Power transmission line image processing method and device, chip, electronic equipment and medium

Abstract

The invention discloses a power transmission line image processing method, a power transmission line image processing device, a chip, electronic equipment and a computer readable storage medium, and the processing method comprises the steps: obtaining a power transmission line video, extracting an image frame in the power transmission line video, carrying out the preprocessing of the image frame, and obtaining a to-be-processed power transmission line image, and processing the power transmission line image according to the image processing model to obtain a target power transmission line image meeting the target demand. Thus, image frame extraction can be carried out on the power transmission line video, the image frames are preprocessed to obtain the power transmission line image, the situation that the processing difficulty is high due to the fact that the power transmission line image is too large is avoided, the power transmission line image is processed through the image processing model, and the target power transmission line image meeting the target requirement is output. The target power transmission line image can be used for a subsequent monitoring task, and stable operation of the power transmission line is guaranteed.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method for processing a power transmission line image, a processing device for a power transmission line image, a chip, an electronic device, and a computer-readable storage medium. Background Art

[0002] In related technologies, in order to ensure the stable operation of power transmission lines, monitoring equipment is usually installed near the power transmission lines to achieve real-time monitoring of the power transmission lines. However, when night falls, or when the power transmission lines are hit by storms, heavy rains, etc., the power transmission lines may be blurred or even unrecognizable in the images captured by the monitoring equipment, which cannot meet the needs of real-time monitoring, affecting the stable operation of the power transmission lines. Summary of the invention

[0003] The present application provides a method for processing a power transmission line image, a chip, an electronic device, and a computer-readable storage medium.

[0004] The present application provides a method for processing a power transmission line image, including:

[0005] Get video of power transmission lines;

[0006] Extracting image frames from the power transmission line video;

[0007] Preprocessing the image frame to obtain a power transmission line image to be processed, wherein the preprocessing is configured to normalize pixel values ​​and / or image sizes of the image frame;

[0008] According to the image processing model that has been pre-trained and can be deployed in the electronic device, the transmission line image is processed to obtain the target transmission line image that meets the target requirements, wherein the image processing model is used to perform at least one of defogging, super-resolution processing and dark light enhancement processing on the received image, and the value range of the model parameter of the image processing model is [30*10 6 , 50*10 6 ].

[0009] In the method for processing a power line image provided in the embodiment of the present application, a power line video can be obtained, and image frames in the power line video can be extracted, and the image frames can be pre-processed to obtain a power line image to be processed, and a model parameter value range of [30*10 6 , 50*10 6 ]’s image processing model is used to process the transmission line image, thereby obtaining the target transmission line image that meets the target requirements.

[0010] Thus, in the implementation mode of the present application, the image frame extraction can be performed on the acquired power transmission line video, and the extracted image frame can be preprocessed such as normalization and image size standardization to obtain the power transmission line image to be processed, thereby avoiding the situation that the power transmission line image is too large and the processing difficulty of the power transmission line image is high. Then, the power transmission line image can be pre-trained, can be deployed locally, and the value range of the number of model parameters is [30*10 6 , 50*10 6 ] is used to perform at least one of dehazing, super-resolution and dark light enhancement processing on the transmission line image, thereby outputting a target transmission line image that meets the target requirements, realizing offline processing of the transmission line image, and making the target transmission line image output by the electronic device usable for subsequent monitoring tasks, thereby ensuring the stable operation of the transmission line to a certain extent.

[0011] In certain embodiments of the present application, the image processing model includes a defogging module, and the defogging module includes four first convolutional layers and a first output layer connected in sequence;

[0012] The four first convolutional layers are configured to perform convolution processing on the power transmission line image layer by layer to determine the optical transmittance corresponding to the power transmission line image to be processed;

[0013] The first output layer is configured to perform dehazing processing on the power transmission line image according to the optical transmittance and an atmospheric light intensity parameter, wherein the atmospheric light intensity parameter is determined according to pixel values ​​of pixels in the power transmission line image.

[0014] Thus, in the implementation mode of the present application, the transmission line image can be defogged by the defogging module in the image processing model.

[0015] In some embodiments of the present application, a subsequent first convolutional layer is connected to at least one previous first convolutional layer.

[0016] In some embodiments of the present application, the image processing model includes a super-resolution module, and the super-resolution module includes a first feature extraction unit, a second convolution layer, an activation function layer, a splicing layer, a second feature extraction unit, a third convolution layer and a second output layer connected in sequence;

[0017] The first feature extraction unit is configured to perform feature extraction processing on the power transmission line image to determine a first feature extraction result;

[0018] A second convolution layer, configured to perform linear convolution processing on the first feature extraction result to obtain a first convolution result;

[0019] The activation function layer is configured to perform linear activation processing on the first convolution result to determine a linear activation result;

[0020] The splicing layer is configured to splice the power transmission line image and the linear activation result to determine a first splicing result;

[0021] The second feature extraction unit is configured to perform feature extraction processing on the first splicing result to determine a second feature extraction result;

[0022] The third convolution layer is configured to perform linear convolution processing on the second feature extraction result to obtain a second convolution result;

[0023] The second output layer is configured to perform super-resolution processing on the power transmission line image according to the second convolution result.

[0024] Thus, in the implementation mode of the present application, the transmission line image can be subjected to super-resolution processing through the super-resolution module in the image processing model.

[0025] In some embodiments of the present application, the first feature extraction unit includes a first normalization layer, a first attention layer, a second normalization layer, and a first perception layer connected in sequence;

[0026] The first standardization layer is configured to perform standardization processing on the power transmission line image to determine a first standardization result;

[0027] The first attention layer is configured to determine a first attention score of the first normalized result based on the first normalized result;

[0028] The second normalization layer is configured to normalize the second splicing result to determine a second normalization result, wherein the second splicing result is determined by splicing the first attention score and the power transmission line image;

[0029] The first perception layer is configured to determine the first feature extraction result according to the second normalization result.

[0030] In some embodiments of the present application, the first feature extraction unit includes a third normalization layer, a second attention layer, a fourth normalization layer and a second perception layer connected in sequence;

[0031] The third standardization layer is configured to perform standardization processing on the first splicing result to determine a third standardization result;

[0032] The second attention layer is configured to determine a second attention score of the standardized result according to the third standardized result;

[0033] The fourth normalization layer is configured to perform normalization processing on the third splicing result to determine a second normalization result, wherein the third splicing result is determined by performing splicing processing on the second attention score and the first splicing result;

[0034] The second perception layer is configured to determine the first feature extraction result according to the second normalization result.

[0035] In some embodiments of the present application, the image processing model includes a dark light enhancement module, and the dark light enhancement module includes an encoding unit and a decoding unit connected in sequence;

[0036] The encoding unit is configured to perform encoding processing on the power transmission line image to obtain an encoding result;

[0037] The decoding unit is configured to perform decoding processing on the encoding result to determine the transmission line image after dark light enhancement.

[0038] Thus, in the implementation mode of the present application, the transmission line image can be subjected to dark light enhancement processing through the dark light enhancement module in the image processing model.

[0039] In certain embodiments of the present application, the encoding unit includes a plurality of fourth convolutional layers connected in sequence, the decoding unit includes a plurality of fifth convolutional layers connected in sequence, and each of the fourth convolutional layers is also connected to a fifth convolutional layer.

[0040] In certain embodiments of the present application, the power transmission line image is processed according to an image processing model that is pre-trained and can be deployed in an electronic device to obtain a target power transmission line image that meets the target requirements, including:

[0041] When the current time is within the target time, the power transmission line image is processed according to the dark light enhancement module to obtain a target power transmission line image that meets the preset light intensity requirement, wherein the target time is determined according to a preset sunset time and sunrise time.

[0042] Thus, in the implementation mode of the present application, the electronic device can perform dark light enhancement processing on the transmission line image through the dark light enhancement module when the current time is within the target time, thereby ensuring the effective execution of the dark light enhancement processing.

[0043] The present application provides a device for processing a power transmission line image, including:

[0044] A video acquisition module, used to acquire transmission line video;

[0045] An image extraction module, used to extract image frames from the power transmission line video;

[0046] A preprocessing module, used for preprocessing the image frame to obtain a power transmission line image to be processed, wherein the preprocessing module is configured to normalize pixel values ​​and / or image sizes of the image frame;

[0047] The image processing module is used to process the transmission line image according to the image processing model that has been pre-trained and can be deployed in the electronic device to obtain the target transmission line image that meets the target requirements. The image processing model is used to perform at least one of defogging, super-resolution processing and dark light enhancement processing on the received image. The value range of the model parameter of the image processing model is [30*10 6 , 50*10 6 ].

[0048] Thus, in the implementation mode of the present application, the image frame extraction can be performed on the acquired power transmission line video, and the extracted image frame can be preprocessed such as normalization and image size standardization to obtain the power transmission line image to be processed, thereby avoiding the situation that the power transmission line image is too large and the processing difficulty of the power transmission line image is high. Then, the power transmission line image can be pre-trained, can be deployed locally, and the value range of the number of model parameters is [30*10 6 , 50*10 6 ] is used to perform at least one of dehazing, super-resolution and dark light enhancement processing on the transmission line image, thereby outputting a target transmission line image that meets the target requirements, realizing offline processing of the transmission line image, and making the target transmission line image output by the electronic device usable for subsequent monitoring tasks, thereby ensuring the stable operation of the transmission line to a certain extent.

[0049] In certain embodiments of the present application, the image processing model includes a defogging module, and the defogging module includes four first convolutional layers and a first output layer connected in sequence;

[0050] The four first convolutional layers are configured to perform convolution processing on the power transmission line image layer by layer to determine the optical transmittance corresponding to the power transmission line image to be processed;

[0051] The first output layer is configured to perform dehazing processing on the power transmission line image according to the optical transmittance and an atmospheric light intensity parameter, wherein the atmospheric light intensity parameter is determined according to pixel values ​​of pixels in the power transmission line image.

[0052] Thus, in the implementation mode of the present application, the transmission line image can be defogged by the defogging module in the image processing model.

[0053] In some embodiments of the present application, a subsequent first convolutional layer is connected to at least one previous first convolutional layer.

[0054] In some embodiments of the present application, the image processing model includes a super-resolution module, and the super-resolution module includes a first feature extraction unit, a second convolution layer, an activation function layer, a splicing layer, a second feature extraction unit, a third convolution layer and a second output layer connected in sequence;

[0055] The first feature extraction unit is configured to perform feature extraction processing on the power transmission line image to determine a first feature extraction result;

[0056] A second convolution layer, configured to perform linear convolution processing on the first feature extraction result to obtain a first convolution result;

[0057] The activation function layer is configured to perform linear activation processing on the first convolution result to determine a linear activation result;

[0058] The splicing layer is configured to splice the power transmission line image and the linear activation result to determine a first splicing result;

[0059] The second feature extraction unit is configured to perform feature extraction processing on the first splicing result to determine a second feature extraction result;

[0060] The third convolution layer is configured to perform linear convolution processing on the second feature extraction result to obtain a second convolution result;

[0061] The second output layer is configured to perform super-resolution processing on the power transmission line image according to the second convolution result.

[0062] Thus, in the implementation mode of the present application, the transmission line image can be subjected to super-resolution processing through the super-resolution module in the image processing model.

[0063] In some embodiments of the present application, the first feature extraction unit includes a first normalization layer, a first attention layer, a second normalization layer, and a first perception layer connected in sequence;

[0064] The first standardization layer is configured to perform standardization processing on the power transmission line image to determine a first standardization result;

[0065] The first attention layer is configured to determine a first attention score of the first normalized result based on the first normalized result;

[0066] The second normalization layer is configured to normalize the second splicing result to determine a second normalization result, wherein the second splicing result is determined by splicing the first attention score and the power transmission line image;

[0067] The first perception layer is configured to determine the first feature extraction result according to the second normalization result.

[0068] In some embodiments of the present application, the first feature extraction unit includes a third normalization layer, a second attention layer, a fourth normalization layer and a second perception layer connected in sequence;

[0069] The third standardization layer is configured to perform standardization processing on the first splicing result to determine a third standardization result;

[0070] The second attention layer is configured to determine a second attention score of the standardized result according to the third standardized result;

[0071] The fourth normalization layer is configured to perform normalization processing on the third splicing result to determine a second normalization result, wherein the third splicing result is determined by performing splicing processing on the second attention score and the first splicing result;

[0072] The second perception layer is configured to determine the first feature extraction result according to the second normalization result.

[0073] In some embodiments of the present application, the image processing model includes a dark light enhancement module, and the dark light enhancement module includes an encoding unit and a decoding unit connected in sequence;

[0074] The encoding unit is configured to perform encoding processing on the power transmission line image to obtain an encoding result;

[0075] The decoding unit is configured to perform decoding processing on the encoding result to determine the transmission line image after dark light enhancement.

[0076] Thus, in the implementation mode of the present application, the transmission line image can be subjected to dark light enhancement processing through the dark light enhancement module in the image processing model.

[0077] In certain embodiments of the present application, the encoding unit includes a plurality of fourth convolutional layers connected in sequence, the decoding unit includes a plurality of fifth convolutional layers connected in sequence, and each of the fourth convolutional layers is also connected to a fifth convolutional layer.

[0078] In certain embodiments of the present application, the image processing module is also used to process the transmission line image according to the dark light enhancement module when the current time is within the target time, so as to obtain a target transmission line image that meets the preset light intensity requirements, and the target time is determined based on the pre-set sunset time and sunrise time.

[0079] Thus, in the implementation mode of the present application, the electronic device can perform dark light enhancement processing on the transmission line image through the dark light enhancement module when the current time is within the target time, thereby ensuring the effective execution of the dark light enhancement processing.

[0080] An embodiment of the present application provides a chip, which is used in the above-mentioned method for processing a power transmission line image.

[0081] An embodiment of the present application provides an electronic device, comprising the above-mentioned chip.

[0082] An embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program. When the computer program is executed by one or more processors, the above-mentioned method for processing a power transmission line image is implemented.

[0083] The chip, electronic device and computer-readable storage medium provided in the embodiments of the present application can extract image frames from the acquired transmission line video, and perform preprocessing such as normalization and image size standardization on the extracted image frames to obtain the transmission line image to be processed, thereby avoiding the situation where the transmission line image is too large and the processing difficulty of the transmission line image is high. Then, the transmission line image can be pre-trained and deployed locally, and the value range of the number of model parameters is [30*10 6 , 50*10 6 ] is used to perform at least one of dehazing, super-resolution and dark light enhancement processing on the transmission line image, thereby outputting a target transmission line image that meets the target requirements, realizing offline processing of the transmission line image, and making the target transmission line image output by the electronic device usable for subsequent monitoring tasks, thereby ensuring the stable operation of the transmission line to a certain extent.

[0084] Additional aspects and advantages of the embodiments of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the embodiments of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0086] Figure 1 A schematic flow chart of a method for processing a power transmission line image in certain implementation modes of the present application;

[0087] Figure 2 A schematic diagram of a device for processing a power transmission line image in certain embodiments of the present application;

[0088] Figure 3A schematic diagram of an electronic device in some embodiments of the present application;

[0089] Figure 4 A schematic diagram of application scenarios in certain embodiments of the present application;

[0090] Figure 5 A schematic diagram of application scenarios in certain embodiments of the present application;

[0091] Figure 6 A schematic diagram of application scenarios in certain embodiments of the present application;

[0092] Figure 7 A schematic diagram of a feature extraction unit in some embodiments of the present application;

[0093] Figure 8 A schematic diagram of application scenarios in certain embodiments of the present application;

[0094] Fig. 9 A schematic diagram of application scenarios in certain embodiments of the present application;

[0095] Fig.10 A schematic diagram of application scenarios in certain embodiments of the present application;

[0096] Fig.11 It is a flowchart of a method for processing a power transmission line image in certain implementation modes of the present application. DETAILED DESCRIPTION

[0097] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions from beginning to end. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the embodiments of the present application, and cannot be understood as limiting the embodiments of the present application.

[0098] As an important part of the power system, the safe and stable operation of the transmission line is crucial to the normal holding of social activities in the region. Therefore, in the relevant technology, power workers will install monitoring devices (or monitoring equipment) on the transmission lines, and then the workers can monitor the transmission lines in real time through the monitoring devices.

[0099] Furthermore, in related technologies, in order to ensure effective monitoring of transmission lines and accurate determination of the operating status of transmission lines, image processing tasks such as target detection and defect recognition can be performed on the images collected by the monitoring device, so as to complete real-time monitoring of transmission lines with the assistance of artificial intelligence.

[0100] Furthermore, in order to ensure the effective implementation of image processing tasks such as target detection and defect recognition, ISP (Image Signal Processing) operations are usually performed on the images collected by the monitoring device before performing image processing tasks such as target detection and defect recognition, so as to achieve noise removal, bad pixel removal and white balance of the images collected by the monitoring device, thereby improving the quality of the images collected by the monitoring device, thereby ensuring that image processing tasks such as target detection and defect recognition can be performed based on reliable image data. It can be understood that ISP operations are of great significance in improving image quality, enhancing user experience, and ensuring safe and stable operation of power transmission lines.

[0101] It is also understandable that traditional ISP operations include but are not limited to mosaic removal based on interpolation algorithms, noise reduction based on filtering algorithms, etc. These traditional ISP operations have achieved a high level of image quality improvement, but with the gradual development of the field of computer vision, the defects and limitations of traditional ISP operations have gradually been exposed.

[0102] Specifically, traditional ISP operations usually rely on pre-set fixed parameters, and the execution effect of such ISP operations based on fixed parameters may be difficult to achieve expected results in some complex scenarios or some special scenarios.

[0103] Secondly, under low-light conditions, the effects of traditional ISP operations on noise reduction, contrast enhancement, and low-light reinforcement are relatively limited, and the execution effects of traditional ISP operations are difficult to adapt to the increasingly high image quality requirements.

[0104] Furthermore, traditional ISP operations are mostly implemented using hardware chips, but with the continuous expansion of parameter libraries, the difficulty of hardware debugging and the hardware development cycle are gradually increasing. In addition, the hardware chips that implement traditional ISP operations also face limitations and challenges in terms of chip area occupancy, power consumption, and bandwidth.

[0105] In addition, in terms of algorithm design and function implementation, traditional ISP operations are usually set up for specific application scenarios, and therefore it is difficult to meet the requirements of multi-scenario versatility and flexible algorithm adjustment.

[0106] In general, although traditional ISP operations perform relatively well in certain application scenarios, their limitations are gradually becoming apparent as technology advances and demand changes. Therefore, with the development of AI (Artificial Intelligence) technology, AI technology has been widely and deeply applied in the scenario of power line image processing.

[0107] However, in the relevant technologies, the AI ​​technology applied to the scene of power line image processing supports relatively single functions, such as only supporting defogging or night scene enhancement, etc. Moreover, the implementation of such AI technology is relatively difficult, and is usually based on the joint action of multiple algorithm models. This results in the need for AI-based power line image processing to be set up in a dedicated image processing server, resulting in a corresponding increase in the processing cost of power line images and the impact of network communication.

[0108] Based on the above problems you may encounter, please refer to Figure 1 The present application embodiment provides a method for processing a power transmission line image, comprising:

[0109] 01: Get the video of the power transmission line;

[0110] 02: Extract image frames from the power transmission line video;

[0111] 03: Preprocessing the image frame to obtain a power transmission line image to be processed, wherein the preprocessing is configured to normalize pixel values ​​and / or image sizes of the image frame;

[0112] 04: According to the image processing model that has been pre-trained and can be deployed in electronic devices, the transmission line image is processed to obtain the target transmission line image that meets the target requirements, wherein the image processing model is used to perform at least one of defogging, super-resolution processing and dark light enhancement processing on the received image, and the value range of the model parameter of the image processing model is [30*10 6 , 50*10 6 ].

[0113] See also Figure 2 , the embodiment of the present application provides a transmission line image processing device 100. The transmission line image processing method of the embodiment of the present application can be implemented by the transmission line image processing device 100 of the embodiment of the present application. Specifically, the processing device 100 includes a video acquisition module 110, an image extraction module 120, a preprocessing module 130 and an image processing module 140. Among them, the video acquisition module 110 is used to acquire the transmission line video. The image extraction module 120 is used to extract the image frame in the transmission line video. The preprocessing module 130 is used to preprocess the image frame to obtain the transmission line image to be processed, and the preprocessing configuration is to normalize the pixel value and / or image size of the image frame. The image processing module 140 is used to process the transmission line image according to an image processing model that has been pre-trained and can be deployed in an electronic device to obtain a target transmission line image that meets the target requirements. The image processing model is used to perform at least one of defogging, super-resolution processing and dark light enhancement processing on the received image. The value range of the model parameter quantity of the image processing model is [30*106 , 50*10 6 ].

[0114] The embodiment of the present application also provides a chip. The method for processing the transmission line image of the embodiment of the present application can be implemented by the chip of the embodiment of the present application. Specifically, the chip is used to obtain the transmission line video, and to extract the image frame in the transmission line video, and to pre-process the image frame to obtain the transmission line image to be processed, and to process the transmission line image according to the image processing model that has been pre-trained and can be deployed in the electronic device to obtain the target transmission line image that meets the target requirements. The pre-processing configuration is to normalize the pixel value and / or image size of the image frame. The image processing model is used to perform at least one of defogging, super-resolution processing and dark light enhancement processing on the received image. The value range of the model parameter of the image processing model is [30*10 6 , 50*10 6 ].

[0115] Specifically, in an embodiment of the present application, an electronic device (or a chip in an electronic device) may extract one or more frames of images from a transmission line video to be processed, and perform preprocessing operations such as pixel value normalization and image size standardization on the extracted image frames to obtain the transmission line image to be processed, and call a pre-trained, lightweight image processing model that can be deployed in an electronic device to perform at least one ISP operation of dehazing, super-resolution processing, and dark light enhancement processing on the transmission line image, thereby obtaining a transmission line image that meets target requirements (such as brightness requirements, resolution requirements, etc.).

[0116] It is understood that in the implementation of the present application, the electronic device can be understood as a device that can be used to monitor and shoot a power transmission line, such as a camera. Figure 3 , Figure 3 Schematic diagram of an electronic device in some embodiments of the present application. That is, in the embodiments of the present application, the electronic device 200 may be composed of a camera 201, a chip 202 for implementing an image processing model, and a communication module 203.

[0117] Among them, the camera 201 is used to shoot the transmission line to obtain the corresponding transmission line video. The chip 202 is used to extract one or more image frames in the transmission line video, and pre-process the image frames to obtain the transmission line image to be processed, and call the lightweight image processing model that has been pre-trained and deployed locally to process the transmission line image to output the transmission line image that meets the target requirements. The communication module 203 is used to send the transmission line image output by the chip 202 to the outside world (such as a remote server).

[0118] It should be noted that, in order to ensure that the transmission line is clearly visible, in the embodiment of the present application, the camera 201 has a high shooting accuracy, or in other words, the resolution of the transmission line video captured by the camera 201 is high, and thus, the resolution of each frame of the transmission line video is high. In one example, the resolution of the transmission line video and the image frame in the transmission line video is 2K, that is, 2560×1440.

[0119] It should also be noted that in one example, in order to ensure the richness of image information, the power transmission line video and each image frame in the power transmission line video are raw image encoding data, that is, RAW format data. It can be understood that RAW format data preserves the data information transmitted from the camera sensor to the greatest extent, so the image processing model can perform ISP operations such as dark light enhancement, defogging and super resolution based on the rich image information.

[0120] Furthermore, since the resolution of each frame of the transmission line video is relatively high, if each frame of the transmission line video is processed directly through an image processing model, the image processing model will need to include a large number of model parameters and the computing power required to run the image processing model will be relatively high, making it more difficult for electronic devices to call the image processing model.

[0121] Therefore, in the embodiment of the present application, after obtaining the power transmission line video output by the camera 201, the chip 202 can perform frame extraction processing on the power transmission line video according to a pre-set frame extraction strategy, and store the extracted image frames in a storage unit according to certain storage rules. Then, the extracted and stored image frames are subjected to a series of image preprocessing operations, such as pixel value normalization, image size standardization, etc.

[0122] Further, in the embodiment of the present application, the chip 202 can perform a scaling operation on the image frame of the RAW format data to scale the image frame to a size of 544x960 pixels, and the scaled image frame includes four channels of R, G, B, and G, so that the returned data format is 12 bits, and then the pixels of the four channels of the image frame RGBG are divided by 2047 (2 12 -1), thereby obtaining the power transmission line image to be processed.

[0123] In one example, the chip 202 may be composed of multiple NPU (Neural Processing Unit) cores and multiple CPU (Central Processing Unit) cores. The NPU core is responsible for the reasoning of the image processing model, thereby realizing the ISP operation of the image, and the CPU core is responsible for the control of the electronic device body. The image processing model can be deployed in the system Flash.

[0124] It can also be understood that the communication module 203 can send the transmission line image output by the chip 202 to a remote server, and the remote server then sends the transmission line image to a computer device in a specific organization (such as a power station), or the communication module 203 directly sends the transmission line image output by the chip 202 to a computer device in a specific organization (such as a power station). After receiving the transmission line image, the computer device can control the display component to display the transmission line image. For a clearer description of the embodiments of the present application, please refer to Figure 4 , Figure 4 Schematic diagram of application scenarios in certain embodiments of the present application. That is, the display component 301 of the computer device can display the received transmission line image 302 in the corresponding window interface for the staff to browse, and then know the status of the transmission line and the surrounding of the transmission line, so as to confirm whether the transmission line is entangled by foreign objects (such as balloons, kites, etc.), and confirm whether it is necessary to dispatch front-line personnel to the transmission line for processing.

[0125] In addition, it can be understood that in the embodiment of the present application, the image processing model that can be deployed in the electronic device can be understood as a lightweight model, that is, the value range of the number of model parameters is [30*10 6 , 50*10 6 ] neural network model, so the computing power required to run model reasoning is relatively low. Therefore, compared with the solution that requires setting up a dedicated image processing server to implement AI-based power line image processing, the power line image processing has higher processing efficiency and can realize offline processing of power line image processing, thereby avoiding the situation where the power line image processing is affected by the network environment.

[0126] In one example, the number of model parameters is the sum of the number of weight parameters and the number of bias parameters in image processing.

[0127] It can also be understood that in the embodiments of the present application, the image processing model can be used to perform one or more of the three ISP operations such as haze removal (i.e., dehazing processing), super-resolution processing, and dark light enhancement processing on the received image.

[0128] In summary, in the implementation mode of the present application, the image frame extraction can be performed on the acquired power transmission line video, and the extracted image frame can be preprocessed such as normalization and image size standardization to obtain the power transmission line image to be processed, thereby avoiding the situation where the power transmission line image is too large and the processing difficulty of the power transmission line image is high. Then, the power transmission line image can be pre-trained and deployed locally, and the value range of the number of model parameters is [30*10 6 , 50*10 6 ] is used to perform at least one of dehazing, super-resolution and dark light enhancement processing on the transmission line image, thereby outputting a target transmission line image that meets the target requirements, realizing offline processing of the transmission line image, and making the target transmission line image output by the electronic device usable for subsequent monitoring tasks, thereby ensuring the stable operation of the transmission line to a certain extent.

[0129] In certain embodiments of the present application, the image processing model includes a defogging module, and the defogging module includes four first convolutional layers and a first output layer connected in sequence. The four first convolutional layers are configured to perform convolution processing on the transmission line image layer by layer to determine the optical transmittance corresponding to the transmission line image to be processed. The first output layer is configured to perform defogging processing on the transmission line image according to the optical transmittance and the atmospheric light intensity parameter, and the atmospheric light intensity parameter is determined according to the pixel value of the pixel point in the transmission line image.

[0130] To more clearly illustrate the implementation of this application, please refer to Figure 5 , Figure 5 Schematic diagram of application scenarios in some embodiments of the present application. That is, in the embodiment of the present application, the input power transmission line image (or non-defogged image) 400 is input to the first convolutional layer (i.e. Figure 5 The first output layer (not shown) determines the optical transmittance according to the outputs of the four first convolutional layers, and then processes the power line image 400 in combination with the atmospheric light intensity parameters determined by the power line image 400 to obtain the dehazed power line image 401.

[0131] It should be noted that the atmospheric degradation model plays a vital role in image dehazing. The purpose of image dehazing is to restore a clear, high-contrast image from an image affected by fog. The atmospheric degradation model is the core of the dehazing method based on the physical model, which describes the visual impact of haze on the image. Specifically, the atmospheric degradation model consists of two main parts: direct attenuation and the contribution of ambient light. Direct attenuation explains the reduction in the intensity of light when it passes through a medium (such as haze), while ambient light refers to the influence of scattered light in the surrounding environment. The combined effect of these two causes images taken on foggy days to often appear blurred and distorted in color. In an example, the atmospheric degradation model formula can be found in equation (1), that is:

[0132] I=t×J+(1-t)A (1)

[0133] In formula (1), I is the observed foggy image (i.e., the power line image 400). J is the fog-free image to be restored (i.e., the power line image 401 after defogging). A is the global atmospheric light, i.e., the atmospheric light intensity parameter in the embodiment of the present application. t is the transmittance.

[0134] To solve J, formula (1) can be transformed into formula (2), that is:

[0135]

[0136] It can be understood that there are unknown parameters t and A in formula (2). Among them, A can be obtained by taking an approximate estimate of several pixels in the brightness area of ​​the image. It can also be understood that there is an error in A obtained by taking an approximate estimate of several pixels in the image, so a corresponding error parameter b can be set according to the error, and b can be 0.3, 0.6 and 0.9. Therefore, formula (2) can be transformed into formula (3), that is:

[0137]

[0138] It can be understood that in formula (3), the unknown parameters only include t. In the embodiment of the present application, the parameter t can be learned through a lightweight network structure, or in other words, the relationship between the input image and the parameter t can be learned through a dehazing module (Dense Dehaze Block, DDB) including a 4-layer convolution structure, and then after the model obtains the image input, the corresponding parameter t can be output.

[0139] It can also be understood that, when the parameter t is obtained, according to formula (3), the input data (i.e., the input image, or the power line image 400) can be multiplied by 1 / t, subtracted from A×1 / t, and then added to A and b, thereby obtaining the power line image 401 after dehazing.

[0140] Thus, in the implementation mode of the present application, the transmission line image can be defogged by the defogging module in the image processing model.

[0141] In addition, it can be understood that in the embodiment of the present application, the first convolutional layer of the four layers in the defogging module (i.e. Figure 5 The specific connection method of the convolution layer 1 501, the convolution layer 2 502, the convolution layer 3 503 and the convolution layer 4 504 in the example can be set according to the actual situation. For example, please refer to Figure 5 In an embodiment of the present application, a subsequent first convolutional layer is connected to at least one preceding first convolutional layer.

[0142] That is, in the embodiment of the present application, the four first convolutional layers in the defogging module can be connected based on a dense connection mode. For example, convolutional layer 502 No. 2 is connected to convolutional layer 501 No. 1, convolutional layer 503 No. 3 is connected to convolutional layer 501 No. 1 and convolutional layer 502 No. 2, and convolutional layer 4 is connected to convolutional layer 502 No. 2 and convolutional layer 503 No. 3.

[0143] In an embodiment of the present application, the image processing model includes a super-resolution module, which includes a first feature extraction unit, a second convolutional layer, an activation function layer, a splicing layer, a second feature extraction unit, a third convolutional layer and a second output layer connected in sequence.

[0144] The first feature extraction unit is configured to perform feature extraction processing on the power transmission line image to determine a first feature extraction result. The second convolution layer is configured to perform linear convolution processing on the first feature extraction result to obtain a first convolution result. The activation function layer is configured to perform linear activation processing on the first convolution result to determine a linear activation result. The splicing layer is configured to splice the power transmission line image and the linear activation result to determine a first splicing result. The second feature extraction unit is configured to perform feature extraction processing on the first splicing result to determine a second feature extraction result. The third convolution layer is configured to perform linear convolution processing on the second feature extraction result to obtain a second convolution result. The second output layer is configured to perform super-resolution processing on the power transmission line image according to the second convolution result.

[0145] To more clearly illustrate the implementation of this application, please refer to Figure 6 , Figure 6Schematic diagram of application scenarios in certain embodiments of the present application. That is, in the embodiments of the present application, the power transmission line image (i.e., input image 402) input to the super-resolution module (Super-ResolutionBlock, SRB) can be based on the first feature extraction unit 601, the second convolution layer 602, the activation function layer 603, the splicing layer 604, the second feature extraction unit 605 and the third convolution layer 606, and feature extraction, convolution processing, linear activation, splicing, feature extraction and convolution processing are performed in sequence, and finally the second output layer (not shown) performs super-resolution processing on the input power transmission line image (i.e., input image 402) according to the second convolution result, so that the resolution of the input power transmission line image (i.e., input image 402) is improved, thereby changing to a power transmission line image that meets the resolution requirement (i.e., output image 403).

[0146] Thus, in the implementation mode of the present application, the transmission line image can be subjected to super-resolution processing through the super-resolution module in the image processing model.

[0147] In one example, the second convolution layer 602 and the third convolution layer 606 are both convolution layers with a convolution kernel size of 3×3.

[0148] In one example, the activation function layer 603 is used to implement LeakyRelu (Leaky Rectified linear unit).

[0149] It is understandable that the power transmission line image input to the super-resolution module can be set according to actual conditions. For example, in one example, the power transmission line image input to the super-resolution module is the original image without any ISP operation, such as directly Figure 3 In another example, the transmission line image input to the super-resolution module is the output of the aforementioned defogging module, that is, Figure 5 401 in the figure, in other words, 401 and 402 in the embodiments of the present application may refer to the same power transmission line image that has been defogged.

[0150] It can also be understood that in certain embodiments of the present application, the power transmission line image input to the super-resolution module can be understood as LR (Low Resolution data), and then, the first feature extraction unit 601 and the second feature extraction unit 605 are both used to extract LR features.

[0151] It can be understood that in some embodiments of the present application, the specific structures of the first feature extraction unit 601 and the second feature extraction unit 605 can be set according to actual conditions.

[0152] For example, in certain embodiments of the present application, the first feature extraction unit includes a first standardization layer, a first attention layer, a second standardization layer, and a first perception layer connected in sequence. The first standardization layer is configured to perform standardization processing on the power transmission line image to determine a first standardization result. The first attention layer is configured to determine a first attention score of the first standardization result based on the first standardization result. The second standardization layer is configured to perform standardization processing on the second splicing result to determine a second standardization result, and the second splicing result is determined by splicing the first attention score and the power transmission line image. The first perception layer is configured to determine the first feature extraction result based on the second standardization result.

[0153] And, in certain embodiments of the present application, the first feature extraction unit includes a third standardization layer, a second attention layer, a fourth standardization layer and a second perception layer connected in sequence. The third standardization layer is configured to perform standardization processing on the first splicing result to determine a third standardization result. The second attention layer is configured to determine a second attention score of the standardization result based on the third standardization result. The fourth standardization layer is configured to perform standardization processing on the third splicing result to determine a second standardization result, and the third splicing result is determined by splicing the second attention score and the first splicing result. The second perception layer is configured to determine the first feature extraction result based on the second standardization result.

[0154] To explain the implementation of this application more clearly, please participate Figure 7 , Figure 7 Schematic diagram of a feature extraction unit in some embodiments of the present application. That is, in the embodiments of the present application, the specific structures of the first feature extraction unit 601 and the second feature extraction unit 605 can be as follows: Figure 7 As shown, it includes a pre-normalization layer 607, an attention layer 608, a post-normalization layer 609 and a perception layer 610.

[0155] In one example, both the pre-normalization layer 607 and the post-normalization layer 609 are Layer Normalization.

[0156] In one example, the attention layer 608 is Window-Multihead SelfAttention (W-MSA).

[0157] In one example, the perception layer 610 is a multilayer perceptron (MLP).

[0158] It can be understood that when the pre-normalization layer 607 and the post-normalization layer 609 are both Layer Normalization, the attention layer 608 is the moving window self-attention, and the perception layer 610 is a multi-layer perceptron, the first feature extraction unit 601 and the second feature extraction unit 605 can both be understood as STL (Swin-Transformer Layer).

[0159] Furthermore, it can be understood that the moving window self-attention in STL can close the adaptive receptive field and supplement the feature information learned by the low-level network by paying attention to multi-level spatial information, so as to extract LR features in a targeted and efficient manner, thereby improving the performance of the super-resolution module in the SISR (Single Image Super Resolution) task.

[0160] In certain embodiments of the present application, the image processing model includes a dark light enhancement module, and the dark light enhancement module includes an encoding unit and a decoding unit connected in sequence. The encoding unit is configured to perform encoding processing on the power transmission line image to obtain an encoding result. The decoding unit is configured to perform decoding processing on the encoding result to determine the power transmission line image after dark light enhancement.

[0161] To more clearly illustrate the implementation of this application, please refer to Figure 8 , Figure 8 Schematic diagram of application scenarios in certain embodiments of the present application. That is, in the embodiments of the present application, the power transmission line image 404 input to the dark light enhancement module (Image Enhancement Block, EB) can be encoded based on the encoding module (Encoder) 710 to obtain a down-sampled encoding result, and the encoding result can be up-sampled and decoded by the decoding unit (Decoder) 720 to obtain the dark light enhanced power transmission line image 405.

[0162] Thus, in the implementation mode of the present application, the transmission line image can be subjected to dark light enhancement processing through the dark light enhancement module in the image processing model.

[0163] It can be understood that, in the implementation of the present application, the power line image input to the dark light enhancement module can be set according to the actual situation. For example, in one example, the power line image input to the super-resolution module is the original image without any ISP operation, such as directly through Figure 3 In another example, the transmission line image input to the super-resolution module is the output of the super-resolution module, that is, Figure 6403 in the figure, in other words, 403 and 404 in the embodiments of the present application may refer to the same power transmission line image that has undergone super-resolution processing.

[0164] It can also be understood that in the implementation of the present application, the specific structure and connection mode of the decoding unit 710 and the encoding module 720 can be set according to actual conditions. For example, in one example, the encoding unit 710 includes a plurality of fourth convolutional layers connected in sequence, and the decoding unit includes a plurality of fifth convolutional layers connected in sequence, and each fourth convolutional layer is also connected to a fifth convolutional layer.

[0165] For more information, please refer again to Figure 8 The encoding unit 710 is composed of a convolutional layer No. 4 711, a convolutional layer No. 5 712 and a convolutional layer No. 6 713 connected in sequence, and the decoding unit 720 is composed of a convolutional layer No. 7 721, a convolutional layer No. 8 722 and a convolutional layer No. 9 723 connected in sequence.

[0166] It can be understood that both the encoding unit 710 and the decoding unit 720 are as follows Figure 8 As shown, when composed of three convolutional layers, the dark light enhancement module in the implementation manner of the present application can be understood as a lightweight UNet structure.

[0167] It can also be understood that, in the embodiment of the present application, each fourth convolutional layer is also connected to a fifth convolutional layer. Figure 8 In the figure, convolution layer No. 4 711 is connected to convolution layer No. 9 723, convolution layer No. 5 712 is connected to convolution layer No. 8 722, and convolution layer No. 6 713 is connected to convolution layer No. 7 721.

[0168] Furthermore, it can be understood that based on Figure 8 The connection method shown, or based on Figure 8 The skip connection method shown can make the feature extraction of each convolutional layer more effective and improve the feature engineering capability of the entire module.

[0169] In certain embodiments of the present application, step 04 includes:

[0170] When the current time is within the target time, the transmission line image is processed according to the dark light enhancement module to obtain a target transmission line image that meets the preset light intensity requirement, wherein the target time is determined according to the preset sunset time and sunrise time.

[0171] The image processing module of the implementation mode of the present application is also used to process the transmission line image according to the dark light enhancement module when the current time is within the target time, so as to obtain a target transmission line image that meets the preset light intensity requirements, wherein the target time is determined based on the pre-set sunset time and sunrise time.

[0172] The chip of the embodiment of the present application is also used to process the transmission line image according to the dark light enhancement module when the current time is within the target time, so as to obtain a target transmission line image that meets the preset light intensity requirements, wherein the target time is determined based on the pre-set sunset time and sunrise time.

[0173] Specifically, in the implementation manner of the present application, the electronic device can determine whether it is necessary to perform a dark light enhancement operation on the power transmission line image through the dark light enhancement module according to the current time.

[0174] For example, in one example, the sunrise time is 05:51, the sunset time is 19:09, and the electronic device obtains a power transmission line image through the camera at 12:09. Since 12:09 is not within the period from 19:09 on the same day to 05:51 on the next day, the electronic device can skip the dark light enhancement operation of the power transmission line image. In another example, the sunrise time is 05:51, the sunset time is 19:09, and the electronic device obtains a power transmission line image through the camera at 20:09. Since 20:09 is within the period from 19:09 on the same day to 05:51 on the next day, the electronic device can perform the dark light enhancement operation on the power transmission line image through the dark light enhancement module.

[0175] In addition, it can be understood that the sunset time and the sunrise time can be determined according to the location (or longitude and latitude) of the electronic device that captures the transmission line image.

[0176] Thus, in the implementation mode of the present application, the electronic device can perform dark light enhancement processing on the transmission line image through the dark light enhancement module when the current time is within the target time, thereby ensuring the effective execution of the dark light enhancement processing.

[0177] In certain embodiments of this application, please refer to Fig. 9 , Fig.10 and Fig.11 , Fig. 9 This is a schematic diagram of application scenarios in certain embodiments of the present application. Fig.10 This is a schematic diagram of application scenarios in certain embodiments of the present application. Fig.11 It is a flowchart of a method for processing a power transmission line image in certain implementation modes of the present application.

[0178] Specifically, in the implementation manner of the present application, the electronic device can shoot the power transmission line through a camera to obtain a video image of the power transmission line, and transmit the video image of the power transmission line to the video analysis module.

[0179] Next, the video analysis module can extract at least one image frame from the power transmission line video image, and perform a series of image preprocessing operations on the image frame, including but not limited to normalization, size standardization, etc., so as to obtain a power transmission line image 700 that can be input into the image processing model.

[0180] In one example, the video analysis module performs a scaling operation on the image frame of RAW format data to scale the image frame to 544x960 pixels, and the scaled image frame includes four channels of R, G, B, and G, so that the returned data format is 12 bits, and then the pixels of the four channels of RGBG of the image frame are divided by 2047 (2 12 -1), thereby obtaining a power line image 700 that can be input into an image processing model.

[0181] Next, when the power transmission line image 700 that can be input into the image processing model is obtained, the obtained power transmission line image 700 can first be defogged, that is, the atmospheric fog model parameters are calculated by the defogging module 810 and the power transmission line image 700 is defogged using the atmospheric fog model parameters to obtain the defogging image data (such as Figure 5 401 in the above example).

[0182] Then, the image data after the defogging process is input to the super-resolution module 820 to improve the clarity of the image data, thereby obtaining the image data with improved clarity (such as Figure 6 403 in the above example).

[0183] Then, when the image data with improved clarity is obtained, it can be determined whether a dark light enhancement operation is needed according to the current time. If necessary, the dark light enhancement operation is input to the dark light enhancement module 830 for processing.

[0184] It is understandable that if the current time is within the target time, the dark light enhancement module 830 can perform dark light enhancement processing on the image data with improved clarity to obtain dark light enhanced image data (such as Figure 8 405 in the image processing model), the current round of calling the image processing model ends. If the current time is not within the target time, the dark light enhancement processing can be skipped and the current round of calling the image processing model ends directly.

[0185] Finally, the final "image data 701 after dehazing and clarity improvement" or "image data 701 after dehazing, clarity improvement and dark light enhancement" is used for subsequent transmission line monitoring tasks, such as transmission line foreign body detection, defect detection, etc.

[0186] An embodiment of the present application provides an electronic device, comprising the above-mentioned chip.

[0187] The embodiment of the present application further provides a computer-readable storage medium storing a computer program. When the computer program is executed by one or more processors, the above-mentioned method for processing the power transmission line image is implemented.

[0188] In the description of this specification, the descriptions with reference to the terms "specifically", "further", "particularly", "understandably", etc. are intended to mean that the specific features, structures, materials or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms are not intended to refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0189] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0190] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for processing a power transmission line image, characterized in that: include: Get video of power transmission lines; Extracting image frames from the power transmission line video; Preprocessing the image frame to obtain a power transmission line image to be processed, wherein the preprocessing is configured to normalize pixel values ​​and / or image sizes of the image frame; According to the image processing model that has been pre-trained and can be deployed in the electronic device, the transmission line image is processed to obtain the target transmission line image that meets the target requirements, wherein the image processing model is used to perform at least one of defogging, super-resolution processing and dark light enhancement processing on the received image, and the value range of the model parameter of the image processing model is [30*10 6 , 50*10 6 ].

2. The method according to claim 1, characterized in that: The image processing model includes a defogging module, and the defogging module includes four first convolutional layers and a first output layer connected in sequence; The four first convolutional layers are configured to perform convolution processing on the power transmission line image layer by layer to determine the optical transmittance corresponding to the power transmission line image to be processed; The first output layer is configured to perform dehazing processing on the power transmission line image according to the optical transmittance and an atmospheric light intensity parameter, wherein the atmospheric light intensity parameter is determined according to pixel values ​​of pixels in the power transmission line image.

3. The method according to claim 2, characterized in that A subsequent first convolutional layer is connected to at least one preceding first convolutional layer.

4. The method according to claim 1, characterized in that: The image processing model includes a super-resolution module, which includes a first feature extraction unit, a second convolution layer, an activation function layer, a splicing layer, a second feature extraction unit, a third convolution layer and a second output layer connected in sequence; The first feature extraction unit is configured to perform feature extraction processing on the power transmission line image to determine a first feature extraction result; A second convolution layer, configured to perform linear convolution processing on the first feature extraction result to obtain a first convolution result; The activation function layer is configured to perform linear activation processing on the first convolution result to determine a linear activation result; The splicing layer is configured to splice the power transmission line image and the linear activation result to determine a first splicing result; The second feature extraction unit is configured to perform feature extraction processing on the first splicing result to determine a second feature extraction result; The third convolution layer is configured to perform linear convolution processing on the second feature extraction result to obtain a second convolution result; The second output layer is configured to perform super-resolution processing on the power transmission line image according to the second convolution result.

5. The method according to claim 4, characterized in that The first feature extraction unit includes a first normalization layer, a first attention layer, a second normalization layer and a first perception layer connected in sequence; The first standardization layer is configured to perform standardization processing on the power transmission line image to determine a first standardization result; The first attention layer is configured to determine a first attention score of the first normalized result based on the first normalized result; The second normalization layer is configured to normalize the second splicing result to determine a second normalization result, wherein the second splicing result is determined by splicing the first attention score and the power transmission line image; The first perception layer is configured to determine the first feature extraction result according to the second normalization result.

6. The method according to claim 4, characterized in that The first feature extraction unit includes a third normalization layer, a second attention layer, a fourth normalization layer and a second perception layer connected in sequence; The third standardization layer is configured to perform standardization processing on the first splicing result to determine a third standardization result; The second attention layer is configured to determine a second attention score of the standardized result according to the third standardized result; The fourth normalization layer is configured to perform normalization processing on the third splicing result to determine a second normalization result, wherein the third splicing result is determined by performing splicing processing on the second attention score and the first splicing result; The second perception layer is configured to determine the first feature extraction result according to the second normalization result.

7. The method according to claim 1, characterized in that The image processing model includes a dark light enhancement module, and the dark light enhancement module includes an encoding unit and a decoding unit connected in sequence; The encoding unit is configured to perform encoding processing on the power transmission line image to obtain an encoding result; The decoding unit is configured to perform decoding processing on the encoding result to determine the transmission line image after dark light enhancement.

8. The method according to claim 7, characterized in that The encoding unit includes a plurality of fourth convolutional layers connected in sequence, the decoding unit includes a plurality of fifth convolutional layers connected in sequence, and each of the fourth convolutional layers is further connected to a fifth convolutional layer.

9. The method according to claim 7, characterized in that: The method of processing the power transmission line image according to the image processing model that has been pre-trained and can be deployed in the electronic device to obtain a target power transmission line image that meets the target requirements includes: When the current time is within the target time, the power transmission line image is processed according to the dark light enhancement module to obtain a target power transmission line image that meets the preset light intensity requirement, wherein the target time is determined according to a preset sunset time and sunrise time.

10. A device for processing a power transmission line image, characterized in that: include: A video acquisition module, used to acquire transmission line video; An image extraction module, used to extract image frames from the power transmission line video; A preprocessing module, used for preprocessing the image frame to obtain a power transmission line image to be processed, wherein the preprocessing module is configured to normalize pixel values ​​and / or image sizes of the image frame; The image processing module is used to process the transmission line image according to the image processing model that has been pre-trained and can be deployed in the electronic device to obtain the target transmission line image that meets the target requirements. The image processing model is used to perform at least one of defogging, super-resolution processing and dark light enhancement processing on the received image. The value range of the model parameter of the image processing model is [30*10 6 , 50*10 6 ].

11. The device according to claim 10, characterized in that The image processing model includes a defogging module, and the defogging module includes four first convolutional layers and a first output layer connected in sequence; The four first convolutional layers are configured to perform convolution processing on the power transmission line image layer by layer to determine the optical transmittance corresponding to the power transmission line image to be processed; The first output layer is configured to perform dehazing processing on the power transmission line image according to the optical transmittance and an atmospheric light intensity parameter, wherein the atmospheric light intensity parameter is determined according to pixel values ​​of pixels in the power transmission line image.

12. The device according to claim 11, characterized in that A subsequent first convolutional layer is connected to at least one preceding first convolutional layer.

13. The device according to claim 10, characterized in that The image processing model includes a super-resolution module, which includes a first feature extraction unit, a second convolution layer, an activation function layer, a splicing layer, a second feature extraction unit, a third convolution layer and a second output layer connected in sequence; The first feature extraction unit is configured to perform feature extraction processing on the power transmission line image to determine a first feature extraction result; A second convolution layer, configured to perform linear convolution processing on the first feature extraction result to obtain a first convolution result; The activation function layer is configured to perform linear activation processing on the first convolution result to determine a linear activation result; The splicing layer is configured to splice the power transmission line image and the linear activation result to determine a first splicing result; The second feature extraction unit is configured to perform feature extraction processing on the first splicing result to determine a second feature extraction result; The third convolution layer is configured to perform linear convolution processing on the second feature extraction result to obtain a second convolution result; The second output layer is configured to perform super-resolution processing on the power transmission line image according to the second convolution result.

14. The device according to claim 13, characterized in that The first feature extraction unit includes a first normalization layer, a first attention layer, a second normalization layer and a first perception layer connected in sequence; The first standardization layer is configured to perform standardization processing on the power transmission line image to determine a first standardization result; The first attention layer is configured to determine a first attention score of the first normalized result based on the first normalized result; The second normalization layer is configured to normalize the second splicing result to determine a second normalization result, wherein the second splicing result is determined by splicing the first attention score and the power transmission line image; The first perception layer is configured to determine the first feature extraction result according to the second normalization result.

15. The device according to claim 13, characterized in that The first feature extraction unit includes a third normalization layer, a second attention layer, a fourth normalization layer and a second perception layer connected in sequence; The third standardization layer is configured to perform standardization processing on the first splicing result to determine a third standardization result; The second attention layer is configured to determine a second attention score of the standardized result according to the third standardized result; The fourth normalization layer is configured to perform normalization processing on the third splicing result to determine a second normalization result, wherein the third splicing result is determined by performing splicing processing on the second attention score and the first splicing result; The second perception layer is configured to determine the first feature extraction result according to the second normalization result.

16. The device according to claim 10, characterized in that The image processing model includes a dark light enhancement module, and the dark light enhancement module includes an encoding unit and a decoding unit connected in sequence; The encoding unit is configured to perform encoding processing on the power transmission line image to obtain an encoding result; The decoding unit is configured to perform decoding processing on the encoding result to determine the transmission line image after dark light enhancement.

17. The device according to claim 16, characterized in that The encoding unit includes a plurality of fourth convolutional layers connected in sequence, the decoding unit includes a plurality of fifth convolutional layers connected in sequence, and each of the fourth convolutional layers is further connected to a fifth convolutional layer.

18. The device according to claim 16, characterized in that The image processing module is also used to process the transmission line image according to the dark light enhancement module to obtain a target transmission line image that meets a preset light intensity requirement when the current time is within a target time, and the target time is determined based on a preset sunset time and sunrise time.

19. A chip, characterized in that: The chip is used to execute the method according to any one of claims 1 to 9.

20. An electronic device, characterized in that: Comprising the chip as claimed in claim 19.

21. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by one or more processors, the method according to any one of claims 1 to 9 is implemented.

Images