67results about How to "Improve reconstruction accuracy" patented technology

The invention provides a hyperspectral image reconstruction method based on space-spectral characteristic fusion. The method comprises the following implementation steps: acquiring a training sample set and a test sample set; constructing a reconstruction network model based on space-spectral characteristic fusion; performing iterative training on the reconstructed network model; and obtaining a reconstruction result of the hyperspectral image. According to the invention, the spatial-spectral characteristic guiding network in the encoder carries out spatial-spectral characteristic extraction on input characteristics, the spatial-spectral characteristic guiding network in the bottleneck layer carries out spatial-spectral characteristic fusion, the spatial-spectral characteristic guiding network in the decoder carries out spatial-spectral characteristic detail recovery, spatial structure information and spectral distribution information in the input characteristics are fully extracted, and the spatial-spectral characteristic guiding network in the decoder carries out spatial-spectral characteristic fusion. The spatial features and the spectral features cooperatively participate in the reconstruction process, and cooperative recovery of the spatial information and the spectral information of the hyperspectral image is realized, so that the reconstruction precision of the hyperspectral image is effectively improved.

The invention provides a diffusion-driven channel adaptive point cloud semantic communication method, which belongs to the technical field of wireless communication, and comprises the following steps: obtaining a training point cloud data set, constructing a point cloud feature extraction network based on a hypergraph convolutional neural network as a semantic encoder, and constructing a channel adaptive enhancement module as a channel decoder, a channel adaptive recovery module is constructed at a receiving end as a channel decoder, a diffusion reconstruction network is constructed as a semantic decoder, end-to-end joint training is performed on the point cloud feature extraction network, the channel adaptive enhancement module, the channel adaptive recovery module and the diffusion reconstruction network, the diffusion reconstruction network predicts original point cloud distribution, and the point cloud feature extraction network and the channel adaptive enhancement module are subjected to end-to-end joint training. And a loss function is constructed based on the chamfering distance between the predicted point cloud distribution and the original point cloud. In the communication process, the reconstructed semantic features are input into the diffusion reconstruction network to reconstruct an original point cloud structure. The high-order semantic features of the point cloud can be effectively extracted so as to improve the adaptive capacity of the method to the incomplete point cloud.

A space-based microwave and laser fused ocean subsurface layer temperature field reconstruction method solves the technical problems that current ocean subsurface layer temperature profile acquisition lacks direct participation of subsurface layer parameters and the reconstruction precision is low, and the ocean subsurface layer temperature field reconstruction method comprises the following steps: step 1, determining an error of laser radar inversion ocean subsurface layer temperature; 2, generating an input data set and an output data set; 3, constructing a sub-surface layer temperature profile reconstruction model, and performing training; and step 4, performing efficiency evaluation on the trained model. According to the method, the temperature field reconstruction precision is remarkably improved, and the traditional technical limitation depending on a pure sea surface parameter model is broken through.

The invention discloses an orthopedic surgery navigation positioning method and system based on three-dimensional registration, and belongs to the technical field of orthopedic surgery navigation. According to the method, through preoperative multi-modal image fusion reconstruction, intraoperative dynamic point cloud collection and an improved mixed registration algorithm, precise alignment of a preoperative model and an intraoperative actual skeleton is achieved, and by combining real-time error monitoring and dynamic compensation, precise positioning parameters of a surgical instrument are output to guide surgical operation; the system corresponds to the method and comprises an image acquisition module, a three-dimensional reconstruction module, a registration calculation module, a navigation guiding module, an error monitoring module and a control module, all the modules work cooperatively, and it is ensured that the navigation positioning precision is smaller than or equal to 0.3 mm, the registration time is smaller than or equal to 3.5 s, and the dynamic error compensation response time in the operation is smaller than or equal to 50 ms. The method does not need to depend on a physical marker, does not need a doctor to manually intervene the registration process, is high in anti-interference capability, is suitable for various orthopedic surgery scenes, effectively reduces the surgical risk, and improves the surgical efficiency and the standardization level.

The present application relates to the technical field of computer graphics, computer vision and three-dimensional reconstruction, and particularly relates to an immersive scene reconstruction rendering method and system based on adaptive 3D Gaussian sputtering, which acquires scene initial point cloud data and camera parameters, constructs a 3D Gaussian point cloud model containing multiple attributes and generates sky point cloud; based on the mechanism of pulse neurons, the opacity of Gaussian points is adaptively controlled, and according to the gradient information and geometric attributes of Gaussian points, splitting, copying or pruning operations are performed to optimize the distribution of point cloud; through a deep learning network, the camera pose is corrected, large-scale scenes are divided into blocks for parallel training, and a bilateral grid post-processing is combined with multiple loss functions to optimize rendering, and the sky point cloud is subjected to exclusive constraint, so that the technical defects of the prior art are solved, and the immersive scene application requirements in the fields of VR / AR, digital twin and the like are met.

The application discloses a ramjet lightweight flow field reconstruction method guided by multi-scale feature mapping, and belongs to the technical field of engine combustion monitoring and artificial intelligence application, and the method comprises the following steps: S1, acquiring wall surface pressure time sequence signals and schlieren images, and constructing a combustion flow field reconstruction data set under different inflow conditions; S2, performing data preprocessing on the combustion flow field reconstruction data set to obtain a preprocessed combustion flow field reconstruction data set; S3, inputting the preprocessed combustion flow field reconstruction data set into a space-time multi-scale feature alignment and fusion model to obtain a lightweight flow field reconstruction model; and S4, real-time receiving wall surface pressure time sequence signals and inputting the wall surface pressure time sequence signals into the lightweight flow field reconstruction model to reconstruct a high-time-resolution two-dimensional transient flow field. The method realizes high-precision and high-efficiency reconstruction of extreme combustion non-steady-state multi-physical fields under high-speed flight conditions, and solves the problems of data sparseness and inability to perceive combustion characteristics at future time points in advance in the traditional method.

The invention discloses a method for constructing, updating and retrieving an action memory library, which belongs to the technical field of computer vision and comprises the following steps of: constructing a training data source with time sequence diversity; initializing an action memory library containing a plurality of learnable prototype matrixes and a double-flow interaction network; memory bank evolution is executed, an action prototype is retrieved by utilizing a query stream, a current memory state is dynamically generated by combining a memory state updating gate mechanism with a historical state, and dynamic memory is injected into a feature space by utilizing memory driving graph convolution; synchronously updating parameters based on multi-target loss, and driving a memory bank to evolve into optimal structured prior; and finally, performing structured reasoning on the to-be-detected sequence by using the optimal memory bank. According to the method, structured priori is constructed by mining a spatio-temporal topology mode of a human body action sequence, and hierarchical memory evolution and double-flow depth interaction are combined, so that the problem of depth ambiguity in a monocular vision task is effectively solved, geometric structure distortion is corrected, and the accuracy of action posture estimation is remarkably improved.

The application relates to the technical field of image processing, in particular to an image super-resolution reconstruction method based on a multi-scale content perception mixer, which is realized by using an adaptive processing mechanism. The method comprises the following steps: shallow feature extraction is performed on a low-resolution image to be reconstructed, so as to obtain an initial shallow feature map; feature enhancement based on a feature pyramid and an attention mechanism is performed on the shallow feature map, so as to obtain a deep feature map; multi-scale content perception prediction is performed based on the deep feature map, so as to generate guide information for guiding calculation allocation, the guide information comprising a window classification binary mask and a window size; different image regions are allocated to different calculation paths for processing based on the guide information; the feature maps output by the calculation paths are recombined and fused, and then enlarged to a target resolution, so that a high-resolution image is finally obtained. The method realizes accurate classification of image regions and on-demand allocation of calculation resources, and significantly reduces the calculation complexity and the memory occupation.

The application relates to a multispectral image reconstruction system, and belongs to the field of image processing. The multispectral image reconstruction system comprises an intensity difference mask convolution demosaicking module, a multi-level convolution feature extraction module, a multi-dimensional attention feature fusion module and a CNN residual reconstruction module; an intensity difference mosaic image is obtained by subtracting a multispectral mosaic image from a single-channel intensity image; the same convolution weight is assigned to pixels belonging to the same MSFA position by adopting a position mask convolution method, different convolution weights are assigned to pixels belonging to different MSFA positions, and a convolution feature map is obtained; multi-level convolution fusion features are obtained based on the convolution feature map; high-frequency convolution features and cross-attention fusion features are extracted based on the multi-level convolution fusion features, and multi-dimensional adaptive fusion features are obtained; residual features are extracted from the multi-dimensional adaptive fusion features, and a multispectral image is obtained. The application can comprehensively recover the missing information of the multispectral mosaic image and improve the reconstruction accuracy.

The present application relates to the technical field of vehicle trajectory construction, and discloses a vehicle trajectory reconstruction method and device based on an interpretable attention, which comprises the following steps: obtaining trajectory data of a first vehicle at multiple time steps; determining a relative position relationship between a second vehicle and the first vehicle, wherein the relative position relationship comprises the second vehicle being a preceding vehicle or a following vehicle of the first vehicle; determining a target model from a first pre-trained deep learning model and a second pre-trained deep learning model according to the relative position relationship, wherein the first deep learning model is used for reconstructing a preceding vehicle trajectory, and the second deep learning model is used for reconstructing a following vehicle trajectory; and inputting the trajectory data into the target model to obtain a reconstructed trajectory of the second vehicle. The present application realizes the reconstruction of the preceding and following vehicles by using targeted models respectively, thereby improving the accuracy of the reconstruction.

The application discloses a three-dimensional reconstruction method and system based on frequency domain phase consistency and multi-cue fusion, and relates to the technical field of three-dimensional reconstruction. The method comprises the following steps: collecting images of a target semiconductor packaging structure, forming a focus stack, setting analysis parameters, extracting four-dimensional depth cues from three dimensions of a frequency domain, a time domain and a space domain in parallel, obtaining a four-dimensional cue matrix, tracking time sequence peaks of each cue in the four-dimensional cue matrix pixel by pixel, performing sub-pixel fitting, fusing the fitted four-dimensional cues through Bayesian maximum a posteriori probability estimation, generating a coarse depth map and a confidence map, performing Markov random field space regularization optimization to generate an optimized depth map, adaptively smoothing the optimized depth map based on the fused confidence map, and generating a three-dimensional depth map and a full-focus image. The application can provide an efficient, stable and engineerable new technical path for three-dimensional online detection of wafer-level advanced packaging.

This invention relates to a power data sparsity compression observation method for non-intrusive load monitoring, comprising the following steps: Step 1, acquiring power monitoring data and determining its electricity consumption behavior pattern; Step 2, initializing wavelet sparse basis for sample data based on simple power behavior patterns; Step 3, based on the determination results of the electricity consumption behavior pattern of the completed training samples and the results after wavelet basis initialization in Steps 1 and 2, performing improved K-SVD training and generating sparse basis; Step 4, training the sample dataset Y based on Step 3 to obtain a sparse dictionary D, and performing power monitoring data compression observation based on the improved sparse basis.

This invention proposes a hyperspectral image fusion and reconstruction method based on cross-modal interaction and difference perception. The steps are as follows: acquiring training and testing sample sets; constructing a reconstruction network model based on cross-modal interaction and difference perception; iteratively training the reconstruction network model; and obtaining the reconstruction results of the hyperspectral image. The residual module of this invention utilizes multiple cross-modal interaction modules (CIMB) to perform cross-modal collaborative interaction on the spectral-spatial feature maps of LR-HSI and HR-MSI, enabling the mining of interaction features and contextual features between the two modalities. The difference perception fusion module (DPFM) performs difference perception and fusion on the cross-modal interaction features output by CIMB, effectively enhancing key details while mitigating distortion caused by differences between the two modal features, thus significantly improving reconstruction accuracy.

This application discloses a 3D reconstruction method, storage medium, and computer program product, relating to the field of computer vision technology. The method includes: acquiring a sequence of unlabeled multi-frame video images; calling a feedforward 3D reconstruction model; obtaining ordinal depth loss, local affine invariant loss, and multi-view consistency constraint loss; the ordinal depth loss constrains the relative depth ordering of pixel pairs, the local affine invariant loss preserves locally aligned surface details, and the multi-view consistency constraint loss is used to determine the camera position and pose; jointly updating the model parameters using the ordinal depth loss, local affine invariant loss, and multi-view consistency constraint loss to obtain a 3D reconstruction optimization model; and inputting the multi-frame video image sequence into the 3D reconstruction optimization model for forward propagation to obtain 3D reconstruction information. By jointly updating the model parameters using three different loss functions, the reconstruction accuracy of dynamic scenes is improved without relying on expensive annotations and while preserving inference efficiency.

This invention provides a method and apparatus for reconstructing the three-dimensional temperature and humidity field of a tropical cyclone based on deep learning and sparse observations. The method includes: acquiring radiosonde observation data and satellite remote sensing data for a target tropical cyclone region; mapping the radiosonde observation data onto a two-dimensional grid on multiple preset pressure layers to generate an initial temperature and humidity field containing missing data regions, and generating a mask to identify the missing data locations; interpolating and aligning the satellite remote sensing data onto the two-dimensional grid to form an auxiliary remote sensing field; combining the initial temperature and humidity field, the mask, and the auxiliary remote sensing field at each pressure layer to form a multi-channel input tensor, which is then input into a pre-trained deep neural network model. The model infers and completes the missing data regions, outputting the complete two-dimensional temperature and humidity field for the current layer; and vertically synthesizing the complete two-dimensional temperature and humidity fields corresponding to all pressure layers to obtain the three-dimensional temperature and humidity field of the target tropical cyclone.

The invention discloses a deepwater structure health monitoring method and system. The method comprises the following steps: establishing a discrete model and constructing a unit strain matrix; collecting single-side surface strain data and establishing a mapping relation between single-side surface theoretical strain and node displacement; constructing a unilateral inverse finite element data fitting item comprising a surface strain fitting item and a lateral shear strain penalty item; constructing a geometric guidance hybrid regularization term consisting of a membrane regularization term and a bending regularization term; assembling a global equation to solve full-field displacement; and recovering full-field strain and stress and outputting a monitoring result. The system comprises a data acquisition module, a preprocessing and inverse reconstruction module, a stabilization solving module and a visual output module. According to the full-field response sensing scheme based on unilateral inverse finite element and geometric guidance mixed regularization, stable and high-precision inversion of full-field displacement, strain and stress of the structure can be realized under the condition of only depending on unilateral sparse sensing data, and the dependence of a traditional method on bilateral sensing arrangement is effectively broken through.

A model training method, a scene reconstruction method, a device, equipment and a medium are disclosed. A training data set of a target scene is obtained, each sample including at least multi-view images, a target camera pose, depth information, semantic constraint data of a visual marker and a target pose; the multi-view images and the target camera pose of each training sample are input into a neural radiance field model to obtain scene data including a predicted scene color image and a predicted scene depth image; based on the predicted scene depth image, the target camera pose and the target pose, predicted semantic constraint data is determined; based on each training sample, corresponding scene data and predicted semantic constraint data, a preset loss function is used to train the neural radiance field model, the preset loss function including a photometric reconstruction loss term, a depth consistency loss term and a semantic constraint loss term, three types of loss terms are used to cooperatively optimize the model, and the reconstruction accuracy, global consistency and rendering quality of the model are effectively improved.

The invention discloses a historical building multi-source data fusion and intelligent modeling method and system, and relates to the technical field of buildings, and the method comprises the steps: S1, carrying out the initial registration of obtained historical building multi-source data, and obtaining a registration parameter set; s2, on the basis of the registration parameter set, calculating a structure compliance deviation of each data segment relative to a building construction rule, and generating a credibility lambda vector on the basis of the structure compliance deviation; s3, executing geodesic optimization search in the registration parameter space according to the credibility lambda vector, dynamically adjusting the data fragment sampling weight according to the geodesic curvature of the optimized trajectory, and outputting a re-estimated registration result and an optimization difficulty sequence; s4, determining the initial space configuration of the hypergraph nodes based on the re-estimation registration result, and constructing a dynamic symbol hypergraph network on the basis of the hypergraph nodes; and based on the dynamic symbol hypergraph network and the credibility lambda vector, generating a condition-associated creation rule analysis tree to output a historical building three-dimensional model.

This invention provides a method and system for geomagnetic data reconstruction based on reference channel constraints and a physical information neural network. The method includes: S1: segmenting the geomagnetic data to be processed and inputting it into a classification model to identify it as high-quality geomagnetic data segments or noisy / anomaly-containing geomagnetic data segments; S2: inputting the high-quality geomagnetic data segments into an MVMD model based on reference channel constraints to obtain a series of modal components; S3: inputting each modal component into an MFFE-LSTM-PI network model to obtain a predicted value, and then superimposing the predicted values ​​corresponding to each modal component to obtain the geomagnetic data reconstructed for the next time period of the high-quality geomagnetic data segment. The technical solution of this invention combines MVMD and MFFE-LSTM-PI networks to achieve multi-band decomposition and reconstruction, and further optimizes the classification model by using an Improve-TCN network for preliminary processing of geomagnetic data to achieve geomagnetic signal-to-noise identification.

The invention discloses an unmanned aerial vehicle surface reconstruction method based on point cloud semantic communication, and belongs to the technical field of point cloud semantic communication. According to the method, the problem that in the surface reconstruction process of the unmanned aerial vehicle, due to the defects of limited flight time and computing resources of the unmanned aerial vehicle, information needs to be completely transmitted in a transmission bandwidth limited scene, so that surface reconstruction is smoothly carried out is solved. The invention provides a new semantic communication algorithm based on the point cloud, an auto-encoder architecture with category importance permission is introduced for selective transmission, and unnecessary data transmission is reduced, so that the bandwidth is saved, and information more related to a surface reconstruction task is transmitted. Secondly, an end-to-end training system framework is developed, joint optimization of deep learning is utilized, data transmission and surface reconstruction processes are simplified, and the overall efficiency and reconstruction precision of the system are improved. Wide verification and simulation results prove that the framework can complete surface reconstruction with less point cloud data transmission quantity, the superiority of the framework in efficiency and performance is verified, and the application of the system in different data types and scenes is expanded.

The application discloses a reflection type laminated imaging method based on off-axis diffraction and a related device, relates to the field of laminated diffraction imaging, and comprises the following steps: acquiring diffraction images of different scanning positions in a target region of a sample; according to the diffraction images, applying an improved rPIE iterative algorithm to perform diffraction light field reconstruction to obtain the amplitude and phase of the sample reconstruction; and the improved rPIE iterative algorithm refers to replacing a diffraction propagation model in a traditional rPIE iterative algorithm with a parallel plane off-axis diffraction transmission model. In the application, the diffraction propagation model in the traditional rPIE iterative algorithm is replaced with the parallel plane off-axis diffraction transmission model, a relatively simple light field transmission model is established and applied to realize the reconstruction of the reflection type laminated imaging, no complex interpolation calculation is involved, the reconstruction method is simplified, and the reconstruction accuracy is improved from the aspects of the definition, contrast and resolution of the reconstruction result.

The application discloses a kind of space neighborhood reconstruction self-encoding multimodal fusion vector generation method, belong to geographic information science, computer vision and artificial intelligence cross technical field.The present mask self-encoding technique is only stopped at the technical limitation of grid level spatial unit, cannot process the heterogeneous modal vector of geographical entity with independent identity, the application proposes a kind of geographical entity-oriented multimodal mask self-encoding training framework: the multiple heterogeneous modal vectors of geographical entity with independent entity identification are randomly masked, the heterogeneous modal vector information of adjacent entity in the space neighborhood determined by k-ring algorithm is used, the masked modal is reconstructed by multimodal fusion network using the remaining modal vector of the entity not masked;With the weighted combination of reconstruction loss and contrast learning loss, the fusion network parameters are jointly optimized.The application also provides a joint optimization method and a cold start and gradual switching method.

The application belongs to the technical field of structural health management, and particularly relates to a multi-sensor failure compensation method based on high-dimensional feature reconstruction, comprising: initializing algorithm parameters, acquiring monitoring data by using multi-sensors, and converting the monitoring data into a matrix; standardizing observation values of the matrix to obtain a standardized data matrix, and obtaining a covariance matrix in a subspace after dimension reduction; based on an ADMM method, decomposing the covariance matrix in the subspace after dimension reduction to obtain a low-rank matrix and a sparse matrix; reconstructing missing sensor data based on the low-rank matrix and the sparse matrix to obtain a reconstructed sensor vector. The aviation aircraft sensor failure compensation algorithm based on high-dimensional feature reconstruction fully considers the coupling relationship between sensors and real-time requirements, significantly improves the prediction accuracy under the condition of missing fault sensors, and is suitable for wide application of various aircrafts such as civil aviation passenger planes and unmanned aerial vehicles.

This disclosure provides a model training method, comprising: extracting features of an image, the features including channel dimension, height dimension, and width dimension; dividing the features into multiple feature blocks along the channel dimension; calculating entropy constraints of the features based on the encoding entropy of each feature block; and training an encoder based on the entropy constraints of the features, the encoder being used to generate compressed codes for the image. This disclosure also provides a method for training a generator using the encoder trained by this model training method, and for generating compressed codes for input image labels based on the generator.

The invention discloses a spatial-temporal feature enhancement-based dynamic point cloud geometric compression method and related equipment, and the method comprises the steps: carrying out the multi-layer down-sampling of a reference point cloud and an original point cloud, and obtaining L layers of reference compression point clouds and original compression point clouds; entropy coding and decoding are carried out on the L layers of original compressed point clouds to obtain L layers of reconstructed point clouds; l = L; performing up-sampling on the lth-layer reconstruction point cloud to obtain (l-1)-layer extension point cloud, performing geometric coding and decoding on the (l-1)-layer extension point cloud and the (l-1)-layer original compression point cloud to obtain (l-1)-layer coarse reconstruction point cloud; performing time domain modeling based on the (l-1)-layer coarse reconstruction point cloud and the reference compression point cloud to obtain (l-1)-layer fusion alignment features; obtaining (l-1) layers of reconstruction point clouds based on the (l-1) layers of fusion alignment features and the original compression point clouds; performing progressive decrease operation on l, and performing cyclic reconstruction until l is equal to 1; and performing up-sampling according to the first-layer reconstructed point cloud to obtain a final reconstructed point cloud. The method can adapt to complex point cloud compression requirements and can be widely applied to the technical field of point cloud data processing.

An MIMO array sparse rapid design method of a fuzzy function generator based on deep learning comprises the following steps: firstly, directly learning a fuzzy function mapping relation corresponding to virtual array configuration and an incident angle by using a deep learning model so as to replace a traditional analysis fuzzy function calculation process depending on a manifold matrix; and then performing efficient performance evaluation on the candidate sparse array based on the fuzzy function generator, and obtaining a sparse array structure meeting the requirements of high resolution and low sidelobe in combination with an optimization strategy. By means of the technology, acceleration of more than 35 times can be achieved on the premise of not explicitly calculating the fuzzy function, the fuzzy function prediction root mean square error (RMSE) lower than 0.1 is kept, and therefore the sparse array design time is greatly shortened. Experiments on a 2T4R K-band MIMO radar prove that the designed array achieves 10 dB sidelobe level in a 100-degree field of view, the angular resolution is improved by about 32.8%, and the effectiveness and engineering application value of the method in high-resolution MIMO sparse array design are proved.

The application provides a curve reconstruction method and device based on deep learning, equipment and medium, the method comprises the following steps: obtaining the logging curve data of the non-expanded section, and establishing a training set and a test set; a density curve reconstruction model is established and the training set is used to train the density curve reconstruction model, the density curve reconstruction model is a deep learning model; the test set is used to test the density curve reconstruction model, and a reconstructed density curve is generated, the density curve reconstruction model is adjusted until the correlation of the reconstructed density curve and the original density curve of the non-expanded section and the relative error of the density curve reconstruction model reach a preset condition; the density curve of the target well is reconstructed by using the adjusted density curve reconstruction model. The application simplifies the operation process, especially in complex formation conditions, can accurately reflect the actual situation of the formation, and improves the calculation accuracy of the reservoir parameters.

The application discloses a muscle synergy analysis method based on surface electromyogram signals, and comprises the following steps: S1, collecting multi-channel surface electromyogram signals; S2, determining the number of muscle synergies by taking explained variance as the standard of the number of muscle synergies; S3, extracting muscle synergy features, and according to the envelope signal matrix Z i×r and the number of muscle synergies n, performing non-negative matrix decomposition on the envelope signal matrix Z i×r , and introducing a sparse constraint in the decomposition process, so that the process of extracting muscle synergy features is converted into an optimization problem; S4, obtaining a reconstructed matrix HY res obtained according to the optimization problem in step S3, that is, a decomposition result; S5, performing normalization processing on the synergy structure matrix H res , obtaining muscle activation states under different muscle synergy modules and analyzing the muscle activation states, obtaining a muscle synergy mode in the movement process, and calculating a reconstruction precision and a sparse degree. The method can achieve higher reconstruction precision and sparse degree in muscle synergy analysis, so that the quality and interpretation ability of data decomposition are improved.

The application discloses an ECG signal acquisition method and device based on an rPPG signal, equipment and a medium, relates to the technical field of computer software, and comprises the following steps: obtaining a PPG signal based on an rPPG signal, and generating rPPG conditional coding through a PPG encoder. The low-frequency, high-frequency and spatial domain ECG tokens of the mask state are used as the current mask token, the rPPG conditional coding is used as the constraint, and the mask reconstruction is completed through a bidirectional Transformer. The mask is gradually removed according to the confidence, and the reconstruction is circularly performed until all the masks are removed to obtain target tokens. The low-frequency and high-frequency feature tokens in the target tokens are subjected to corresponding decoding and inverse short-time Fourier transform to generate a frequency domain ECG waveform; and the spatial domain feature tokens are subjected to an ECG decoder to obtain a spatial domain ECG waveform. The target ECG signal is obtained by fusing the two waveforms, and a high-precision and high-fidelity ECG signal can be obtained based on the rPPG signal.

This invention relates to the field of ionospheric technology, specifically to a method for reconstructing ionospheric scintillation maps based on neural network error compensation. First, ionospheric scintillation observation data from discrete GNSS monitoring stations and relevant background environmental parameters are acquired. After preprocessing, ordinary Kriging interpolation is used to generate an initial ionospheric scintillation map. The difference between the interpolated estimated values ​​and the actual values ​​at each station is extracted as the interpolation residual. Combined with the background environmental parameters, the original dataset is obtained. Samples are constructed using a sliding window mechanism and input into a Long Short-Term Memory (LSTM) network model for training, resulting in an LSTM residual prediction model. The environmental parameters of the area to be reconstructed are input into the trained model to output the prediction residual. Based on spatial distance, a Gaussian decay function is used to adaptively allocate weights, and the residual is dynamically fused with the initial interpolated map to obtain a high-fidelity reconstructed map. This invention effectively solves the smoothing degradation problem of traditional interpolation in blind areas without stations, while also considering robustness in densely populated areas, thus improving the accuracy of global ionospheric scintillation reconstruction.