33results about How to "Improve separability" patented technology

The invention discloses a multi-modal contrast learning fault diagnosis method for a small sample scene, and the method comprises the following steps: carrying out the enhancement of a one-dimensional signal and two-dimensional image fused fault data set through physical simulation for the small sample scene with scarce industrial fault data, and constructing a positive and negative sample pair through a plurality of data enhancement strategies; based on heterogeneous multi-modal fault data, designing a double-flow encoder architecture of a time sequence branch and an image branch, extracting depth features and mapping the depth features to a unified feature space through a projection head; performing supervised contrast learning pre-training based on intra-modal and inter-modal dual contrast loss; supervision fine tuning is carried out based on multiple loss functions such as physical guidance, so that accurate diagnosis of equipment faults is realized in a small sample scene. According to the fault diagnosis method under the unbalanced sample and limited labeling conditions, the problem that a traditional data driving model depends on large-scale labeling samples is effectively relieved through supervised comparative learning and cross-modal information alignment.

The invention discloses a power electronic transformer working mode recognition method and device, equipment, a medium and a product, and relates to the field of artificial intelligence, and the method comprises the steps: collecting original inductive current data during the operation of a power electronic transformer; performing adaptive segmentation normalization processing on the original inductive current data to generate a normalized inductive current sequence; calculating a wavelet packet energy entropy and a time domain differential entropy of the normalized inductive current sequence, and splicing the wavelet packet energy entropy and the time domain differential entropy into a two-dimensional fusion feature vector; inputting the normalized inductive current sequence and the two-dimensional fusion feature vector into a trained deep learning model to obtain a prediction probability vector; based on the prediction probability vector, the working mode category with the maximum probability value is selected as the recognition result, and the recognition precision of the working modes of the power electronic transformer can be guaranteed under the working conditions of high noise interference or rapid mode switching.

The invention discloses a pavement crack accurate extraction method and system based on multi-scale image segmentation, and belongs to the technical field of pavement detection, and the method comprises the steps: obtaining a to-be-detected pavement image, carrying out the brightness normalization and geometric correction of the pavement image, and constructing a multi-scale image set with different resolutions; generating a spatial texture channel and a spectral domain response channel for each scale image in the multi-scale image set, and fusing the spatial texture channel and the spectral domain response channel to obtain a spectral-space coupling input tensor, wherein the spectral domain response channel is obtained by extracting energy characteristics of a plurality of frequency bands after performing spectrum transformation on the scale image; a multi-scale image set is constructed for road surface images, and coupling expression of spatial texture information and spectral domain response information is introduced, so that feature consistency of fine cracks under different resolutions is enhanced, and scale drift and response distortion caused by shadows, light reflection and seam textures are inhibited at the same time; therefore, the separability and the stability of the crack in a complex scene are improved.

The application discloses a dynamic detection method for direct current fault arc based on improved stochastic resonance and generative deduction, which comprises the following steps: sampling a direct current loop current, and dynamically strengthening a weakened fault feature by using a parameter self-adaptive adaptive improved stochastic resonance model based on an asymmetric bistable potential function; and then deducing a generalization feature sample set through a generative adversarial network, and realizing detection of the fault arc by using a random forest classification and a multi-time window joint decision. The dynamic detection method for the direct current fault arc based on the improved stochastic resonance and the generative deduction solves the problems of insufficient detection reliability and insufficient sample coverage when the fault feature is weakened in the prior art.

This invention discloses a smart contract vulnerability detection model and device. The model includes a code structure feature extraction module, a thought chain text feature extraction module, a semantic space adaptation module, a multi-head cross-attention module, and a vulnerability classification module. Based on code structure information modeling, this invention further introduces vulnerability reasoning thought chain text features and achieves effective alignment between code structure semantics and vulnerability reasoning semantics through the synergistic effect of semantic space adaptation and multi-head cross-attention. Compared to detection methods that rely solely on overall semantic modeling of a single code modality, this invention no longer depends solely on the overall statistical features of the source code for vulnerability identification. Instead, guided by vulnerability reasoning semantics, it further focuses on core code segments strongly related to the vulnerability triggering logic, reducing the interference of irrelevant business code and redundant code on the feature extraction process, and improving the accuracy and robustness of vulnerability detection.

This invention discloses an infrared small target detection method based on a frequency-space collaborative gating attention network, belonging to the field of infrared small target detection. The method includes: inputting an initial feature map into a pre-trained frequency-space collaborative gating attention network model, the model including an encoder module, a wavelet prior module, and a decoder module; the shallow coding layer of the encoder module extracts low-level spatial features, and the deep coding layer combines frequency-space collaborative gating enhancement with frequency-domain prior information to obtain an encoder-enhanced feature map; the wavelet prior module obtains frequency-domain prior information based on the low-level spatial features and inputs it into the deep coding layer; the shallow decoding layer of the decoder module performs frequency-space collaborative gating enhancement, and the deep decoding layer extracts spatial detail features; the infrared small target detection result is obtained based on the output of the decoder module. This invention can improve the target detection rate and reduce the false alarm rate under conditions of complex backgrounds, strong noise, low signal-to-noise ratio, and insignificant target features, enhancing stability and robustness in practical engineering environments.

The invention relates to the field of wafer code reading, and provides a wafer code reading method and system based on intelligent enhancement and dynamic optimization, and the method comprises the steps: firstly obtaining a wafer character image, and carrying out the image quality evaluation and image preprocessing; then, the preprocessed wafer character image is input into a deep transfer learning character large model for wafer character recognition, a wafer character recognition result and recognition confidence are obtained, and the model structure of the deep transfer learning character large model comprises a double-branch convolutional neural network and a Transform fusion framework; an incremental learning mechanism is introduced for abnormal sample learning; and finally, according to the wafer character recognition result and the recognition confidence coefficient, performing verification through a checksum algorithm, and judging whether a wafer code is directly obtained or a wafer character image corresponding to the wafer character recognition result is marked as an abnormal sample. According to the method, the accuracy, the adaptive capability and the engineering availability of a wafer code reading system in a semiconductor production environment can be remarkably improved.

This invention discloses a radar target detection method based on graph node dual-channel feature attention fusion, belonging to the field of radar signal detection technology. The method includes the following steps: Step 1: Divide the received frame radar echo data into graph nodes; Step 2: Extract time-domain amplitude and time-frequency features from the echo time-series data corresponding to each graph node; Step 3: Construct a feature preprocessing subnetwork; Step 4: Construct a node feature fusion subnetwork; Step 5: Construct a signal classification graph neural network; Step 6: Connect the feature preprocessing subnetwork, the node feature fusion subnetwork, and the signal classification graph neural network in series to form a radar target detection neural network; Step 7: Input the test set into the trained radar target detection neural network and output a binary classification result indicating whether the corresponding node is a target or clutter signal. This method can improve the radar's target detection capability in clutter environments.

The invention relates to the technical field of remote sensing image fine-grained target detection, in particular to a two-way guided remote sensing image fine-grained target detection method. Comprising the steps of 1, acquiring image data, and dividing the image data into a training set and a test set; 2, constructing a target detection network, and obtaining image multi-scale features from the preprocessed image through a feature extraction network; then, a coarse-grained detector screens out a target candidate area and judges whether the target candidate area is a foreground or a background; then, a fine-grained detector carries out finer classification and positioning on the candidate areas; and step 3, carrying out performance evaluation on the test set so as to test the detection capability of the model on the fine-grained target in an actual scene. According to the method, for the problems of class sample imbalance and easy confusion of fine-grained classes in fine-grained target detection, the remote-sensing image fine-grained target detection capability is effectively improved by constructing a layered detection network architecture and combining an instantaneous confidence coefficient signal and class accumulation learning quality.

An aspect of the present disclosure relates to a method for producing a semiconducting single-walled carbon nanotube dispersion that includes a step of centrifuging a separation target SWCNT dispersion containing SWCNTs that include semiconducting SWCNTs and metallic SWCNTs, an aqueous medium, and a polymer, and then collecting a supernatant liquid containing the semiconducting SWCNTs. The polymer is a copolymer that includes a structural unit A derived from a monomer represented by Formula (1) below and a structural unit B derived from a monomer represented by Formula (3) below.         CH2=CR0-COOM     (1)         CH2=CR6-COO-(EO)p(PO)qR6     (3)

This invention discloses a medical image segmentation method based on dynamic deformable convolution and a sliding window adaptive complementary attention mechanism, which improves the perception of small lesions and highly deformed targets in medical images, as well as the ability to distinguish between segmented targets and the background. Dynamic deformable convolution, through task-adaptive learning, can flexibly change weight coefficients and deformation biases, enhancing the ability to express local image features and achieving adaptive extraction of spatial features. The sliding window adaptive complementary attention mechanism achieves cross-dimensional global modeling of medical images through a self-attention branch of adaptively learned weight coefficients. This mechanism overcomes the shortcomings of conventional methods in modeling cross-dimensional relationships between space and channels, and can capture long-distance cross-dimensional correlation features in images. Furthermore, the parallel interactive approach combines local and global features at different resolutions to enhance representation learning, maximizing the preservation of both local and global features in medical images.

The application discloses a kind of spatio-temporal perception remote sensing image change detection methods based on SAM2.The application introduces spatio-temporal perception adapter as the weight sharing twin encoder, difference modeling and attention enhancement are carried out to double time phase features, the consistency of unchanged area features is improved, and the difference of change area features is expanded, so as to effectively alleviate the interference of radiation difference.Simultaneously, a dynamic adaptive scanning change decoder is designed, and the scanning position and scanning path are adaptively adjusted according to the local change characteristics, so as to enhance the modeling capability of irregular change area under complex background.Finally, the proposed method cooperates to improve the double time phase feature representation and change information extraction capability from the encoding end and the decoding end, and effectively improves the accuracy of change detection under the condition of serious radiation difference interference.

The application discloses a pavement crack precise extraction method and system based on multi-scale image segmentation, belongs to the technical field of pavement detection, and comprises the following steps: acquiring a to-be-detected pavement image, performing brightness normalization and geometric correction on the pavement image, and constructing a multi-scale image set with different resolutions; generating a spatial texture channel and a spectral domain response channel for each scale image in the multi-scale image set, and fusing the spatial texture channel and the spectral domain response channel to obtain a spectral space coupling input tensor, wherein the spectral domain response channel is obtained by extracting energy features of multiple frequency bands after performing frequency spectrum transformation on the scale image; the multi-scale image set is constructed through the pavement image, and the coupling expression of the spatial texture information and the spectral domain response information is introduced, so that the feature consistency of fine cracks under different resolutions is strengthened, and the scale drift and response distortion caused by shadows, reflections and joint textures are inhibited, so that the crack separability and stability under a complex scene are improved.

The invention discloses an algae community detection method and system based on polarized light characteristics, and relates to the field of water environment monitoring, and the method comprises the steps: obtaining an algae single particle polarization scattering signal through multiple observation angles, resolving a Stokes parameter, further extracting a polarization phase to construct a polarization phase divergence matrix, and carrying out the structural representation of the polarization difference between the angles; meanwhile, a morphological complexity index is obtained through multi-angle scattering intensity statistics and is used for depicting scattering distribution characteristics; on the basis, a polarization-form coupling coefficient is provided, polarization phase differences of different angle combinations are subjected to weighted convergence and modulated by form complexity, and stable fusion of polarization information and form statistical information is achieved; and the coupling coefficients are subjected to statistical convergence in a time window to form an algal group cooperation index, so that unified output of single-particle species classification and community risk early warning is realized. Therefore, the distinction degree, the stability and the time sequence characterization capability of algae community detection are improved.

A head and neck cancer local recurrence information acquisition method based on multi-modal supervised contrast learning, specifically comprising: sample set division and image preprocessing, multi-view data enhancement, constructing an image coding network based on multi-modal supervised contrast learning, training the constructed coding network, view pooling fusion image coding, encoding clinical data according to an attribute-based hybrid coding strategy, constructing a classifier network based on multi-modal decoupling bilinear pooling fusion, and training the constructed classifier network; the original contrast learning method is increased based on the multi-modal structure, the intra-modal and inter-modal contrast losses are optimized together, label information is added, the decoupling bilinear pooling mode is used to fuse the extracted image features, and the accuracy of the head and neck cancer recurrence signal acquisition result is improved.

The present application relates to the technical field of fabric detection, and particularly relates to a fabric flaw intelligent detection method and system based on machine vision, which comprises the following steps: performing gray-scale processing, geometric correction and denoising, illumination compensation and texture enhancement processing on the fabric image collected under controlled illumination, extracting edges and tracking contour point sequences; calculating outer angles, directionality and signed curvatures based on local point sets, combining concave-convex discrimination, key point reservation, thinning and anti-aliasing smoothing, and reconstructing continuous defect boundary trajectories; and then matching with a defect feature interval library to output defect types and boundary positioning information. Through the present application, the problems of missing detection / mis-detection and unstable positioning of defect boundaries under complex texture and illumination fluctuation conditions in the prior art are solved.

The present application relates to the technical field of fluid machinery fault diagnosis, and discloses an emulsion pump fault intelligent diagnosis method and system based on multi-dimensional entropy feature fusion, which comprises the following steps: constructing a multi-dimensional physical field state monitoring space, synchronously collecting vibration acceleration, outlet pressure and flow time series signals; reconstructing the vibration signal through adaptive variational modal decomposition and effective sensitive component screening, calculating normalized vibration sub-number arrangement entropy through phase space reconstruction and sub-number mapping, extracting pressure fluctuation entropy and flow pulsation variance and fusing them into a multi-dimensional fault feature vector, inputting the multi-classification support vector machine to output an operation state label, and executing hierarchical closed-loop control; through sub-number mapping and multi-physical field fusion, the present application effectively overcomes the problems of entropy value feature distortion caused by non-Gaussian impact noise and weak fault feature masking caused by liquid-machine coupling, and improves the robustness and accuracy of emulsion pump fault diagnosis under complex working conditions.

This invention belongs to the field of wireless communication and relates to a machine learning-based method for wireless communication channel estimation. The method includes: acquiring pilot signals, antenna array geometric parameters, and multipath propagation parameters from the transceiver end of a millimeter-wave massive MIMO system to construct a three-dimensional propagation structure model; determining candidate regions for the channel sparse support set and generating a prior constraint matrix for the channel sparse support; inputting the prior constraint matrix and pilot observation data into a sparse reconstruction algorithm for channel sparse feature extraction and dimensionality reduction; performing super-oscillatory feature enhancement on the dimensionality-reduced channel sparse features to form an enhanced channel feature vector; inputting the enhanced channel feature vector and the prior support vector matrix into a complex domain deep expansion network to output the channel estimation result; and outputting the final beamforming weights. Its beneficial effects are improved sparse support set matching accuracy and channel estimation precision with low pilot overhead, and enhanced system throughput and interference suppression capabilities.

The application discloses a rumor detection method and device fusing expression emotion perception and semantic evolution, electronic equipment and a storage medium. The method comprises the following steps: firstly, a propagation tree is constructed in units of social events; then, multi-source features are extracted, semantic features are extracted based on a Roberta model processing text content, text semantic features of the propagation tree are combined with user statistical features, and expression emotion features of the propagation tree level are constructed based on an expression emotion symbol information base; then, semantic evolution features of the combined representation are captured based on a BiLSTM network model, and a self-attention mechanism is introduced to highlight key propagation semantic information; finally, the semantic evolution features and the expression emotion features are fused, so that enhanced emotion-semantic features are formed and used for rumor discrimination of the social events.

The invention provides an adaptive SSMVEP-MI fusion brain-computer interface method and system based on eye movement tracking. The method comprises the following steps: determining a current effective fixation area of a user in real time through eye movement tracking; based on the region, only activating a corresponding local SSMVEP visual stimulation target, and generating associated specific motor imagery prompt information; electroencephalogram signals are synchronously collected, and SSMVEP features and MI features are extracted; self-adaptively selecting a decoding path according to whether the accumulated number of the MI effective samples with the labels reaches a threshold value or not: if not, independently decoding by using the SSMVEP features; if yes, fusion decoding based on signal quality evaluation is carried out on the bimodal features; and finally outputting a control signal to drive external equipment or update an interactive interface. According to the method, the visual interference is remarkably reduced, the recognition accuracy and the training efficiency at the initial stage of motor imagery are improved, and the robustness and the availability of the system are ensured through a dual adaptive mechanism of the sample number and the signal quality.

The application discloses an AI image discrimination method based on frequency domain and noise domain consistency, and belongs to the technical field of image processing, computer vision and artificial intelligence security. The method comprises the following steps: acquiring an image to be discriminated and performing pretreatment, converting the image into a brightness channel and acquiring frequency spectrum information; inputting the image into a dual-domain physical guidance feature extraction module to extract directional spectrum features and multi-scale noise flow features; inputting the directional spectrum features and the multi-scale noise flow features into a cross-domain physical consistency module to generate a local consistency distance map and a global consistency score through shared embedding mapping and feature distance calculation, and obtaining consistency perception features; inputting the directional spectrum features, the multi-scale noise flow features and the consistency perception features into a dual-flow fusion module for cross-flow interactive fusion and global feature aggregation to obtain discrimination feature representation; and inputting the discrimination feature representation into a classification head to output a discrimination result of whether the image to be discriminated is a real image or an AI generated image. The application realizes effective discrimination of AI generated images by jointly modeling the physical consistency relationship between frequency domain features and noise domain features, and improves the robustness and generalization ability of the model under the conditions of cross-generator, cross-dataset and image compression, blurring and other post-processing conditions while ensuring detection accuracy.

The invention provides a bearing fault diagnosis method based on Fisher criterion adaptive optimization RCMFDE and CAOA-SVM, and the method comprises the steps: firstly obtaining a vibration acceleration signal of a rolling bearing, and discretizing a continuous time sequence into sample segments based on a non-overlapping sliding window technology; secondly, constructing a fractional order divergence entropy of a nonlinear gain characteristic, obtaining a fine composite multi-scale fractional order divergence entropy based on a fine composite multi-scale strategy, taking the fine composite multi-scale fractional order divergence entropy as a feature quality evaluation function based on a kernel Fisher discrimination criterion, adaptively searching an optimal fractional order in a preset interval, and extracting a fault feature vector with high discrimination; and inputting the extracted feature vector into a fault diagnosis model, and outputting the fault category of the rolling bearing. According to the method, the problems of blind parameter selection and difficult weak feature extraction in bearing fault diagnosis are solved, and the diagnosis efficiency and generalization ability are improved.

This invention belongs to the field of satellite remote sensing and red tide detection technology, and discloses an integrated method and system for generating and identifying broadband 3D samples of dominant red tide species. The method constructs an integrated model for generating broadband 3D red tide samples and identifying dominant species based on deep learning. It utilizes a red tide image generation module based on spectral feature transfer to obtain expanded samples of different dominant red tide species. A red tide dominant species identification module is constructed by integrating multi-scale feature extraction and cross-spatial attention mechanisms. The module is trained using a student-teacher semi-supervised learning model. Experimental results show that the model has good adaptability and can effectively identify three dominant red tide species: *Noctiluca scintillans*, *Noctiluca chloroticus*, and *Hacochloa erythrophora*, with an overall identification accuracy of 94.48%, which is 3.52%–8.66% higher than other comparative methods, providing a basis for red tide prevention and management.

The invention discloses a small sample hyperspectral change detection method based on surface feature category prompt, and relates to the technical field of remote sensing image processing, and the method comprises the following steps: S1, constructing surface feature category prompt, and carrying out the preprocessing of a hyperspectral image; s2, extracting semantic modal features corresponding to surface feature category prompts and visual modal features corresponding to the preprocessed hyperspectral image; s3, fusing the semantic modal features and the visual modal features to obtain transformed deep visual features; s4, obtaining a multi-scale change feature subjected to multi-modal feature attention enhancement; and S5, decoding the multi-scale change features subjected to multi-modal feature attention enhancement by using a decoder of the hyperspectral change detection network to obtain a binary change graph. According to the method, the network can still obtain reliable change discrimination capability when only a small number of labeled samples are used, so that the dependence on large-scale hyperspectral manual labeled data or additional source domain data is reduced.

The application discloses a motor fault early warning method and device based on artificial intelligence and a storage medium, and relates to the technical field of motor fault detection. The working condition data and vibration signals of the motor during operation are acquired, the vibration signals are denoised to obtain a target vector, and frequency domain transformation and time-frequency analysis are performed to obtain a first frequency domain feature and a first time-frequency feature respectively. The working condition data, the frequency domain feature and the time-frequency feature are input into a pre-trained fault diagnosis model, the model adopts a double-branch structure to extract the frequency domain and time-frequency features respectively, a second frequency domain feature and a second time-frequency feature are obtained, and the features are enhanced in combination with the working condition data, fault diagnosis is performed based on the enhanced features, a diagnosis result is output, and early warning information is generated. The method fuses working condition information and multi-domain vibration features, effectively distinguishes feature deviation caused by working condition changes, improves early weak fault recognition capability, and improves diagnosis accuracy and early warning timeliness.

This invention discloses a method and system for snow and ice identification based on remote sensing images. The method includes: acquiring remote sensing images of the target area and performing sensor-specific preprocessing and topographic radiometric correction to obtain a standardized surface reflectance image; extracting improved spectral index features, scale-adaptive texture features, and topographic occlusion compensation features based on multi-level feature coupling rules to construct a multi-dimensional feature set; generating a snow and ice probability map through a feature pyramid network based on an attention mechanism; generating an initial snow and ice mask using a dynamic window adaptive threshold segmentation strategy, combined with elevation zonation constraints and multi-temporal change trajectories; and finally, removing transient coverage and noise through spatiotemporal consistency joint optimization to generate an accurate snow and ice coverage map. This invention effectively solves the problem of spectral confusion between snow and ice, clouds, and bare rock, overcomes the influence of topographic shadows and transient interference, and significantly improves the accuracy and reliability of snow and ice identification in complex environments.

A multi-class motor imagery brain-computer interface decoding method combining EEGNet and FBCSP includes: 1) segmenting the continuous EEG signals of the subject into single-test signals and using filter banks for sub-band filtering; 2) using the training data of all subjects to perform transfer learning training on EEGNet, and classifying the test data to obtain classification scores for all tasks; 3) based on the OVO strategy, training the test subjects' training data on FBCSP to obtain spatial filters for each sub-band and each pair of tasks, concatenating the training and test feature signals of all sub-bands for each pair of tasks after spatial filtering, using the former to train an SVM to classify the latter to obtain classification scores for each pair of tasks, and summing all scores for each task; 4) standardizing the score vectors obtained from EEGNet and FBCSP and summing them, with the maximum value corresponding to the task as the test signal label. This invention can improve the classification accuracy of BCI and promote its practical application.

The application discloses a discharge detection device and method for a general high-low voltage switch cabinet for mines, and belongs to the technical field of switch cabinet detection. The device comprises a modularized sensor assembly, a signal acquisition unit, an edge computing node unit and an upper monitoring platform, and adopts a UHF sensor, an ultrasonic sensor and a transient ground voltage sensor for cooperative detection. The method comprises multi-source signal synchronous acquisition, adaptive denoising based on a sparrow search algorithm optimization, double-domain feature extraction, discharge type identification based on an attention-enhanced convolutional neural network, and multi-sensor information fusion and hierarchical early warning based on evidence theory. The application compensates for the blind area of a single detection method through cooperative work of multiple types of sensors, improves the denoising effect through adaptive parameter optimization, enhances the recognition accuracy through double-domain features and an attention mechanism, and improves the diagnosis reliability through multi-sensor fusion, so that accurate detection and intelligent diagnosis of local discharge of a mine switch cabinet are realized.

This invention discloses a method and apparatus for detecting series fault arcs based on a residual dilated convolutional neural network, comprising: acquiring fault arc waveforms and normal arc waveforms under various loads; extracting three physically interpretable feature quantities—time-domain amplitude, spectral centroid, and instantaneous frequency—and calculating the Gram angle field matrix for each of the three feature quantities; enhancing the generated Gram matrix based on the physical characteristics of the AC arc by calculating the zero-crossing re-ignition coefficient to generate an adaptive pseudo-color image; constructing a residual dilated convolutional neural network model, wherein the model contains three sets of residual sequence modules with progressively increasing expansion rates; training the model using pseudo-color image data; and outputting the detection result by inputting the test sample. This invention achieves efficient and high-precision detection of series fault arcs under conditions of limited load types and limited samples through the synergistic design of physically interpretable multi-domain feature fusion and residual dilated network architecture.

This invention discloses a super-resolution reconstruction method for the carbon skeleton of a fuel cell catalyst layer using FIB-SEM. The invention employs a high-precision 3D digital structure generation method to construct high-resolution supervision labels required for super-resolution training. Based on the same digital structure, a corresponding low-resolution SEM image is generated through Monte Carlo electron scattering simulation, thus establishing a one-to-one correspondence training sample set between "low-resolution imaging representation and high-resolution real structure". Since the high-resolution supervision labels are directly derived from the digital structure generation process, smaller unit scales can be used to express the nanopores and skeleton details of the catalyst layer during the structure construction stage. The low-resolution input is obtained from physically consistent SEM imaging simulation, reflecting the resolution loss and signal degradation in the real imaging process. This bypasses the difficulty in obtaining real high- and low-resolution paired samples and the training data limitations imposed by the approximately 3 nm lateral resolution of real FIB-SEM, effectively improving the recognizability and reconstruction capability of the catalyst layer microstructure features.