4865 results about "Low frequency" patented technology

The invention relates to an underwater image enhancement method based on frequency domain analysis and visual Mama, comprising the following steps: constructing a multistage wavelet transform network, the network being an encoder-decoder structure, the encoder comprising a space-frequency feature fusion module and discrete wavelet transform; the decoder comprises a space-frequency feature fusion module and inverse discrete wavelet transform; an underwater image is decomposed into a low-frequency part and a high-frequency part through discrete wavelet transform in an encoder, spatial feature extraction and frequency feature extraction are conducted on the low-frequency part and the high-frequency part through corresponding space-frequency feature fusion modules based on Mama and frequency domain convolution respectively, and splicing is conducted on the low-frequency part and the high-frequency part along a channel in the decoder according to the sequence of low frequency and high frequency. Reconstruction is carried out through inverse discrete wavelet transform; and inputting the underwater image data into the multistage wavelet transform network to obtain enhanced underwater image data. Compared with the prior art, the method has the advantages of improving the enhancement quality, reducing the complexity, quickly enhancing and the like.

The invention discloses a data management system and method based on distributed cloud storage, and relates to the technical field of distributed cloud storage, and the system comprises an access log collection module, a node state monitoring module, a data thermal evaluation module, a redundancy strategy decision module and a storage execution scheduling module. A fragmentation access behavior is recorded through an access log acquisition module, a node state monitoring module acquires a node resource state in real time, and a data thermal evaluation module constructs a three-dimensional thermal model based on an access frequency, a network hop count and a storage cost to calculate a thermal value and supports dynamic weight adjustment; the redundancy strategy decision module divides the fragments into high frequency, medium frequency and low frequency according to the thermodynamic value, and single copy + local cache, erasure code storage and dynamic adjustment strategies are adopted respectively; the storage execution scheduling module executes fragmentation operation and optimizes reading node selection, and data access efficiency is improved through load awareness and a bandwidth-delay joint model; according to the invention, dynamic balance between data access efficiency and storage cost is realized.

The invention discloses an infrared and visible light image fusion method based on a conditional diffusion model, and belongs to the technical field of image fusion. Compared with a current popular infrared and visible light image fusion algorithm, the algorithm is obviously improved in the aspects of color fidelity, texture detail maintenance and fusion quality, and feature extraction and fusion tasks are completed through combination of a conditional denoising diffusion implicit model and a multi-channel high and low frequency parallel fusion model; the problem that the denoising diffusion implicit model lacks a reference true value of a fusion result when executing an image fusion task is relieved, the unique advantages of the denoising diffusion implicit model in image processing are further shown, and the method is excellent in the aspects of high-quality image generation and mathematical interpretability. The optimization of the existing method and the upgrading of the existing technology are realized.

A network monitor may execute, or communicate with, one or more stored machine learning models that are trained to predict a failure probability for one or more ports and / or links within a network fabric. Systems and methods may monitor a set of ports and / or links to generate predictions for failure probabilities using a first trained model and low frequency telemetry data. For a subset of ports and / or links with failure probabilities exceeding a first threshold, high speed telemetry data may be used by a second trained model to generate predictions for failure probabilities for the subset of ports. Suspicious ports may then be isolated and undergo various remediation and / or monitoring actions prior to de-isolating the isolated ports.

The invention belongs to the technical field of remote sensing image segmentation, and particularly relates to a method for generating an adversarial frequency perception image segmentation network, which comprises the following steps of: 1, acquiring a public remote sensing image data set; 2, inputting the remote sensing image data set into a self-adaptive frequency driving feature enhancement module, and decoupling high-frequency features and low-frequency features; the frequency perception adversarial feature enhancement generates more real high and low frequency features through a generative adversarial network, and the resolution capability of the model is improved; and the CLIP text encoder converts the text description into semantic feature vectors to provide support for subsequent cross-modal matching. According to the remote sensing image semantic segmentation method, innovative design is carried out in the aspects of multi-modal fusion, high and low frequency information extraction, adversarial training and the like for remote sensing image semantic segmentation tasks with complex modals, and the segmentation precision, generalization ability and robustness are effectively improved.

The invention discloses a cross-modal pedestrian re-identification method and system based on wavelet transform, and relates to the technical field of pedestrian re-identification, and the method comprises the steps: constructing a two-stream ResNet-50 backbone network; fusing the features of the shallow layers of the two streams of ResNet-50 backbone networks through an ICB module to obtain fused features; carrying out feature extraction on the fusion features through a WEB branch module to obtain modal sharing features of high and low frequency parts; and a final identification result is obtained through the modal sharing feature output by the dual-branch center guiding loss optimization WEB branch module and the fusion feature output by the ICB module. According to the invention, through the information compensation module, the lost valuable information in the network feature extraction process is compensated by combining the features of the shallow layer of the wavelet transform aggregation network at different stages, and the quality of the final feature vector is improved; and through a wavelet enhancement module, dual-branch center guide loss is used to guide the network to mine modal invariant information in a wavelet sub-graph, so that the model performance is improved.

The invention discloses a cable oscillation ultralow-frequency withstand voltage and partial discharge integrated test method and system, and relates to the technical field of distribution network cable detection, and the method comprises the steps: applying an oscillation ultralow-frequency voltage signal to a tested cable; the oscillation ultra-low frequency voltage signal is optimized and adjusted through a manifold learning algorithm and is used as a common excitation source for a withstand voltage test, dielectric loss detection and partial discharge detection; acquiring a voltage signal and a current signal of the tested cable under the excitation of the oscillation ultralow-frequency voltage signal, analyzing the deviation degree of the voltage signal and the current signal from the pre-established characteristic manifold, and extracting a dielectric loss characteristic signal and a partial discharge pulse signal; calculating a dielectric loss parameter based on the dielectric loss characteristic signal, determining a partial discharge position based on the partial discharge pulse signal, and generating a cable insulation state diagnosis report according to the dielectric loss parameter and the partial discharge position. According to the invention, the cable testing efficiency is improved, the equipment burden and the testing time are reduced, comprehensive evaluation and accurate fault positioning of the insulation state are realized, and a scientific basis is provided for preventive maintenance of a power system.

The invention provides an open earphone audio optimization method based on a self-adaptive compensation algorithm, and aims to improve the audio auditory experience of a user. The method comprises the following steps: firstly, acquiring an environmental noise signal through an earphone microphone, receiving a target audio signal, analyzing auditory perception of a user in a current environment by combining an ISO226 loudness curve and the like, and determining auditory sensitivity of the user to different frequency bands (including high frequency, intermediate frequency and low frequency); and then, enhancing the high-frequency part, deepening the low-frequency part and performing smooth transition compensation on the intermediate-frequency part by using a self-adaptive compensation algorithm. And then, dynamically adjusting the low-frequency part by using a hyperbolic filter, generating an adjusted audio signal, and generating a harmonic signal based on the amplitude and frequency characteristics to further optimize the tone quality. Finally, the optimized audio signal and the generated harmonic signal are provided for a user together, and the effects of automatically optimizing audio output in different environments and enhancing tone quality performance are achieved.

The invention discloses a wind turbine generator concrete tower safety monitoring, early warning and evaluation method, which belongs to the field of wind power, and comprises the following steps: analyzing different stress stages of a concrete tower based on a stress mechanism of the concrete tower; determining the main wind direction and the orientation of the inhaul cable with the lowest frequency as the monitoring orientation; arranging a transverse acceleration sensor and a strain gauge in a monitoring direction; determining an early warning value of the monitored concrete tower drum; acceleration and strain signals are collected, first-order frequency of the inhaul cable and actual strain of concrete are obtained through calculation and compared with an early warning value, and safety monitoring and early warning of the concrete tower drum of the wind turbine generator are achieved. According to the stress characteristics of the wind turbine generator concrete tower tube structure, the inhaul cable frequency and the concrete strain are combined, accurate early warning and evaluation of the inhaul cable force loss and the concrete limit strain are achieved, and the safety of the concrete tower tube structure is effectively guaranteed.

The invention belongs to the technical field of time series prediction, and more specifically relates to a time series data prediction method based on a deep integrated learning model and high and low frequency separation. The method comprises the following steps: collecting time sequence data in a system, and preprocessing the obtained data; vMD variational mode decomposition is carried out on original time sequence data, and a time sequence is decomposed into K mode components with limited bandwidth; combining a maximum information coefficient method and a reconstruction error analysis method to determine an optimal decomposition mode number K; dividing high and low frequency components for all decomposed modes by using a zero-crossing rate and a center frequency; appropriate prediction models are established for the high-frequency component and the low-frequency component respectively; and superposing all modal prediction results to obtain a final time sequence prediction result. According to the method, multiple deep learning models are combined, so that the prediction method has higher flexibility and adaptability, and proper models can be selected for components with different characteristics for prediction, thereby improving the overall performance.

The invention discloses a synthetic aperture quantitative phase imaging method based on mixed digital holography and Fourier lamination, which comprises the following steps of: installing a programmable LED (light-emitting diode) array which is annularly arranged on an off-axis digital holography system for providing inclined illumination of angle transformation, obtaining accurate low-frequency information of an object which contains noise through digital holography, and taking the information as initial estimation; and then, object high frequency is reconstructed through a Fourier lamination algorithm based on intensity measurement, so that synthetic aperture quantitative phase imaging is realized. According to the invention, the accurate low frequency provided by holography is used as initial estimation, so that the accuracy of phase recovery and wavefront aberration reconstruction is ensured, a matched illumination condition does not need to be strictly met, and meanwhile, the number of intensity graphs required by Fourier lamination is reduced; the non-interferometric phase recovery method based on Fourier lamination improves the imaging resolution to an incoherent diffraction limit, and effectively solves the inherent speckle noise problem in digital holography.

The invention discloses a nondestructive testing method and system and a data processing method, and relates to the technical field of nondestructive testing. Comprising the steps of obtaining initial detection data of a detected component; establishing at least one reference model according to the initial detection data, and obtaining simulation data corresponding to each reference model; determining target data according to the initial detection data and simulation data; and obtaining subsequent detection data, and comparing the subsequent detection data with the target data to form an evaluation result. According to the invention, high and low frequency mixed excitation signals are adopted, so that the penetrating power and resolution of detection are improved, surface defects and buried defects can be detected at the same time, and the comprehensiveness and accuracy of detection are remarkably enhanced; a picked complex response signal is decoupled by using an orthogonal phase locking method, a magnetic field signal and an electric field signal are effectively separated, the signal-to-noise ratio and the detection precision are improved, and accurate discrimination of defect types, depths and orientations is realized by establishing a reference model and acquiring analog data.

The embodiment of the invention provides a signal processing circuit, a chip and electronic equipment. The signal processing circuit comprises an analog-to-digital conversion module and a signal amplification module, the first chopping switch module is used for modulating the differential input signal to high frequency under the control of a preset clock signal and inputting the signal to the signal amplification module; the second chopping switch module is used for modulating the differential output signal output by the signal amplification module to a low frequency under the control of a preset control signal and inputting the low frequency to the analog-to-digital conversion module; and the second chopping switch module also enables the analog-to-digital conversion module to establish sampling under the control of a preset control signal. According to the application, the second chopping switch module is multiplexed as a sampling switch of the analog-to-digital conversion module, so that the problem that on-resistance is increased due to the fact that a chopping demodulation switch and the sampling switch are arranged at the same time can be avoided.

Methods and systems for testing and treating spinal cord stimulation (SCS) patients are disclosed. Patients are eventually treated with sub-perception (paresthesia free) therapy. However, supra-perception stimulation is used during “sweet spot searching” during which active electrodes are selected for the patient. This allows sweet spot searching to occur much more quickly and without the need to wash in the various electrode combinations that are tried. After selecting electrodes using supra-perception therapy, therapy is titrated to sub-perception levels using the selected electrodes. Such sub-perception therapy has been investigated using pulses at or below 10 kHz, and it has been determined that a statistically significant correlation exists between pulse width (PW) and frequency (F) in this frequency range at which SCS patients experience significant reduction in symptoms such as back pain. Beneficially, sub-perception stimulation at such low frequencies significantly lowers power consumption in the patient's neurostimulator.

The invention discloses a photovoltaic power generation energy storage device control method based on artificial intelligence, and the method comprises the steps: S1, distributing synchronous sampling nodes at a PV side, a bus side and an energy storage side, collecting six paths of quantities, and obtaining a time sequence matrix through incremental buffering, time correction and abnormity elimination; s2, detecting a locking event through a load power step, performing window interception and differential normalization, and combining reactance and correlation to generate an event space-time diagram; s3, the irradiance and the impact amplitude are superposed, and event features are extracted through multi-scale space-time convolution and high and low frequency coupling; s4, calculating a risk score, estimating tail distribution by using mixed Pareto when the risk score exceeds a threshold, and giving a confidence upper bound; s5, constructing an objective function based on the tail distribution and the event coupling probability; s6, the EMS executes reference power according to SOC amplitude limiting, slope detection triggers compensation or calls a plan, and derating feedback is carried out during border crossing; s7, collecting pressure, residual error and loss, and adaptively updating convolution and scheduling parameters; according to the invention, real-time high-precision prediction of the power peak generated by superposition of minute-level photovoltaic fluctuation and millisecond-level load step is realized.

The invention discloses a variable-frequency ash removal device for improving photovoltaic power generation efficiency, and a control system of the variable-frequency ash removal device can dynamically adjust the output frequency and amplitude of a signal generator according to cleaning requirements. Particularly, when different types of dust are cleaned, the system can automatically adjust the vibration frequency, so that the cleaning process is more accurate. Large-particle dust is effectively removed through low-frequency and large-amplitude vibration, and small-particle dust is finely cleaned through high-frequency micro-vibration. The dynamic adjustment of the frequency and the amplitude ensures that each dust type can be thoroughly removed, and the problem of poor effect caused by fixed frequency in a traditional cleaning method is avoided.

The invention belongs to the technical field of vibration control, and provides a CFD numerical simulation-based TMDI control bridge rigid main beam low-frequency vertical vortex vibration research method, which comprises the following steps of: establishing a rigid main beam-TMDI vertical coupling vibration equation in a wind field, carrying out order reduction processing on the coupling vibration equation by constructing a state equation, embedding a solving process into Fluent software by utilizing a user-defined function, and carrying out low-frequency vertical vortex vibration research on a rigid main beam-TMDI control bridge rigid main beam. And the vertical vortex-induced vibration response of the controlled main beam is obtained by alternately solving the flow field domain and the structural domain. Compared with a theoretical method, the fluid-solid coupling effect can be directly simulated, a vortex-induced force model does not need to be assumed, and the result is more reliable; compared with a wind tunnel test method, various parameters of the structure and the TMDI can be accurately set, physical space constraint is avoided, machining errors do not exist, the method is not limited to low wind speed, and the vibration suppression effect can be verified more conveniently; compared with an on-site actual measurement method, the method does not need to passively wait for vertical vortex vibration of a real bridge for a long time, can obtain richer information, facilitates the development of vibration suppression mechanism analysis, and greatly reduces the research difficulty and cost.

The invention provides an image snow removal method based on dual-frequency domain enhancement, which adopts a discrete wavelet transform technology to decompose high and low frequencies and fuse high-frequency enhancement features and low-frequency enhancement features after the high and low frequencies pass through a feature enhancement network so as to enhance the recovery of detail textures of a snowflake-degraded image. According to the method provided by the invention, the high-frequency characteristics of the snow degradation image are enhanced by using a differential convolutional network, and meanwhile, the serious degradation influence caused by large-scale snowflakes is reduced by using a low-frequency enhancement network, so that the performance of the snow degradation image in different frequency domains during snowflake removal and image restoration is improved. And meanwhile, global modeling is carried out on a seriously degraded area through a degradation sensing routing attention module, so that seriously degraded snowflakes are further removed more thoroughly.

The invention provides a DEM super-resolution reconstruction method and system fusing gradient feature Transform, and belongs to the technical field of image processing, the reconstruction method is composed of a plurality of residual feature fusion blocks, a multi-scale pyramid segmentation Transform layer in the residual feature fusion blocks fuses multi-head self-attention and pyramid segmentation attention, and the multi-scale pyramid segmentation Transform layer fuses the multi-head self-attention and the pyramid segmentation attention. In order to make the reconstructed DEM closer to reality, a double-filtering convolution block is used, high and low frequency features of the DEM are fully extracted through a parallel structure of two different filtering blocks, and finally super-resolution reconstruction is carried out under double constraints of a gradient domain and a height domain. The method adapts to topographic features of different scales and complexities, and solves the problem that DEM details cannot be accurately recovered by a traditional algorithm.

The invention discloses a distribution line broadband waveform sampling method, and belongs to the technical field of power system monitoring. By collecting quantum vacuum fluctuation noise, temperature, humidity and electromagnetic parameters, a quantum environment coupling entropy model is constructed, an optimal sparse base dictionary is dynamically generated, and a noise entropy break variable is calculated. And when the entropy break variable exceeds a threshold value, triggering sub-Nyquist compressed sampling, otherwise, adopting Nyquist rate sampling to realize dynamic matching of sampling resources and signal complexity. A high-frequency component of a compressed signal is reconstructed by combining an adversarial compressed sensing network with a residual network, a low-frequency component is generated based on an original signal, the distribution consistency of high and low frequency entropies is optimized by using a Wasserstein distance, and finally a three-dimensional waveform matrix containing a timestamp, a frequency bin and quantum entropy confidence is output. The problems of low sampling efficiency, poor sparse representation adaptability, slow noise response, inconsistent reconstruction and the like in a non-stationary signal and complex noise environment are solved, and the precision and energy efficiency of broadband waveform sampling are remarkably improved.

The invention relates to the technical field of treatment scheme optimization, and provides an emergency treatment critical patient first-aid path optimization method and system based on big data, and the method comprises the steps: building a standardized event sequence through extracting historical diagnosis and treatment data of a hospital information system, and analyzing the time sequence correlation between events through a graph structure mining technology, key diagnosis and treatment nodes and high and low frequency paths are identified, and a layered first-aid path state transition network is constructed based on the key diagnosis and treatment nodes and high and low frequency paths. When a real-time patient is processed, the system matches an electronic health record of the patient with a network core state, dynamically plans a subsequent standard transfer path, and avoids a low-frequency risk path marked as a taboo, so that an emergency path optimization scheme of risk avoidance and path dynamic optimization is realized, and the treatment efficiency of emergency critical patients is improved. The treatment effect is improved, and the bad prognosis risk is reduced.

The embodiment of the invention relates to the technical field of remote sensing image processing, and provides a panchromatic sharpening method and device for a satellite remote sensing image, electronic equipment and a medium, and the method comprises the steps: obtaining a to-be-processed satellite remote sensing image in a panchromatic mode and a multispectral mode; performing up-sampling processing on the multispectral modal image until the target output size of the target panchromatic sharpening image is reached, and obtaining a multispectral modal first satellite remote sensing image; performing multi-scale feature extraction on the panchromatic modal image to obtain different scale features of the panchromatic modal satellite remote sensing image; performing frequency selectivity enhancement processing on image features of the first satellite remote sensing image and the panchromatic modal image with the same scale as the first satellite remote sensing image to obtain high-frequency features and low-frequency features with different scales; and carrying out feature fusion on the high and low frequency features with the same scale, and carrying out image reconstruction to obtain a target panchromatic sharpening image. Therefore, the detail richness and the visual coordination of the reconstructed image can be improved.

Systems, methods, and devices for generating tunable low frequency signals, such as microwave and radiofrequency signals are disclosed. A first waveguide in optical communication with a first ring resonator can be configured to receive a light input and generate frequencies via optical parametric oscillation. A second waveguide with second ring resonator can receive the frequencies and optically synchronize the frequencies via optical frequency division to generate a low frequency signal. One or more heating elements can be connected to each ring resonator to tune the resonators and reduce noise associated with the low frequency signal.

According to the double-end traveling wave distance measurement wave head calibration method and device suitable for the power distribution network and the medium, a proper wavelet high-frequency component is selected through a characteristic frequency band, and a proper mode extreme point is calibrated to serve as initial estimation of a wave head starting point; constructing a smooth threshold value on the low-frequency component corresponding to the characteristic frequency band based on the preliminary estimation and the hard trigger point; a more accurate wave head starting point is obtained through the self-adaptive smooth threshold value search; and finally, calculating the position of a fault point according to a double-end traveling wave fault location formula. The invention provides a novel principle and method for completing coordinated point marking by taking waveform break point calibration of a wave head as wave head starting point calibration and fully utilizing appropriate-scale high and low frequency components of a wavelet domain, and practices prove that the novel calibration principle and method are not easily influenced by traveling wave attenuation, distortion and noise interference caused by traveling wave dispersion and wave impedance; and meanwhile, the problem of low calibration precision caused by inconsistent wave head characteristic frequency spectrums of the traveling waves at the two ends can be solved.

The invention discloses an overlapped finite element method-based underwater target low and medium frequency sound scattering field forecasting method, which comprises the following steps of: generating a virtual node in an overlapped finite element, distributing a group of quadratic complete shift polynomial basis functions in a local analysis domain of each field node to construct local field approximation, and multiplying the local field approximation by a unit decomposition function of the overlapped finite element, constructing a high-order overlapping finite element approximation function; the method comprises the following steps: cutting off an infinite fluid domain by using an artificial boundary, and introducing a Dirichlet-to-Neumann mapping technology to realize a non-reflection boundary condition. For the problem of underwater target sound scattering in middle and low frequency bands, an overlapped finite element numerical calculation model of elastic target sound scattering is established. According to the method, high-order field variable approximation is realized by using the standard low-order finite element grid, the numerical dispersion error and the numerical anisotropy are effectively inhibited, and the low and medium frequency sound scattering field of the elastic target in the infinite water area can be accurately forecasted.

The invention discloses an ocean-ionized layer ultralow-frequency electromagnetic target detection simulation method based on ADI-FDTD and SIBC, and belongs to the field of computational electromagnetics. The ocean-ionized layer ultralow-frequency electromagnetic target detection simulation method comprises the following steps: introducing a frequency domain surface impedance boundary condition SIBC into an ADI-FDTD model; the ocean-ionosphere simulation model is used for carrying out ocean-ionosphere ultra-low frequency electromagnetic wave attenuation analysis on the ocean-ionosphere ultra-low frequency electromagnetic wave attenuation. According to the method, the frequency domain surface impedance boundary condition (SIBC) is combined with the ADI-FDTD method for the first time, so that the calculation space for analyzing the influence of the seabed on electromagnetic wave propagation is reduced. The calculation requirement is obviously reduced, and the efficiency is improved. In addition, the SIBC model based on the ADI-FDTD method can be used for simulating the scattering effect of electromagnetic waves on a high-conductivity and low-conductivity area interface.

The invention discloses a cross-subject electroencephalogram emotion recognition model training method, a cross-subject electroencephalogram emotion recognition model training system and a cross-subject electroencephalogram emotion recognition method based on spatial interaction characteristics. The model training method comprises the step of providing an electroencephalogram signal emotion recognition method in combination with differential warm-up learning (DWL), a high and low frequency window (HFLFW) and a space interactive Colmogov-Arnod network (SIK). Firstly, electroencephalogram signals are converted into electroencephalogram feature maps through three-dimensional mapping, then a model is warmed up before training through differential warm-up learning, and the ability of distinguishing physiological differences of subjects is learned in advance. And processing local refinement and global structure features of the electroencephalogram feature map through high and low frequency windows respectively. And finally, capturing spatial interaction characteristics of electroencephalogram signals of different channels among different testees through a spatial interaction KAN so as to enhance the modeling ability of the model for different electroencephalogram modes.

The invention discloses a two-dimensional and three-dimensional fused fabric defect detection method, and belongs to the technical field of machine vision recognition. The technical problem of limitation of a traditional two-dimensional machine vision system is solved. According to the technical scheme, the method is characterized in that multi-angle light source imaging is carried out, a detected fabric is polished through light sources at different angles, and two-dimensional images of illumination at different angles are obtained through shooting of a linear array camera; and two-dimensional and three-dimensional information fusion: carrying out pixel-level fusion on the shot two-dimensional grey-scale map and the reconstructed gradient map and depth map with the three-dimensional information, carrying out high and low frequency decomposition and fusion on the two-dimensional information and the three-dimensional information through non-subsampled contourlet transformation, and fusing the two-dimensional defect feature information and the three-dimensional defect feature information into one map. After the non-subsampled contourlet transform decomposes the image, the frequency domain space is divided into wedge-shaped directional sub-bands; defects are identified based on the deep learning model. The method has the effect of improving the accuracy of textile quality detection.

The invention provides a leather defect identification method based on double-branch frequency domain feature decoupling and a dynamic channel enhancement network. For the problem that the characteristics of a leather image are difficult to correctly extract due to the characteristics of complexity and non-uniformity of the leather image, a high and low frequency domain decoupling thought is introduced in the invention to extract main body and detail information respectively; a DCT (Discrete Cosine Transformation) guided high / low-frequency double-branch network is designed to respectively capture a local structure (high-frequency branch) of a microdefect and a continuity feature (low-frequency branch) of a macroscopic texture, a structure in which an SE (Square-and-Excitation) module is embedded in a feature fusion stage is designed, and channel weight and spatial position response are jointly optimized, so that the feature fusion is realized. And the significance expression of the key defect area is enhanced. In some public data sets, the method has better classification performance and higher diagnosis speed.

The invention relates to the technical field of intelligent power grid operation and maintenance, and discloses a power grid node health index assessment method suitable for remote operation and maintenance, and the method comprises the steps: S1, carrying out the fusion collection of multi-source data; s2, dynamic weight self-adaptive distribution is carried out; s3, dynamically correcting the life reference; step S4, layered health index calculation; s5, health state decision making and visualization are carried out; the data weight is automatically adjusted according to the timeliness difference of high and low frequency data through a dynamic weight adaptive distribution mechanism, so that the problem of health score long-term distortion possibly caused by low frequency data updating lag is solved, key parameters can be monitored in real time through a real-time anomaly detection and response mechanism, a parameter overrun early warning threshold value is set, and the early warning accuracy is improved. When the key parameters continuously exceed the limit, an online diagnosis mode is triggered, the data sampling frequency is increased, it is ensured that abnormal conditions can be quickly found and processed, the ability of the system to deal with sudden faults is improved, and the fault risk is reduced.