A method, system, device and storage medium for enhancing power grid monitoring video quality

Abstract

The present invention discloses a method, system, device, and storage medium for enhancing the quality of power grid monitoring videos, which are applied to the field of image processing technology. The method comprises determining an electromagnetic interference area; obtaining a plurality of continuous video frames of a monitoring video; performing multi-scale feature extraction processing on all video frames to obtain spatial domain features and frequency domain features of each video frame; processing the spatial domain features using a motion vector-based deconvolution algorithm to obtain a detail-enhanced video frame; processing the frequency domain features to obtain a denoised time domain image; processing the detail-enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature image; and inputting the multi-domain fusion feature image into a visual quality reconstruction model to obtain a quality-enhanced power grid monitoring video. The method provided by the embodiments of the present invention can address the impact of strong electromagnetic interference on video transmission, enhance the quality of power grid monitoring videos, and thereby ensure the application of power grid monitoring videos in power equipment status analysis and fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method, system, device and storage medium for enhancing the quality of power grid monitoring video. Background Art

[0002] With the rapid development of power systems, power grid monitoring videos are increasingly used in equipment status monitoring, fault diagnosis and intelligent inspections.

[0003] However, when power grid monitoring videos are transmitted in strong electromagnetic environments such as substations, strong electromagnetic interference will interfere with the video transmission link. In existing technologies, since electromagnetic interference is not processed, clear videos cannot be obtained for transmission and storage, which seriously restricts the application of power grid monitoring videos in power equipment status analysis and fault diagnosis.

[0004] It can be seen that how to solve the impact of strong electromagnetic waves on video transmission and enhance the quality of power grid monitoring video has become a technical problem that needs to be urgently solved by those skilled in the art. Summary of the Invention

[0005] The present invention provides a method, system, device and storage medium for enhancing the quality of power grid monitoring video, so as to solve the influence of strong electromagnetic fields on video transmission, enhance the quality of power grid monitoring video, and further ensure the application of power grid monitoring video in power equipment status analysis and fault diagnosis.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a method for enhancing the quality of power grid monitoring video, comprising:

[0007] Determine the electromagnetic interference area;

[0008] Acquiring a plurality of continuous video frames of an initial power grid monitoring video in the electromagnetic interference area;

[0009] Performing multi-scale feature extraction processing on all the video frames to obtain spatial domain features and frequency domain features of each of the video frames;

[0010] Processing the spatial domain features using a motion vector-based deconvolution algorithm to obtain detail-enhanced video frames;

[0011] Processing the frequency domain features based on the adaptive frequency band to obtain a denoised time domain image;

[0012] Performing sub-pixel recombining processing on the detail-enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature map;

[0013] The multi-domain fusion feature map is input into the constructed visual quality reconstruction model, and the obtained quality-enhanced video frames are reconstructed into a power grid monitoring video.

[0014] As one preferred solution, the step of obtaining a plurality of continuous video frames of the initial power grid monitoring video in the electromagnetic interference area includes:

[0015] Obtaining a target initial power grid monitoring video in the electromagnetic interference area from a real-time video stream and / or a local server;

[0016] According to a preset time interval, video frames are extracted from the target initial power grid monitoring video to obtain a plurality of continuous video frames.

[0017] As one preferred solution, performing multi-scale feature extraction on all the video frames to obtain spatial domain features and frequency domain features of each video frame includes:

[0018] Inputting all the video frames into a convolutional neural network to extract spatial domain features including at least edge features and texture features;

[0019] All the video frames are subjected to time-frequency decomposition processing using wavelet packet transform to obtain frequency domain features.

[0020] As one preferred solution, the spatial domain features are processed using a motion vector-based deconvolution algorithm to obtain a detail-enhanced video frame. The processing process includes:

[0021] Calculating pixel displacements between adjacent frames in continuous spatial domain features using an optical flow method, and constructing motion vectors based on the pixel displacements;

[0022] A blur kernel is constructed based on the motion vector, and the blur kernel is input into a degradation model constructed by a deconvolution kernel to obtain the detail-enhanced video frame.

[0023] As one preferred solution, the sub-band of the frequency domain feature includes at least low-frequency contour, mid-frequency texture and high-frequency details;

[0024] The processing of the frequency domain features based on the adaptive frequency band to obtain a denoised time domain image includes:

[0025] estimating the noise standard deviation of the frequency domain feature using the median absolute deviation;

[0026] Calculating the energy of each of the sub-bands based on the noise standard deviation;

[0027] Performing adaptive threshold processing based on the energy of each sub-band to obtain a sub-band coefficient after threshold processing;

[0028] Perform inverse wavelet packet transform on the sub-band coefficients to obtain a denoised time domain image.

[0029] As one of the preferred solutions, performing sub-pixel reassembly processing on the detail-enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature map includes:

[0030] Performing dimension elimination processing on the detail-enhanced video frame and the denoised time-domain image to obtain a spatiotemporally aligned detail-enhanced video frame and a denoised time-domain image;

[0031] The spatiotemporally aligned detail-enhanced video frames and denoised time domain images are concatenated along the channel dimension to form a joint feature map;

[0032] The joint feature map is upsampled by sub-pixel convolution to obtain a multi-domain fusion feature map.

[0033] As one preferred solution, the visual quality reconstruction model includes:

[0034] A lightweight feature extraction module, a residual feature enhancement module and a reconstruction module, wherein the lightweight feature extraction module and the residual feature enhancement module are cascade-connected.

[0035] Another embodiment of the present invention provides a system for enhancing the quality of power grid monitoring video, including:

[0036] A determination module, used for determining an electromagnetic interference area;

[0037] An acquisition module, configured to acquire a plurality of continuous video frames of an initial power grid monitoring video in the electromagnetic interference area;

[0038] An extraction module, configured to perform multi-scale feature extraction processing on all the video frames to obtain spatial domain features and frequency domain features of each of the video frames;

[0039] A spatial domain processing module, configured to process the spatial domain features using a motion vector-based deconvolution algorithm to obtain a detail-enhanced video frame;

[0040] A frequency domain processing module, configured to process the frequency domain features based on an adaptive frequency band to obtain a denoised time domain image;

[0041] a recombining module, configured to perform sub-pixel recombining processing on the detail-enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature map;

[0042] A generation module is used to input the multi-domain fusion feature map into the constructed visual quality reconstruction model, and reconstruct the power grid monitoring video based on the obtained quality-enhanced video frames.

[0043] Another embodiment of the present invention provides a power grid monitoring video quality enhancement device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the power grid monitoring video quality enhancement method described above.

[0044] Yet another embodiment of the present invention provides a computer-readable storage medium storing a computer program, wherein when the device where the computer-readable storage medium is located executes the computer program, the power grid monitoring video quality enhancement method described above is implemented.

[0045] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0046] The present invention determines an electromagnetic interference region; obtains several consecutive video frames of an initial power grid monitoring video in the electromagnetic interference region; performs multi-scale feature extraction on all the video frames to obtain spatial and frequency domain features of each video frame; processes the spatial features using a motion vector-based deconvolution algorithm to obtain a detail-enhanced video frame; processes the frequency features based on an adaptive frequency band to obtain a denoised time domain map; performs sub-pixel reconstruction on the detail-enhanced video frame and the denoised time domain map to obtain a multi-domain fusion feature map; inputs the multi-domain fusion feature map into a constructed visual quality reconstruction model, and reconstructs the power grid monitoring video based on the obtained quality-enhanced video frames. Compared with the prior art, the present invention determines the electromagnetic interference region, collects a temporally consecutive sequence of frames in the interfered region, extracts spatial and frequency domain features of the video frames, processes the spatial features to obtain image details and structure, and processes the frequency features to effectively remove noise of specific frequencies, such as electromagnetic interference noise. The spatial and frequency domain processing results are then fused and input into the visual quality reconstruction model to obtain a final quality-enhanced power grid monitoring video. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 1 is a flow chart of a method for enhancing the quality of power grid monitoring video in one embodiment of the present invention;

[0048] Figure 2 is a schematic diagram of a visual quality reconstruction model in one embodiment of the present invention;

[0049] Figure 3 is a structural diagram of a power grid monitoring video quality enhancement system in one embodiment of the present invention;

[0050] Figure 4 It is a structural diagram of a power grid monitoring video quality enhancement device in one embodiment of the present invention.

[0051] Reference numerals:

[0052] Among them, 11, determination module; 12, acquisition module; 13, extraction module; 14, spatial domain processing module; 15, frequency domain processing module; 16, reorganization module; 17, generation module; 21, processor; 22, memory. DETAILED DESCRIPTION

[0053] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0054] In the description of this application, the terms "first," "second," "third," etc. are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first," "second," "third," etc. may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0055] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be a communication between the two components. The terms "vertical", "horizontal", "left", "right", "up", "down" and similar expressions used herein are for illustrative purposes only, and do not indicate or imply that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0056] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this application have the same meanings as those commonly understood by those skilled in the art. The terms used in this specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. Those skilled in the art will understand the specific meanings of the above terms in this application in specific circumstances.

[0057] With the rapid development of power systems, power grid monitoring videos are increasingly used in equipment status monitoring, fault diagnosis and intelligent inspections.

[0058] However, when power grid monitoring videos are transmitted in strong electromagnetic environments such as substations, strong electromagnetic interference will interfere with the video transmission link. In existing technologies, since electromagnetic interference is not processed, clear videos cannot be obtained for transmission and storage, which seriously restricts the application of power grid monitoring videos in power equipment status analysis and fault diagnosis.

[0059] It can be seen that how to solve the impact of strong electromagnetic waves on video transmission and enhance the quality of power grid monitoring video has become a technical problem that needs to be urgently solved by those skilled in the art.

[0060] To this end, an embodiment of the present invention provides a method for enhancing the quality of power grid monitoring video. For details, see Figure 1-Figure 2 , Figure 1 FIG. 1 is a flow chart of a method for enhancing the quality of power grid monitoring video in one embodiment of the present invention. Figure 2 FIG. 1 is a schematic diagram of a visual quality reconstruction model in one embodiment of the present invention, wherein the method includes:

[0061] S1: Determine the electromagnetic interference area;

[0062] S2: Acquire several consecutive video frames of the initial power grid monitoring video in the electromagnetic interference area;

[0063] S3: performing multi-scale feature extraction processing on all the video frames to obtain spatial domain features and frequency domain features of each video frame;

[0064] S4: Processing the spatial domain features using a motion vector-based deconvolution algorithm to obtain a detail-enhanced video frame;

[0065] S5: Processing the frequency domain features based on the adaptive frequency band to obtain a denoised time domain image;

[0066] S6: performing sub-pixel recombining processing on the detail-enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature map;

[0067] S7: Inputting the multi-domain fusion feature map into the constructed visual quality reconstruction model, and reconstructing the obtained quality-enhanced video frames into a power grid monitoring video.

[0068] According to knowledge of power systems, electromagnetic interference mainly comes from the electromagnetic fields generated by power equipment during operation, such as high-voltage transmission lines, switch operations, insulation faults, etc. Common interference areas include at least the near-field areas of transformers and reactors, the vicinity of high-voltage transmission lines, locations with corona discharge, and inside the valve hall of converter stations.

[0069] An electromagnetic sensor array is deployed in the above interference area to determine the electromagnetic interference area affected by the electromagnetic interference.

[0070] Acquire several continuous video frames of the initial power grid monitoring video in the electromagnetic interference area, mainly obtain the target initial power grid monitoring video from the real-time video stream and / or the local server, set the time interval according to the interference time, extract the video frames of the target initial power grid monitoring video, and obtain several continuous video frames.

[0071] Specifically, for switch operation interference, the time interval is set from 2 seconds before the interference operation to 5 seconds after the operation; for discharge interference, the time interval is set according to the sensor trigger duration. This is mainly to ensure that the frame sequence covers the entire life cycle of electromagnetic interference generation, persistence, and dissipation, and accurately locks the interference-sensitive period to avoid invalid data occupying computing resources.

[0072] Specifically, the video frame extraction method is determined by the time interval. For short interference segments less than 30 seconds, full frame rate extraction is performed. For long interference segments greater than or equal to 30 seconds, frames are extracted at an interval of 1 / 3 of the original frame rate to reduce redundancy.

[0073] After obtaining the continuous video frames in step S2, multi-scale feature extraction processing is performed on all video frames to obtain spatial domain features and frequency domain features of each video frame. The specific feature extraction process includes: inputting all the video frames into a convolutional neural network to extract spatial domain features having at least edge features and texture features; using wavelet packet transform to perform time-frequency decomposition processing on all the video frames to obtain frequency domain features.

[0074] Specifically, the video frames are uniformly scaled to a fixed size (such as 512×512 pixels) and normalized to the range of [0,1] to eliminate interference from brightness differences. The normalized video frames are processed using a convolutional neural network. The convolutional neural network includes shallow convolution, deep convolution, and spatial attention modules. The shallow convolution extracts edge features, and the shallow convolution uses 3×3 and 5×5 parallel convolution kernel groups to extract high-frequency details such as device contours and wire edges; the deep convolution extracts texture features through cascaded residual blocks, such as macroscopic patches formed by the oxide layer on the surface of the wire; the spatial attention module assigns a 2-3 times weight to the contact area of ​​the circuit breaker contact to suppress background interference, and finally extracts spatial domain features with at least edge features and texture features.

[0075] The acquisition process of frequency domain features specifically includes: converting the RGB channel into YUV space, decomposing only the luminance component (Y) to reduce chromatic interference, and then performing symmetrical boundary extension on each frame to avoid wavelet transform edge effects.

[0076] Then, a multi-layer full-subtree decomposition is performed using the Daubechies wavelet basis to generate multi-band time-frequency coefficients. Threshold noise reduction is implemented in the low-frequency subbands for power frequency interference positioning, and the impulse noise modulus maximum is detected in the highest-frequency subband, while retaining the characteristic frequency band of the device status. The main purpose of this step is to accurately separate the frequency domain fingerprint of electromagnetic interference, fully retain the equipment operation characteristics, and provide key prior information such as frequency band energy distribution and noise location for subsequent space-frequency domain joint denoising, thereby achieving physical decoupling of electromagnetic interference from the equipment's intrinsic signal.

[0077] The sub-band of the frequency domain feature includes at least low-frequency contour, mid-frequency texture and high-frequency details.

[0078] The Dobechy wavelet basis has compact support and orthogonality, and is a time-frequency analysis tool widely used in signal processing.

[0079] In step S4, the spatial domain features are processed using a motion vector-based deconvolution algorithm to obtain detail-enhanced video frames, specifically including: using an optical flow method to calculate the pixel displacement between adjacent frames in continuous spatial domain features, and constructing a motion vector based on the pixel displacement; constructing a blur kernel based on the motion vector, and inputting the blur kernel into a degradation model constructed by a deconvolution kernel to obtain the detail-enhanced video frame.

[0080] Specifically, the optical flow method is a commonly used method for calculating pixel displacement between adjacent frames. Preferably, the optical flow method adopts the Horn-Schunck optical flow method or the Lucas-Kanade optical flow method. The Horn-Schunck optical flow method is based on the global smoothness assumption and solves the motion field through variational optimization, which is suitable for the overall movement of the equipment; the Lucas-Kanade optical flow method is based on local window linear constraints and calculates the displacement of sparse feature points. It is suitable for local high-speed motion tracking such as the opening and closing of circuit breaker contacts.

[0081] The optical flow method is used to calculate the pixel displacement between adjacent frames, generate a motion vector field, and accurately capture the mechanical motion trajectory of the device. A non-uniform blur kernel is constructed based on the direction and amplitude of the motion vector to simulate the dynamic blur degradation process caused by electromagnetic interference or high-speed motion. The blur kernel is input into the degradation model constructed by the deconvolution kernel. Combined with the device geometric constraints, sub-pixel details are iteratively restored.

[0082] It should be noted that the blur kernel H is the core parameter of the degradation model and participates in the deconvolution solution.

[0083] This method can effectively eliminate motion blur and improve detail quality, meeting the stringent requirements for detail enhancement in power equipment status video analysis.

[0084] In step S5, the frequency domain features are processed based on the adaptive frequency band to obtain a denoised time domain image, including estimating the noise standard deviation of the frequency domain features using the median absolute deviation; calculating the energy of each sub-band based on the noise standard deviation; performing adaptive threshold processing based on the energy of each sub-band to obtain sub-band coefficients after threshold processing; and performing inverse wavelet packet transform processing on the sub-band coefficients to obtain a denoised time domain image.

[0085] The median absolute deviation is used to estimate the noise standard deviation of the frequency domain feature, providing a baseline noise intensity for subsequent threshold processing. In addition, it can also provide a robust estimation of the noise level to prevent outliers (such as pulse interference) from affecting the results.

[0086] The energy of each sub-band, i.e., signal strength, is quantified based on the obtained noise standard deviation. The signal-dominated frequency band and the noise-dominated frequency band are distinguished by the signal strength. The high energy frequency band is the signal-dominated frequency band, and the low energy frequency band is the noise-dominated frequency band.

[0087] At the same time, based on adaptive threshold processing, the threshold of the high-energy sub-band (signal-dominated) is lowered to reduce signal loss, and the threshold of the low-energy sub-band (noise-dominated) is increased to enhance denoising, and finally the sub-band coefficients after threshold processing are obtained.

[0088] The sub-band coefficients after threshold processing are reconstructed by inverse wavelet packet transform, and the frequency domain denoising results are mapped back to the time domain to synthesize the denoised time domain signal, namely the denoised time domain image. The denoised time domain image retains the overall structure and local details of the signal.

[0089] In this process, frequency domain denoising based on adaptive frequency bands effectively suppresses electromagnetic interference noise through sub-band energy analysis and dynamic threshold shrinkage, while retaining key visual information of power grid equipment. This solution can be integrated into the video transmission preprocessing link to lay a high-quality input foundation for subsequent encoding and error-resistant transmission.

[0090] In step S6, the detail-enhanced video frame and the denoised time domain image are subjected to sub-pixel reorganization processing to obtain a multi-domain fusion feature map, including: performing dimension elimination processing on the detail-enhanced video frame and the denoised time domain image to obtain spatiotemporally aligned detail-enhanced video frame and denoised time domain image; splicing the spatiotemporally aligned detail-enhanced video frame and denoised time domain image along the channel dimension to form a joint feature map; and upsampling the joint feature map through sub-pixel convolution to obtain a multi-domain fusion feature map.

[0091] Specifically, the dimension elimination process includes standardization and alignment, which is mainly to ensure that the spatial domain and temporal domain features are in the same numerical range to avoid subsequent fusion deviations. It is also to align the device contour and the dynamically changing area to prevent edge blurring caused by misalignment, that is, to ensure the spatiotemporal consistency of the detail-enhanced video frame and the denoised time domain map.

[0092] The normalized detail-enhanced video frame and the denoised time domain image are spliced ​​along the channel dimension to form a joint feature map. If the number of channels is different, they are adjusted to the same dimension through 1×1 convolution and then the dimension is spliced. In this process, the spatial domain retains edge / color details, the temporal domain suppresses noise and enhances periodic signals.

[0093] Then, through sub-pixel convolution or sub-pixel reconstruction processing, the spatial domain clear frame and the temporal domain denoised image are deeply fused. Through channel reconstruction, the texture information of the spatial domain and the anti-noise characteristics of the temporal domain are mixed to enhance the expression of key areas.

[0094] In step S7, the multi-domain fusion feature map is input into the constructed visual quality reconstruction model, and the obtained quality-enhanced video frames are reconstructed into a power grid monitoring video, wherein the visual quality reconstruction model includes a lightweight feature extraction module, a residual feature enhancement module and a reconstruction module, wherein the lightweight feature extraction module and the residual feature enhancement module are cascaded.

[0095] Among them, the lightweight feature extraction module includes the first 3×3 convolutional layer (Conv), the first wavelet residual block (WRB), the downsampling processing layer, the first wavelet channel aggregation layer (WCA), the first channel attention layer (CAL), the second wavelet channel aggregation layer (WCA), the upsampling processing layer, the second channel attention layer (CAL) and the second wavelet residual block (WRB).

[0096] The structure of the lightweight feature extraction module combines multiple key sub-modules, such as convolutional layers, residual blocks, dense residual networks, wavelet channel attention (WCA), and channel attention (CAL), to form an efficient feature extraction process that can provide accurate solutions to complex problems in power grid monitoring videos.

[0097] First, the convolutional layer extracts preliminary low-level features and expands the number of channels to form a basic feature representation. Next, the wavelet residual block (WRB) uses a wavelet transform to preserve important edge information while removing redundant features, improving feature representation. This is particularly important in complex environments such as noisy or low-light environments.

[0098] To better model global information, the lightweight feature extraction module uses a downsampling layer to reduce the resolution of the feature map and expand the number of channels. Next, the wavelet channel and wavelet attention mechanism adaptively adjust the feature weights to improve the expressiveness of key areas, which is particularly critical in identifying details of power grid equipment such as insulator cracks and broken conductors.

[0099] Finally, the channel attention mechanism gradually adjusts channel features to enhance key features, ensuring the network focuses on critical details of power grid equipment. After multi-level feature extraction and optimization, the LFE module uses upsampling to restore the feature map resolution and further optimizes the final feature representation using wavelet residual blocks (WRBs), ensuring high resolution, low noise, and rich detail information.

[0100] Overall, the lightweight feature extraction module ensures efficient and accurate feature extraction through the combination of convolution, wavelet residual blocks, attention mechanism and dense residual network, providing high-quality input data for the subsequent residual feature enhancement module and reconstruction module, greatly improving the quality of power grid monitoring videos and the accuracy of equipment monitoring.

[0101] Preferably, the first wavelet residual block, downsampling processing unit, first wavelet channel pooling layer, first channel attention layer, second wavelet channel pooling layer, upsampling processing unit, and second channel attention layer can be replaced by a wavelet U-net. The wavelet U-net (WUNet) further optimizes feature extraction and enhances multi-scale feature expression capabilities.

[0102] Preferably, a dense residual network layer can be added to the first wavelet channel set layer and the first channel attention layer. Using a dense residual network can further optimize the feature extraction depth to enhance the robustness of feature extraction.

[0103] Specifically, the residual feature enhancement module includes a pixel unshuffle layer (PixelUnshuffle), a 3×3 second convolutional layer (Conv), a first multi-scale wavelet residual block (MWRB), a second multi-scale wavelet residual block (MWRB), a third multi-scale wavelet residual block (MWRB), a 3×3 third convolutional layer (Conv) and a pixel reshuffle layer (PixelShuffle).

[0104] The residual feature enhancement module maps the input image to a higher-dimensional feature space through a pixel inverse reconstruction layer and generates a 12-channel feature map, reducing the resolution to support subsequent processing.

[0105] Next, a multi-level wavelet residual block (MWRB) and attention mechanism are employed to enhance the representation of key information in power grid monitoring videos and address the issue of detail loss in complex environments. The RFE module utilizes multi-level wavelet residual units (RFFB 1-3) and cascades wavelet residual blocks to enhance the representation of both local details and global information. This is particularly effective in improving the recognizability of critical areas of power grid equipment, such as cracked insulators and broken conductor strands.

[0106] To further remove redundant information, the convolutional layer optimizes the features, and the pixel reconstruction layer restores the spatial resolution to ensure that the final output video frame has a clear and stable visual effect.

[0107] The lightweight feature extraction module and the residual feature enhancement module are cascaded. They are parallel dual-stream structures, forming a complementary relationship, and jointly providing high-quality feature input for the subsequent reconstruction module, significantly enhancing the detail recovery capability of the power grid monitoring video and effectively improving the accuracy of equipment monitoring in complex environments.

[0108] The reconstruction module includes a 1×1 fourth convolutional layer and a 3×3 fifth convolutional layer.

[0109] During the feature reconstruction phase, the network aggregates and refines the extracted and enhanced features through simple convolutional layers, generating high-quality depth images or video frames. This phase is designed primarily to address the need to improve the quality of power grid monitoring videos. Through efficient feature processing, the network aims to ensure that the final video frames have clear and stable details, facilitating status detection and fault diagnosis of power grid equipment.

[0110] Preferably, the method further comprises optimizing the visual quality reconstruction model using an adaptive moment estimation optimizer.

[0111] In deep learning, the Adam optimizer is an adaptive moment estimation optimizer. Because it combines momentum with adaptive learning rate adjustment, it can use different learning rates for high-frequency details (such as device edges) and low-frequency backgrounds (such as device casings), avoiding manual parameter adjustment. Furthermore, in video quality reconstruction tasks, Adam's momentum term helps the model escape local optimality, such as oversmoothing or oversharpening, and approach the global optimal solution.

[0112] An embodiment of the present invention provides a system for enhancing the quality of power grid monitoring video. Figure 3 , Figure 3 FIG. 1 is a schematic diagram showing a method for enhancing the quality of power grid monitoring video according to one embodiment of the present invention. The system includes:

[0113] A determination module 11 is used to determine an electromagnetic interference area;

[0114] An acquisition module 12 is configured to acquire a plurality of continuous video frames of an initial power grid monitoring video in the electromagnetic interference area;

[0115] An extraction module 13 is configured to perform multi-scale feature extraction processing on all the video frames to obtain spatial domain features and frequency domain features of each video frame;

[0116] A spatial domain processing module 14 is configured to process the spatial domain features using a motion vector-based deconvolution algorithm to obtain a detail-enhanced video frame;

[0117] A frequency domain processing module 15 is configured to process the frequency domain features based on an adaptive frequency band to obtain a denoised time domain image;

[0118] a recombining module 16, configured to perform sub-pixel recombining processing on the detail-enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature map;

[0119] The generating module 17 is configured to input the multi-domain fusion feature map into the constructed visual quality reconstruction model, and reconstruct the obtained quality-enhanced video frames into a power grid monitoring video.

[0120] See also Figure 4 , which is a structural diagram of a power grid monitoring video quality enhancement device provided by an embodiment of the present invention. The power grid monitoring video quality enhancement device provided by an embodiment of the present invention includes a processor 21, a memory 22, and a computer program stored in the memory 22 and configured to be executed by the processor 21. When the processor 21 executes the computer program, the steps in the above-mentioned power grid monitoring video quality enhancement method embodiment are implemented, for example Figure 1or, when the processor 21 executes the computer program, the functions of the modules in the above-mentioned device embodiments are realized, such as the determination module 11.

[0121] Exemplarily, the computer program may be divided into one or more modules, which are stored in the memory 22 and executed by the processor 21 to implement the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the power grid monitoring video quality enhancement device. For example, the computer program may be divided into a determination module 11, an acquisition module 12, an extraction module 13, etc., with the specific functions of each module being as follows:

[0122] A determination module 11 is used to determine an electromagnetic interference area;

[0123] An acquisition module 12 is configured to acquire a plurality of continuous video frames of an initial power grid monitoring video in the electromagnetic interference area;

[0124] An extraction module 13 is configured to perform multi-scale feature extraction processing on all the video frames to obtain spatial domain features and frequency domain features of each video frame;

[0125] A spatial domain processing module 14 is configured to process the spatial domain features using a motion vector-based deconvolution algorithm to obtain a detail-enhanced video frame;

[0126] A frequency domain processing module 15 is configured to process the frequency domain features based on an adaptive frequency band to obtain a denoised time domain image;

[0127] a recombining module 16, configured to perform sub-pixel recombining processing on the detail-enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature map;

[0128] The generating module 17 is configured to input the multi-domain fusion feature map into the constructed visual quality reconstruction model, and reconstruct the obtained quality-enhanced video frames into a power grid monitoring video.

[0129] The power grid monitoring video quality enhancement device may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will appreciate that the schematic diagram is merely an example of a power grid monitoring video quality enhancement device and does not limit the power grid monitoring video quality enhancement device. The power grid monitoring video quality enhancement device may include more or fewer components than shown in the diagram, or may combine certain components or different components. For example, the power grid monitoring video quality enhancement device may also include input and output devices, network access devices, buses, and the like.

[0130] The processor 21 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The processor 21 is the control center of the power grid monitoring video quality enhancement device, and utilizes various interfaces and lines to connect various parts of the entire power grid monitoring video quality enhancement device.

[0131] The memory 22 can be used to store the computer programs and / or modules. The processor 21 implements the various functions of the power grid monitoring video quality enhancement device by running or executing the computer programs and / or modules stored in the memory 22 and accessing the data stored in the memory 22. The memory 22 may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as sound playback or image playback); the data storage area may store data generated based on the use of the mobile phone (such as audio data and a phone book). Furthermore, the memory 22 may include high-speed random access memory (RAM) and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0132] If the module integrated into the power grid monitoring video quality enhancement device is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can also implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a removable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium.

[0133] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a computer-readable storage medium, and when executed, the program can include the processes in the above-described method embodiments. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0134] Accordingly, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to perform the steps of the power grid monitoring video quality enhancement method of the above embodiment, for example Figure 1 Steps S1 to S7 described in .

[0135] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0136] The present invention determines an electromagnetic interference region; obtains several consecutive video frames of an initial power grid monitoring video in the electromagnetic interference region; performs multi-scale feature extraction on all the video frames to obtain spatial and frequency domain features of each video frame; processes the spatial features using a motion vector-based deconvolution algorithm to obtain a detail-enhanced video frame; processes the frequency features based on an adaptive frequency band to obtain a denoised time domain map; performs sub-pixel reconstruction on the detail-enhanced video frame and the denoised time domain map to obtain a multi-domain fusion feature map; inputs the multi-domain fusion feature map into a constructed visual quality reconstruction model, and reconstructs the power grid monitoring video based on the obtained quality-enhanced video frames. Compared with the prior art, the present invention determines the electromagnetic interference region, collects a temporally consecutive sequence of frames in the interfered region, extracts spatial and frequency domain features of the video frames, processes the spatial features to obtain image details and structure, and processes the frequency features to effectively remove noise of specific frequencies, such as electromagnetic interference noise. The spatial and frequency domain processing results are then fused and input into the visual quality reconstruction model to obtain a final quality-enhanced power grid monitoring video.

[0137] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A method for enhancing the quality of power grid monitoring video, characterized in that: include: Determine the electromagnetic interference area; Acquiring a plurality of continuous video frames of an initial power grid monitoring video in the electromagnetic interference area; Performing multi-scale feature extraction processing on all the video frames to obtain spatial domain features and frequency domain features of each of the video frames; Processing the spatial domain features using a motion vector-based deconvolution algorithm to obtain detail-enhanced video frames, including: calculating pixel displacements between adjacent frames in continuous spatial domain features using an optical flow method, constructing motion vectors based on the pixel displacements; constructing a blur kernel based on the motion vector, and inputting the blur kernel into a degradation model constructed using a deconvolution kernel to obtain the detail-enhanced video frames; Processing the frequency domain features based on adaptive frequency bands to obtain a denoised time domain image includes: estimating a noise standard deviation of the frequency domain features using a median absolute deviation; calculating the energy of each sub-band based on the noise standard deviation; performing adaptive threshold processing based on the energy of each sub-band to obtain thresholded sub-band coefficients; and performing inverse wavelet packet transform processing on the sub-band coefficients to obtain a denoised time domain image, wherein the sub-bands of the frequency domain features include at least a low-frequency contour, a mid-frequency texture, and a high-frequency detail. Performing sub-pixel recombining processing on the detail-enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature map; The multi-domain fusion feature map is input into the constructed visual quality reconstruction model, and the obtained quality-enhanced video frames are reconstructed into a power grid monitoring video.

2. The method for enhancing the quality of power grid monitoring video according to claim 1, wherein: The acquiring of a plurality of continuous video frames of the initial power grid monitoring video in the electromagnetic interference area includes: Obtaining a target initial power grid monitoring video in the electromagnetic interference area from a real-time video stream and / or a local server; According to a preset time interval, video frames are extracted from the target initial power grid monitoring video to obtain a plurality of continuous video frames.

3. The method for enhancing the quality of power grid monitoring video according to claim 1, wherein: The performing multi-scale feature extraction processing on all the video frames to obtain spatial domain features and frequency domain features of each of the video frames includes: Inputting all the video frames into a convolutional neural network to extract spatial domain features including at least edge features and texture features; All the video frames are subjected to time-frequency decomposition processing using wavelet packet transform to obtain frequency domain features.

4. The method for enhancing the quality of power grid monitoring video according to claim 1, wherein: The sub-pixel recombining process is performed on the detail enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature map, including: Performing dimension elimination processing on the detail-enhanced video frame and the denoised time-domain image to obtain a spatiotemporally aligned detail-enhanced video frame and a denoised time-domain image; The spatiotemporally aligned detail-enhanced video frames and denoised time domain images are concatenated along the channel dimension to form a joint feature map; The joint feature map is upsampled by sub-pixel convolution to obtain a multi-domain fusion feature map.

5. The method for enhancing the quality of power grid monitoring video according to claim 1, wherein: The visual quality reconstruction model includes: A lightweight feature extraction module, a residual feature enhancement module and a reconstruction module, wherein the lightweight feature extraction module and the residual feature enhancement module are cascade-connected.

6. A power grid monitoring video quality enhancement system, characterized in that: include: A determination module, used for determining an electromagnetic interference area; An acquisition module, configured to acquire a plurality of continuous video frames of an initial power grid monitoring video in the electromagnetic interference area; An extraction module, configured to perform multi-scale feature extraction processing on all the video frames to obtain spatial domain features and frequency domain features of each of the video frames; a spatial domain processing module, configured to process the spatial domain features using a motion vector-based deconvolution algorithm to obtain detail-enhanced video frames, including: calculating pixel displacements between adjacent frames in continuous spatial domain features using an optical flow method, constructing motion vectors based on the pixel displacements; constructing a blur kernel based on the motion vectors, and inputting the blur kernel into a degradation model constructed using a deconvolution kernel to obtain the detail-enhanced video frames; a frequency domain processing module, configured to process the frequency domain features based on adaptive frequency bands to obtain a denoised time domain image, comprising: estimating the noise standard deviation of the frequency domain features using a median absolute deviation; calculating the energy of each sub-band based on the noise standard deviation; performing adaptive threshold processing based on the energy of each sub-band to obtain thresholded sub-band coefficients; and performing inverse wavelet packet transform processing on the sub-band coefficients to obtain a denoised time domain image, wherein the sub-bands of the frequency domain features include at least a low-frequency contour, a mid-frequency texture, and a high-frequency detail; a recombining module, configured to perform sub-pixel recombining processing on the detail-enhanced video frame and the denoised time domain image to obtain a multi-domain fusion feature map; A generation module is used to input the multi-domain fusion feature map into the constructed visual quality reconstruction model, and reconstruct the power grid monitoring video based on the obtained quality-enhanced video frames.

7. A power grid monitoring video quality enhancement device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for enhancing the quality of power grid monitoring video according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein when the device where the computer-readable storage medium is located executes the computer program, the power grid monitoring video quality enhancement method according to any one of claims 1 to 5 is implemented.

Images