurality of successively connected first residual modules; wherein, the output of the m-th first residual module is used as the input of the (m + 1)-th first residual module, and m is a positive integer;

[0020] In each first residual module, the input matrix of the first residual module is convolved to obtain a first matrix;

[0021] The first matrix is subjected to global average pooling to obtain a first low-channel matrix;

[0022] The first low-channel matrix is deconvolved to obtain a first low-channel weight matrix;

[0023] The first low-channel weight matrix is multiplied by the first low-channel matrix to obtain a second matrix, and the second matrix is used as the output matrix of the first residual module; wherein, the initial feature matrix is used as the input matrix of the first first residual module, and the second matrix output by the last first residual module is used as the high-frequency feature matrix.

[0024] Preferably, the enhanced U-net network includes: a second encoder, a feature recovery module, and a second decoder;

[0025] Feature extraction is performed on the target image to obtain a light value feature matrix, including:

[0026] Based on the second encoder, multi-level downsampling is performed on the target image to obtain a downsampled feature matrix;

[0027] Based on the feature recovery module, multi-level feature recovery is performed on the downsampled feature matrix to obtain recovery features;

[0028] Based on the second decoder, multi-level upsampling and feature enhancement are performed on the recovery features to obtain a light value feature matrix.

[0029] Preferably, based on the second decoder, multi-level upsampling and feature enhancement are performed on the recovery features to obtain a light value feature matrix, including:

[0030] The input sampling matrix for each level of upsampling is initialized, and the output of the j-th level of upsampling is deconvolved as the input of the (j - 1)-th level of upsampling, and the output of deconvolving the recovery features is used as the input of the highest level of upsampling, where j is a positive integer greater than or equal to 2;

[0031] The input of the j-th level of upsampling is deconvolved to obtain the j-th level of sampled features;

[0032] The downsampled feature matrix with the same feature dimension as the j-th level of sampled features and the j-th level of sampled features are fused and enhanced to obtain the j-th level of enhanced features;

[0033] The j-th level of enhanced features is convolved to obtain the convolved j-th level of enhanced features, and the convolved j-th level of enhanced features is used as the output of the j-th level of upsampling; the convolved first level of enhanced features output by the first level of upsampling is used as the light value feature matrix.

[0034] Preferably, the feature restoration module includes a plurality of connected second residual modules, and the output of the nth second residual module serves as the input of the (n + 1)th second residual module, where n is a positive integer;

[0035] Performing multi-level feature restoration on the downsampled feature matrix based on the feature restoration module to obtain restored features, including:

[0036] In each second residual module, convolving the input matrix of the second residual module to obtain a third matrix;

[0037] Performing global average pooling on the third matrix to obtain a second low-channel matrix;

[0038] Performing deconvolution on the second low-channel matrix to obtain a second low-channel weight matrix;

[0039] Multiplying the second low-channel weight matrix by the second low-channel matrix to obtain a fourth matrix, and using the fourth matrix as the output matrix of this second residual module; among them, using the downsampled feature matrix as the input matrix of the first second residual module, and using the fourth matrix output by the last second residual module as the restored feature.

[0040] Preferably, the method further includes:

[0041] Performing game optimization on the atmospheric light value, transmission rate map, and defogged image based on the discriminator to obtain an optimized defogged image.

[0042] Preferably, the method further includes: constructing a fault identification model, including:

[0043] Obtaining power transmission line sample data, and preprocessing the power transmission line sample data to obtain preprocessed power transmission line sample data;

[0044] Training a convolutional neural network based on the preprocessed power transmission line sample data until the convolutional neural network converges to obtain a trained convolutional neural network, and using the trained convolutional neural network as the fault identification model.

[0045] In a second aspect, the present invention provides a power transmission line fault identification system based on deep learning. The system is used to implement the above-mentioned power transmission line fault identification method based on deep learning. The system includes:

[0046] A data acquisition module, configured to acquire a monitoring image of a transmission line, and preprocess the monitoring image of the transmission line to obtain a target image;

[0047] A light value solving module, configured to solve the light value of the target image based on the enhanced light value estimation model to obtain the atmospheric light value of the target image;

[0048] A feature extraction module, which is used to perform multi-level feature extraction on a target image based on a dense pyramid network to obtain a transmittance map of the target image;

[0049] An image defogging module, which is used to input the atmospheric light value of the target image and the transmittance map of the target image into an atmospheric scattering model to obtain a defogged image;

[0050] A fault identification module, which is used to input the defogged image into a pre-constructed fault identification model and output the fault type of the power transmission line.

[0051] Advantageous effects:

[0052] Through the strong light value estimation model of the present invention, the atmospheric light value of the target image can be accurately calculated. At the same time, by using the dense pyramid network to perform multi-level feature extraction on the target image, an accurate transmittance map of the target image can be obtained; then, by substituting the atmospheric light value and the transmittance map into the atmospheric scattering model, the defogged image can be solved to obtain a high-quality defogged image; by performing fault identification on the high-quality defogged image, the accuracy of line fault identification can be improved. Description of the drawings

[0053] The drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present invention, but do not constitute a limitation to the embodiments of the present invention. In the drawings:

[0054] Figure 1 is a flowchart of a method for identifying faults in a power transmission line based on deep learning provided by an embodiment of the present invention;

[0055] Figure 2 is a block diagram of a system for identifying faults in a power transmission line based on deep learning provided by an embodiment of the present invention. Specific embodiments

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the present invention in combination with the drawings and the descriptions of the embodiments or the prior art. Obviously, the following descriptions of the structures of the drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. It should be noted here that the descriptions of these embodiments are used to help understand the present invention, but do not constitute a limitation to the present invention.

[0057] Embodiment 1

[0058] Figure 1 is a flowchart of a method for identifying faults in a power transmission line based on deep learning provided by an embodiment of the present invention, asFigure 1 As shown in Figure 1 , this embodiment provides a method for identifying faults in power transmission lines based on deep learning. The method includes:

[0059] Step S10: Obtain the monitoring image of the transmission line, and preprocess the monitoring image of the transmission line to obtain a target image.

[0060] In this embodiment, an unmanned aerial vehicle (UAV) is used to collect images of the transmission line, and the monitoring image of the transmission line can be obtained. The preprocessing includes: gray processing, image flipping, image cropping, and geometric correction, etc.

[0061] Step S20: Solve the light value of the target image based on the enhanced light value estimation model to obtain the atmospheric light value of the target image.

[0062] In this embodiment, the existing atmospheric light value often uses an empirical value. Since the specific value of the atmospheric light value varies in different scenarios, using the empirical value as the atmospheric light value of the target image for defogging processing will reduce the defogging effect. Therefore, in order to improve the accuracy of the atmospheric light value, the enhanced light value estimation model is used to solve the atmospheric light value of the target image, and an accurate atmospheric light value can be obtained.

[0063] Specifically, the enhanced light value estimation model includes: an enhanced U-net network, an attention network, and a fusion network;

[0064] The enhanced U-net network is used to: extract features from the target image to obtain a light value feature matrix;

[0065] The attention network is used to: perform feature transformation on the target image based on the attention mechanism to obtain an initial feature matrix; extract high-frequency information from the initial feature matrix to obtain a high-frequency feature matrix; fuse the high-frequency feature matrix with the initial feature matrix to obtain an attention feature matrix;

[0066] The fusion network is used to: fuse the attention feature matrix and the light value feature matrix to obtain an atmospheric light value feature map, and use the atmospheric light value feature map as the atmospheric light value of the target image.

[0067] As a further optimization of this embodiment, extracting high-frequency information from the initial feature matrix to obtain a high-frequency feature matrix includes:

[0068] Step a201: Construct a plurality of first residual modules connected in sequence; where the output of the m-th first residual module is used as the input of the (m + 1)-th first residual module, and m is a positive integer;

[0069] Step a202: In each first residual module, perform convolution on the input matrix of the first residual module to obtain a first matrix;

[0070] Step a203: Perform global average pooling on the first matrix to obtain a first low-channel matrix;

[0071] Step a204: Perform deconvolution on the first low-channel matrix to obtain a first low-channel weight matrix;

[0072] Step a205: Multiply the first low-channel weight matrix by the first low-channel matrix to obtain a second matrix, and use the second matrix as the output matrix of the first residual module; wherein, use the initial feature matrix as the input matrix of the first residual module, and use the second matrix output by the last first residual module as the high-frequency feature matrix.

[0073] In this embodiment, through steps a202 to a205, multiple first residual modules can redistribute the features of different channels, and combined with the initial feature matrix obtained by the attention mechanism, the attention network can have strong fitting ability, can retain more high-frequency feature information, and can also accelerate the training speed of the network.

[0074] Since the attention network has strong fitting ability, but is prone to overfitting, resulting in a decrease in the accuracy of the atmospheric light feature map; in order to prevent overfitting, in this embodiment, the enhanced U-net network includes: a second encoder, a feature recovery module, and a second decoder; as a further optimization of this embodiment, feature extraction is performed on the target image to obtain a light value feature matrix, including:

[0075] Step b201: Perform multi-level downsampling on the target image based on the second encoder to obtain a downsampled feature matrix.

[0076] Step b202: Perform multi-level feature recovery on the downsampled feature matrix based on the feature recovery module to obtain a recovered feature.

[0077] In this embodiment, the feature recovery module includes a plurality of connected second residual modules, and the output of the nth second residual module is used as the input of the (n + 1)th second residual module, where n is a positive integer; then step b02 specifically includes the following steps:

[0078] A1: In each second residual module, perform convolution on the input matrix of the second residual module to obtain a third matrix;

[0079] A2: Perform global average pooling on the third matrix to obtain a second low-channel matrix;

[0080] A3: Perform deconvolution on the second low-channel matrix to obtain a second low-channel weight matrix;

[0081] A4: Multiply the second lowest channel weight matrix by the second lowest channel matrix to obtain a fourth matrix, and use the fourth matrix as the output matrix of the second residual module; wherein, use the downsampled feature matrix as the input matrix of the first second residual module, and use the fourth matrix output by the last second residual module as the restored feature.

[0082] In this embodiment, the feature restoration module composed of multiple second residual modules, steps A1 to A4 can retain as many effective features as possible, making the extracted feature information have good robustness, and enabling the enhanced light value estimation model to have a certain generalization effect.

[0083] Step b203: Based on the second decoder, perform multi-level upsampling and feature enhancement on the restored feature to obtain a light value feature matrix.

[0084] Specifically, performing multi-level upsampling and feature enhancement on the restored feature based on the second decoder to obtain a light value feature matrix includes:

[0085] B1: Initialize the input sampling matrix for each level of upsampling. Use the output of the deconvolution of the j-th level of upsampling as the input of the (j - 1)-th level of upsampling, and use the output of the deconvolution of the restored feature as the input of the highest level of upsampling, where j is a positive integer greater than or equal to 2;

[0086] B2: Perform deconvolution on the input of the j-th level of upsampling to obtain the j-th level of sampling feature;

[0087] B3: Fuse and enhance the downsampled feature matrix with the same feature dimension as the j-th level of sampling feature and the j-th level of sampling feature to obtain the j-th level of enhanced feature;

[0088] B4: Perform convolution on the j-th level of enhanced feature to obtain the convolved j-th level of enhanced feature, and use the convolved j-th level of enhanced feature as the output of the j-th level of upsampling; use the convolved first level of enhanced feature output by the first level of upsampling as the light value feature matrix.

[0089] In this embodiment, during the upsampling process, add and fuse the downsampled feature matrix in the downsampling with the corresponding sampling feature to enhance the sampling feature of the upsampling, thereby enhancing the feature restoration in the decoder.

[0090] In this embodiment, the target image is downsampled and upsampled through an enhanced U-net network to retain the important feature information extracted during the sampling process to the greatest extent, and then combined with the strong fitting ability of the attention network to further improve the robustness of the enhanced light value estimation model. The enhanced light value estimation model can output an atmospheric light map with high accuracy.

[0091] Step S30: Perform multi-level feature extraction on the target image based on the dense pyramid network to obtain the transmittance map of the target image.

[0092] In this embodiment, the dense pyramid network includes: a first encoder and a first decoder, and the interiors of the first encoder and the first decoder are both composed of densely connected dense blocks and multi-level pyramid pooling modules.

[0093] In this embodiment, for the dense pyramid network, first, the dense block is constructed based on the CNN network (Convolutional Neural Networks), and the dense block can effectively extract multi-level features of the image; second, the dense block can enable the entire encoding and decoding model to converge quickly. The encoder and the decoder contain multi-level pyramid pooling modules, and this module completes the estimation of the transmittance map based on the global transmittance information of the image, effectively avoiding the drawback of ignoring the global information due to over-concentration on local detail information, and a relatively accurate transmittance map can be obtained.

[0094] Among them, the first encoder is formed by connecting four dense blocks with a traditional convolutional block, and the output size is 1 / 32 of the original input image. The first decoder after the first encoder is formed by connecting four dense blocks with a convolutional block, and the output size is the same as the original input image. The corresponding modules between the first encoder and the first decoder have a direct connection. Since the feature rate map is not generated by multi-class features extracted from images of different scales, although the densely connected encoder-decoder combines different features of the network, the actually generated transmittance map still does not have the global structural information of multi-scale and multi-features. Considering the above situation, a multi-level pyramid pooling module is added after the encoder-decoder in the embodiment, and different layers of the feature pyramid network fully extract the feature information of the target object in the image, and these feature information will be used for the accurate estimation of the transmittance map. Specifically, in this embodiment, a pyramid pooling module with four-level pooling operations is constructed, and the result sizes output by each level of pooling operation are 1 / 4, 1 / 8, 1 / 16, and 1 / 32 of the original image size respectively. In order to establish a connection between the transmittance map and the original image, it is necessary to upsample the four transmittance maps of different sizes one by one to the original image size, so as to obtain a relatively accurate transmittance map.

[0095] Step S40: Input the atmospheric light value of the target image and the transmittance map of the target image into the atmospheric scattering model to obtain a defogged image.

[0096] In this embodiment, the expression of the atmospheric scattering model is:

[0097] ;

[0098] In the formula, is an image containing fog, is a fog-free image, is the atmospheric light value, is the transmission rate map, and z is the pixel position in the image.

[0099] Therefore, through steps S20 and S30, after calculating the atmospheric light value and the transmission rate map of the target image and substituting them into the atmospheric scattering model, the fog-free image can be calculated. Take the fog-free image as the defogged image.

[0100] As a further optimization of this embodiment, the method further includes: performing game optimization on the atmospheric light value, the transmission rate map, and the defogged image based on a discriminator to obtain an optimized defogged image; in this embodiment, through game optimization, the finally generated transmission rate map and defogged image are as close as possible to the real data, improving the defogging effect.

[0101] Step S50: Input the defogged image into a pre-constructed fault recognition model to output the fault type of the power transmission line.

[0102] As a further optimization of this embodiment, the method further includes: constructing a fault recognition model, including:

[0103] Obtain power transmission line sample data, perform preprocessing on the power transmission line sample data to obtain preprocessed power transmission line sample data; the power transmission line sample data contains labels of defects, such as: conductor faults such as broken wires, loose strands, and insulator damage, insulator faults such as insulator rupture, insulator flashover, and insulator aging, fittings faults such as bolt loosening and poor oxidation contact, tower faults such as tower tilt, tower material corrosion, and tower material fracture, grounding faults such as grounding body corrosion and grounding grid fracture, line short-circuit faults such as ground short-circuit and phase short-circuit, and natural state faults such as ice damage, wind damage, and temperature change;

[0104] Train a convolutional neural network based on the preprocessed power transmission line sample data until the convolutional neural network converges to obtain a trained convolutional neural network, and use the trained convolutional neural network as the fault recognition model.

[0105] The present invention can accurately calculate the atmospheric light value of the target image through the strong light value estimation model, and at the same time use the dense pyramid network to perform multi-level feature extraction on the target image to obtain an accurate transmission rate map of the target image; then substituting the atmospheric light value and the transmission rate map into the atmospheric scattering model can solve the defogged image to obtain a high-quality defogged image; by performing fault recognition on the high-quality defogged image, the accuracy of line fault recognition can be improved.

[0106] Embodiment 2

[0107] Figure 2 is a block diagram of a power transmission line fault identification system based on deep learning provided by an embodiment of the present invention. As Figure 2 shown, this embodiment provides a power transmission line fault identification system based on deep learning. The system is used to implement the power transmission line fault identification method based on deep learning in Embodiment 1. The system includes:

[0108] A data acquisition module, configured to acquire a monitoring image of a power transmission line, preprocess the monitoring image of the power transmission line to obtain a target image;

[0109] A light value solving module, configured to solve the light value of the target image based on an enhanced light value estimation model to obtain the atmospheric light value of the target image;

[0110] A feature extraction module, configured to perform multi-level feature extraction on the target image based on a dense pyramid network to obtain a transmittance map of the target image;

[0111] An image dehazing module, configured to input the atmospheric light value of the target image and the transmittance map of the target image into an atmospheric scattering model to obtain a dehazed image;

[0112] A fault identification module, configured to input the dehazed image into a pre-constructed fault identification model and output the fault type of the power transmission line.

[0113] This embodiment also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the above-mentioned power transmission line fault identification method based on deep learning is implemented.

[0114] The present invention can accurately calculate the atmospheric light value of the target image through the strong light value estimation model. At the same time, by using the dense pyramid network to perform multi-level feature extraction on the target image, an accurate transmittance map of the target image can be obtained; then, by substituting the atmospheric light value and the transmittance map into the atmospheric scattering model, the dehazed image can be solved to obtain a high-quality dehazed image; by performing fault identification on the high-quality dehazed image, the accuracy of line fault identification can be improved.

[0115] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented 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.

[0116] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0117] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A fault identification method for power transmission lines based on deep learning, characterized in that, The method includes: Obtaining a monitoring image of a transmission line, preprocessing the monitoring image of the transmission line to obtain a target image; Solving the light value of the target image based on an enhanced light value estimation model to obtain the atmospheric light value of the target image; Performing multi-level feature extraction on the target image based on a dense pyramid network to obtain the transmission rate map of the target image; Inputting the atmospheric light value of the target image and the transmission rate map of the target image into an atmospheric scattering model to obtain a defogged image; Inputting the defogged image into a pre-constructed fault identification model to output the fault type of the power transmission line; The enhanced light value estimation model includes: an enhanced U-net network, an attention network, and a fusion network; The enhanced U-net network is used for: extracting features from the target image to obtain a light value feature matrix; The attention network is used for: transforming the features of the target image based on an attention mechanism to obtain an initial feature matrix; extracting high-frequency information from the initial feature matrix to obtain a high-frequency feature matrix; fusing the high-frequency feature matrix with the initial feature matrix to obtain an attention feature matrix; The fusion network is used for: fusing the attention feature matrix and the light value feature matrix to obtain an atmospheric light value feature map, and using the atmospheric light value feature map as the atmospheric light value of the target image; Extracting high-frequency information from the initial feature matrix to obtain a high-frequency feature matrix, including: Constructing a plurality of first residual modules connected in sequence; wherein, the output of the m-th first residual module is used as the input of the (m + 1)-th first residual module, and m is a positive integer; In each first residual module, convolving the input matrix of the first residual module to obtain a first matrix; Performing global average pooling on the first matrix to obtain a first low-channel matrix; Performing deconvolution on the first low-channel matrix to obtain a first low-channel weight matrix; Multiplying the first low-channel weight matrix by the first low-channel matrix to obtain a second matrix, and using the second matrix as the output matrix of the first residual module; wherein, using the initial feature matrix as the input matrix of the first first residual module, and using the second matrix output by the last first residual module as the high-frequency feature matrix; The enhanced U-net network includes: a second encoder, a feature recovery module, and a second decoder; Extracting features from the target image to obtain a light value feature matrix, including: Performing multi-level downsampling on the target image based on the second encoder to obtain a downsampled feature matrix; Performing multi-level feature recovery on the downsampled feature matrix based on the feature recovery module to obtain a recovered feature; Performing multi-level upsampling and feature enhancement on the recovered feature based on the second decoder to obtain a light value feature matrix; Performing multi-level upsampling and feature enhancement on the recovered feature based on the second decoder to obtain a light value feature matrix, including: Initializing the input sampling matrix of each level of upsampling, using the output of the j-th level of upsampling after deconvolution as the input of the (j - 1)-th level of upsampling, and using the output of deconvolving the recovered feature as the input of the highest level of upsampling, wherein j is a positive integer greater than or equal to 2; Performing deconvolution on the input of the j-th level of upsampling to obtain the j-th level of sampling feature; Fuse and enhance the downsampled feature matrix with the same feature dimension as the j-th level sampling feature and the j-th level sampling feature to obtain the j-th level enhanced feature; Convolve the j-th level enhanced feature to obtain the convolved j-th level enhanced feature, and use the convolved j-th level enhanced feature as the output of the j-th level upsampling; use the convolved first-level enhanced feature output by the first-level upsampling as the light value feature matrix; The feature recovery module includes a plurality of connected second residual modules, and the output of the n-th second residual module is used as the input of the (n + 1)-th second residual module, where n is a positive integer; Based on the feature recovery module, perform multi-level feature recovery on the downsampled feature matrix to obtain the recovered feature, including: In each second residual module, convolve the input matrix of the second residual module to obtain a third matrix; Perform global average pooling on the third matrix to obtain a second low-channel matrix; Perform deconvolution on the second low-channel matrix to obtain a second low-channel weight matrix; Multiply the second low-channel weight matrix by the second low-channel matrix to obtain a fourth matrix, and use the fourth matrix as the output matrix of this second residual module; among them, use the downsampled feature matrix as the input matrix of the first second residual module, and use the fourth matrix output by the last second residual module as the recovered feature.

2. The method for identifying faults in a power transmission line based on deep learning according to claim 1, characterized in that The dense pyramid network includes: a first encoder and a first decoder, and the interiors of the first encoder and the first decoder are both composed of densely connected dense blocks and multi-level pyramid pooling modules.

3. The method for identifying faults in a power transmission line based on deep learning according to claim 1, wherein, The method further includes: Based on the discriminator, perform game optimization on the atmospheric light value, transmittance map, and defogged image to obtain the optimized defogged image.

4. The method for identifying faults in a power transmission line based on deep learning according to claim 1, wherein, The method further includes: constructing a fault identification model, including: Obtain power transmission line sample data, preprocess the power transmission line sample data to obtain preprocessed power transmission line sample data; Based on the preprocessed power transmission line sample data, train the convolutional neural network until the convolutional neural network converges to obtain a trained convolutional neural network, and use the trained convolutional neural network as the fault identification model.

5. A fault identification system for power transmission lines based on deep learning, the system is used to implement the fault identification method for power transmission lines based on deep learning described in any one of claims 1-4, characterized in that, The system includes: A data acquisition module, configured to acquire a monitoring image of a transmission line, preprocess the monitoring image of the transmission line to obtain a target image; A light value solving module, configured to solve the light value of the target image based on the enhanced light value estimation model to obtain the atmospheric light value of the target image; A feature extraction module, configured to perform multi-level feature extraction on the target image based on the dense pyramid network to obtain the transmittance map of the target image; An image defogging module, configured to input the atmospheric light value of the target image and the transmittance map of the target image into the atmospheric scattering model to obtain a defogged image; A fault identification module, configured to input the defogged image into a pre-constructed fault identification model and output the fault type of the power transmission line.

Images