679 results about "Deconvolution" patented technology

The invention relates to the technical field of fabric detection, and discloses a fabric defect intelligent detection method and system based on AI visual identification. According to the method, motion blur is quantized through motion state data, optical blur caused by fabric motion is eliminated through deconvolution solution, so that motion interference in the fabric transmission process is processed in a targeted mode, self-adaptive balance of the deblurring capacity and the feature retention capacity is achieved, and then based on the optical interference principle, the deblurring capacity and the feature retention capacity are improved. Through a dynamic calibration system combining hardware-level real-time compensation and multi-dimensional optical parameter calibration, dynamic optical parameter calibration of primary correction data is realized, then fabric defect characterization data is extracted to accurately obtain defect features, and finally, a detection-production line control closed loop is constructed through a quality quantitative index and a comprehensive risk value, so that fabric defect detection is realized. The fabric defect detection precision can be improved, so that the problem of high defect missing detection and false detection rate caused by optical data distortion due to movement and environment interference in a traditional method is effectively solved.

The invention provides a medical image segmentation method and system, computer equipment and a storage medium, and belongs to the field of image processing, and the method comprises the steps: extracting preliminary features of a medical image through depth separable convolution, and splicing the preliminary features with original image residuals to obtain a preliminary feature map; after an encoder performs average pooling dimension reduction, local details and global contour features of a dimension reduction feature map are extracted by using left and right branches of a lightweight convolution module LDB, then a downsampling feature map is obtained through channel attention CA weighted fusion, and attention is calculated in combination with a self-attention mechanism module EMHA to obtain a depth feature map and a bottleneck feature map; the decoder weights the depth feature map by means of a channel and space attention to obtain a CBAM enhanced feature map, upsamples the bottleneck feature map and then splices the bottleneck feature map with the CBAM enhanced feature map, features are extracted through an LDB module, and finally a pixel-level segmentation result is output through upsampling and deconvolution, so that image segmentation achieves the effects of high quality, low complexity and low operand.

The invention provides a spatial domain identification method based on data interpolation and cell type deconvolution, and belongs to the technical field of bioinformatics. In order to solve the problems that gap information between adjacent points cannot be utilized in low-resolution spatial transcriptome data and prior information of cell types in a tissue space structure level cannot be fully integrated in a traditional method, the method comprises the following steps: acquiring a spatial transcriptome data set and a single-cell RNA sequencing data set, and performing data preprocessing on the acquired data sets; and carrying out data interpolation on the preprocessed spatial transcriptome data, and carrying out cell type deconvolution in combination with single-cell RNA sequencing data. And constructing a deep learning model based on the graph convolutional network. And training a deep learning model according to gene expression information, spatial position information and cell type information of the spatial transcriptome data after cell type deconvolution by using a self-supervised contrast learning strategy. And performing spatial domain identification on the to-be-detected data based on the trained model.

The invention discloses a fluorescence lifetime imaging phase analysis method based on deep learning, a terminal and a readable storage medium, and the method comprises the steps: obtaining counting data of time-dependent single photon counting of each pixel in a measured sample image, and measuring a response function of a used instrument, inputting the convolution attenuation data and the response function into a deconvolution neural network; the deconvolution neural network outputs a deconvolution signal of each pixel according to the response function and the counting data of each pixel; and performing phase transformation according to the deconvolution signal of each pixel, and obtaining the fluorescence lifetime of each pixel. According to the method, the time migration caused by the instrument response is regarded as the convolution caused by the instrument response function, and then the neural network is adopted to carry out deconvolution on the counting data, so that the time migration caused by the response function is calibrated and corrected, a traditional phase calibration process is not needed, and accurate analysis of the service life is realized.

The invention relates to a power load prediction system based on time sequence self-supervised representation learning, and the system comprises a data preprocessing module, an expansion time convolution network, a trend representation decoupler, a seasonal representation decoupler, a feature converter, and a predictor. The expansion time convolution network performs dynamic adaptive grouping on different variables in the same data processing unit, establishes an intra-group variable relationship by sharing a convolution kernel weight and stacking a plurality of expansion time convolution layers, and establishes an inter-group variable relationship through subsequent multi-group representation splicing operation and single-layer expansion time convolution; the trend representation decoupler is used for separating trend representation Z (T) from feature representation extracted from the expansion time convolution network based on a plurality of parallel one-dimensional causal convolution blocks with different scales; the seasonal representation decoupler adopts discrete Fourier transform to separate seasonal representation Z (S) from the feature representation extracted from the expansion time convolutional network; the feature converter maps a combined feature vector of trend representation and seasonal representation back to an original space from a potential space by stacking a plurality of deconvolution layers; and the predictor maps the feature representation of the original space into a power load prediction result by using a linear projection layer, and the adaptability, robustness and generalization ability of the model for modeling complex power load time series data can be improved.

The invention provides a spectral data optimization method and device, computer equipment and a storage medium, and relates to the technical field of semiconductor detection.The method comprises the steps that original spectral data output by a spectrograph is obtained, and the central wavelength value of the original spectral data of each frame is calculated; querying from a preset database according to the central wavelength value to obtain a corresponding instrument linear function; and processing each frame of original spectral data and the corresponding instrument linear function by adopting a deconvolution algorithm to obtain optimized real spectral data. According to the scheme, algorithm compensation is carried out on each frame of spectral data, so that system distortion introduced by the spectrometer can be accurately compensated in a full spectrum range, the extraction precision of spectral features is greatly improved, and the magnitude order improvement of measurement precision is realized; in addition, the optimized spectral data more truly reflects the physical characteristics of the tested sample, the dependence of an optical model on empirical parameters is reduced, and the generalization ability and prediction reliability of the model are improved.

The invention provides a deep learning-based oral cavity image recognition method, a storage medium and a deep learning-based oral cavity image recognition system. The method comprises the steps of deploying a federated learning framework and collecting a multi-modal oral cavity image data set; extracting local features to obtain image features, and generating a modal adaptive weight map; a multi-head self-attention mechanism is used for fusing the cross-modal features to generate a fused feature map, and deconvolution up-sampling is carried out to form high-resolution multi-modal feature representation. A tooth segmentation mask is generated based on this representation, and an initial diagnostic report is generated. And aggregating the attention weight of each client through an encryption protocol, and generating interpretable decision support data. And finally, generating a structured clinical report by using a natural language. According to the method, the Grad-CAM thermodynamic diagram is combined with the encrypted and aggregated attention weight, so that the privacy security is guaranteed, the model interpretability is enhanced, the clinical credibility and the diagnosis decision efficiency are improved, and the problems of insufficient diagnosis precision of complex lesions and insufficient utilization of multi-modal information in the prior art are solved.

The invention discloses an optical communication network operation and maintenance fault reporting, positioning and tracking system, which relates to the technical field of optical communication network operation and maintenance, and comprises a coarse positioning module for performing phase demodulation and interference signal matching on optical fiber echo data, identifying a sudden change point and an abnormal interval, and generating a potential fault interval and a coarse positioning coordinate; the fine positioning module is used for transmitting a coded pulse signal to a link where the optical fiber echo data is located according to the potential fault interval and the coarse positioning coordinate, and performing matched filtering and deconvolution processing to generate a fine positioning echo characteristic curve; the fault identification module is used for carrying out deep learning classification on the fine positioning echo characteristic curve, identifying a fault type and generating a diagnosis result; and the visualization module is used for mapping the diagnosis result to the topology and geographic position information of the link section where the optical fiber echo data are located, and generating a visual interface. According to the invention, the manual troubleshooting time is obviously reduced, and the fault processing efficiency and reliability are improved.

The embodiment of the invention relates to the technical field of bioinformatics, and provides a bulk data cell analysis method, device and equipment, and the method comprises the following steps: constructing an initial reference matrix according to a single cell data set and a cell type annotation template, each element in the initial reference matrix represents the gene expression quantity of each cell state under each characteristic gene; performing deconvolution on the bulk data to be analyzed according to the initial reference matrix to obtain a first deconvolution result; updating the initial reference matrix according to the first deconvolution result to obtain a first reference matrix; and according to the first reference matrix, performing deconvolution on the bulk data to be analyzed to obtain a second proportion and a second gene expression quantity of each cell type in the bulk data to be analyzed. According to the embodiment of the invention, the accuracy of cell analysis in bulk data can be improved.

Methods for determining a location of a feature in a spatial array with features include: (a) providing an array with a first set of one or more features immobilized on a substrate, a first feature of the first set having a first barcoded oligonucleotide with a first spatial barcode and a first constant sequence, and a second set of one or more features immobilized on the substrate, a second feature of the second set having a second barcoded oligonucleotide with a second spatial barcode and a second constant sequence; (b) attaching the first constant sequence to the second constant sequence to generate a nucleic acid product; (c) determining all or a portion of a sequence of the nucleic acid product or a complement thereof; and (d) associating the second barcoded oligonucleotide with the first barcoded oligonucleotide in the nucleic acid product.

The invention discloses a CT (Computed Tomography) image pulmonary embolism segmentation and classification method combined with quality evaluation, which relates to the technical field of image processing, and comprises the following steps: inputting a 256 * 256 pulmonary embolism CT image and a quality score thereof into a quality score guide encoder, expanding a quality score dimension through linear transformation, and carrying out point product fusion with a feature map extracted by ResNet34 layer by layer to obtain a final product; generating multi-scale coding features; performing wavelet domain decomposition and reconstruction on the coding features through a wavelet transform fusion jump link module, and optimizing feature transmission; a multi-scale cross enhanced decoder is adopted to carry out multi-scale deconvolution fusion on the features, a segmentation result is output in combination with an efficient channel attention mechanism, meanwhile, pulmonary embolism existence judgment is output through a classification head, and the method provides powerful support for early diagnosis of pulmonary embolism, development of an image auxiliary diagnosis system and clinical application, and has good application prospects. Wide application prospects and profound social significance are realized.

The invention relates to the technical field of pulse electric field measurement, and discloses a time drift compensation method of a pulse electric field measurement system based on model parameterization. According to the method, a parameterized model is constructed in simulation software, and the parameterized model is additionally provided with two servo zeroing compensation circuits on the basis of an original measurement system, and the two servo zeroing compensation circuits are connected to the output ends of two in-phase operational amplification circuits respectively to form a negative feedback loop used for dynamically suppressing low-frequency baseline drift; acquiring electric field time domain waveform data of the output end of the third in-phase operational amplifier circuit; calculating a transfer function of a pulsed electric field measurement system in the parameterized model, converting the transfer function from a frequency domain state to a time domain state through inverse Fourier transform, and then performing deconvolution on the transfer function of the time domain state and electric field time domain waveform data to obtain real electric field intensity; according to the invention, time drift caused by operational amplifier input imbalance, power supply fluctuation and temperature change is automatically compensated in real time, and the stability and precision of system measurement are improved.

The invention relates to an overlapped chromatographic peak positioning and splitting method and system based on a numerical stable function. The method comprises the following steps: de-noising chromatographic data and correcting a baseline; calculating a second derivative by using SG filtering, grouping continuous points according to a negative threshold value, and determining an initial position of each peak by using a second-order minimum value; establishing a peak function of bidirectional exponential correction by taking a peak position as an initial value, setting parameters and boundaries such as area, center, width, trailing and the like, constructing a residual error minimization target, and performing robust fitting by adopting a trust region reflection algorithm; and indexes such as suitability, output peak area, center, broadening, separation degree, theoretical plate number and the like are verified by residual errors. According to the method, valley point or prior component information is not needed, the method is insensitive to noise and baseline drift, serious overlapping and trailing / leading edge peaks of multiple components can be effectively processed, convergence is high, the method is insensitive to initial values, parameters can be expanded to different peak shapes, and the method is suitable for qualitative and quantitative analysis of automatic and high-flux chromatographic data.

The invention belongs to the technical field of gravity and magnetic method exploration data processing, and relates to a source body positioning method based on intelligent gravity and magnetic field separation and Euler deconvolution, which comprises the following steps: setting gravity and magnetic abnormal field modeling parameters according to geological and physical property data of a research area; generating an initial seed point in the underground three-dimensional grid and expanding and growing the initial seed point to generate Q independent abnormal field models; mapping the position of the model to a real coordinate, and calculating to obtain a forward modeling abnormal field data set; extracting the maximum value of the global absolute value of the data set and sorting in an ascending order, superposing the first p absolute values to generate a composite field, and constructing a field separation data set by taking the p absolute value as a target field and the composite field; using the field separation data set to train the field separation network until convergence; inputting measured data score departure field and residual field data, and selecting data for iterative separation according to requirements to obtain a field separation result; and obtaining a preliminary positioning result by using a field separation result and combining with an Euler deconvolution method, and determining final source body positioning after screening. The method has the beneficial effects that multi-scale geological target field separation can be realized, and source body positioning can be determined.

The invention discloses a living cell super-resolution imaging method and system based on self-supervised learning. The method comprises the following steps: S1, obtaining a fluorescence microscopic image through a standard fluorescence microscopic system; s2, for the acquired single noise image, generating positive and negative sample pairs of self-supervised training data through an autonomously designed and optimized self-supervised strategy; s3, constructing a front denoising neural network and a rear deconvolution network, and performing network training; and S4, inputting a newly obtained noise image into the trained front denoising network to obtain a denoised image, and inputting the denoised image into the rear deconvolution network to obtain a final super-resolution reconstructed image. According to the invention, the standard fluorescence microscopic system and the self-supervised denoising processing module are combined, so that high-quality denoising and super-resolution reconstruction of the fluorescence microscopic image can be realized in a low-photon signal scene.

The present application relates to the technical field of speech processing, and particularly relates to a mixed speech processing method, an electronic device and a computer readable medium. The method comprises the following steps: collecting mixed speech and environmental influence parameters; performing bionics frequency domain analysis on the mixed speech to obtain low-frequency attenuation compensation feature data; performing multipath effect propagation analysis on the low-frequency attenuation compensation feature data through the environmental influence parameters to generate channel distortion data; performing time domain-frequency domain joint deconvolution processing on the mixed speech by using the channel distortion data to generate direct sound components and reflected sound components; performing adversarial training based on the direct sound components and the reflected sound components to generate anti-multipath speech enhancement data; and constructing a dynamic frequency compensation filter based on preset environmental acoustic characteristics. The present application improves the output quality of mixed speech through multi-stage signal processing, frequency compensation and real-time optimization technology.

The invention relates to an embankment slope stability assessment method based on artificial intelligence, and belongs to the technical field of embankment slope monitoring and assessment. The method comprises the following steps: acquiring and marking embankment slope stress sensing data; dividing the data into a plurality of spatio-temporal data blocks; constructing a stability evaluation model; extracting multi-scale convolution features by adopting multi-scale cavity convolution of a high-frequency channel and pooling-deconvolution operation of a low-frequency channel to obtain a fused multi-scale feature matrix; a hidden state sequence is obtained through a space attention mechanism and a double-door-setting mechanism; calculating time interval saliency based on the hidden state vector, then calculating a weighted feature vector, further obtaining a weighted feature matrix, and processing through deep convolution and point-by-point convolution to obtain a pooling feature vector; carrying out stability evaluation grade classification through learnable category prototype and gating feature transformation; and dynamically adjusting sample weight and constraint attention distribution by adopting a total loss function. According to the method, the progressive instability identification capability can be improved.

The invention belongs to the technical field of earth exploration, and relates to an underground imaging method based on a reformed Marchenko method, which comprises the following steps: removing direct waves from seismic data, and carrying out seismic deconvolution to obtain preprocessed seismic data; selecting an underground imaging point, setting the underground imaging point as an underground focus point, and estimating a direct wave Green function from the underground focus point to a ground surface receiving point according to a background speed model; the direct wave Green function is reversed on a time axis to serve as an initial downlink focusing function, and an initial uplink focusing function is set to be zero; and performing iterative calculation according to the initial downlink focusing function and the initial uplink focusing function to obtain a downlink focusing function and an uplink focusing function after iteration, calculating a downlink green function and an uplink green function according to the uplink focusing function and the downlink focusing function, and imaging the underground imaging point. According to the method, iteration absolute convergence can be ensured by mediating convergence parameters, and finally an underground imaging result which is not influenced by multiple waves is obtained.

The invention discloses a cross-border intelligent advertisement putting optimization method and system based on reinforcement learning, and aims to solve the problems of delayed sparse reward reconstruction and credit distribution under the conditions of privacy attribution and cross-border settlement. The method comprises the following steps: reconstructing instant conversion intensity and estimating uncertainty by constructing a reversible deconvolution integral solution masked by a privacy window; the net profit is calibrated according to a unified reference currency by combining a hierarchical Bayesian model, and bidding, orientation, material selection and budget allocation are optimized under the budget constraint based on risk sensitive reinforcement learning of condition risk value; the technical effects of accurately reconstructing rewards under privacy limitation and cross-border cost fluctuation, stabilizing credit distribution, reducing tail risks and improving putting income and compliance are achieved.

The invention discloses a sound source localization method based on multi-frequency separation and Newton optimization deconvolution, relates to the technical field of array acoustic signal processing, and is used for solving the problem that weak sound sources and multiple sound sources are difficult to identify. According to the method, dominant frequency is extracted through multichannel frequency domain analysis, a cross-spectrum matrix is constructed in combination with a near-field propagation model and a guide vector, delay summation beam forming is executed to obtain sound source preliminary distribution, then the distribution is regarded as a convolution result, a maximum likelihood model is introduced, and a two-stage deconvolution strategy of coarse estimation and Newton method fine optimization is adopted to obtain a high-resolution sound source. According to the multi-sound-source positioning method, subgrid-level analysis of sound source positions and amplitudes is achieved, finally, all frequency results are fused, continuous sound source images are smoothly output through a two-dimensional Gaussian kernel, the resolution and real-time performance of multi-sound-source positioning are remarkably improved, and the multi-sound-source positioning method is suitable for high-precision acoustic imaging in a complex sound field.

Methods for determining a location of a feature on an array include: (a) providing a first array with a first plurality of features immobilized on a first substrate; (b) providing a second array with a second plurality of features immobilized on a second substrate; (c) aligning the first array with the second array; (d) hybridizing a first barcoded oligonucleotide of the first array to a second barcoded oligonucleotide of the second array, thereby producing a combined nucleic acid that includes first and second spatial barcodes; (e) determining all or a portion of the sequence of the combined nucleic acid; and (f) identifying the second barcoded oligonucleotide associated with the first barcoded oligonucleotide in the combined nucleic acid, and determining the location of a second feature in the second array.

Methods for determining a location of a feature on a spatial array include (a) providing an array of features on a substrate, where a feature of the array includes a barcoded oligonucleotide having, in a 5′ to 3′ direction, a spatial barcode, a cleavage domain, and a constant sequence; (b) hybridizing a priming oligonucleotide to the constant sequence; (c) extending the priming oligonucleotide using the barcoded oligonucleotide as a template; and (d) determining all or a portion of a sequence of the extended priming oligonucleotide corresponding to the spatial barcode, or a complement thereof, and a location of the extended priming oligonucleotide, and using the location of the extended priming oligonucleotide to determine the location of the feature on the spatial array.

The invention provides a deep-sea large-aperture array deconvolution near-field high-precision direction finding method based on sound velocity correction. According to the method, an equivalent sound velocity mapping table generated by a BELLHOP ray model is established, a sound ray bending effect is innovatively converted into a real-time table look-up correction mechanism, sound velocity correction time delay is injected in beam forming and deconvolution calculation, the problem of PSF distortion of a traditional linear model in a deep sea near field is thoroughly solved, high-precision direction finding and distance measuring are achieved, and the accuracy of direction finding and distance measuring is improved. And a subversive direction finding solution with physical accuracy, algorithm stability and engineering real-time performance is provided for the deep-sea large-aperture array.

The invention discloses a pipeline inner wall defect detection method and device based on a pipeline robot and a medium, and relates to the technical field of pipeline detection.The method comprises the steps that multi-source detection data of the inner wall of a pipeline are collected and preprocessed, and a comprehensive detection data set of the inner wall of the pipeline is generated; based on the pipeline inner wall comprehensive detection data set, spiral sampling deconvolution processing is carried out to obtain pipeline magnetic signal cylindrical expansion domain data after deconvolution, and a resolution improvement index and an edge feature definition index are generated; calculating defect feature operator distribution according to the deconvoluted pipeline magnetic signal cylindrical expansion domain data and the pipeline inner wall comprehensive detection data set, and generating a suspected position point set; and through the suspected position point set, calculating a co-gradient co-orientation index and carrying out connected domain aggregation to generate a defect judgment set. According to the invention, through spiral sampling deconvolution processing, fine recovery of pipeline magnetic signal space distribution is realized.

The invention discloses an AI-assisted protein purification result analysis method and system. The method comprises the following steps: S1, collecting chromatogram data, electrophoresis image data and mass spectrum peak table data; s2, performing data preprocessing after the chromatogram data, the electrophoresis image data and the mass spectrum peak table data are obtained; s3, carrying out AI identification analysis on the preprocessed data to realize electrophoretic band identification, chromatographic peak identification and mass spectrum deconvolution; s4, performing multi-modal fusion on the chromatography, gel electrophoresis and mass spectrum information sources; and S5, automatically generating a report, and producing a visual chart. The deep learning and multi-modal data fusion technology is introduced, comprehensive automatic analysis of chromatographic data, electrophoresis images and mass spectrum results is achieved, then automatic result analysis of the protein purification process is achieved, an artificial intelligence algorithm is used for recognizing a peak structure and an electrophoresis band, multi-modal data fusion is achieved, and the detection accuracy is improved. And generating a visual analysis report according to a data result.

The invention belongs to the technical field of seismic data processing in geophysical exploration, and relates to a Marchenko multiple suppression method for correcting wave field mismatch based on Wasserstein distance, which comprises the following steps: acquiring a pulse reflection sequence through preprocessing and deconvolution; a pulse reflection sequence is input, and Marchenko iteration initialization is carried out; and remodeled Marchenko iteration for correcting the wave field mismatch based on the Wasserstein distance is carried out. According to the method, the optimal transmission theory is systematically introduced into the Marchenko multiple wave suppression field for the first time, the adaptive capacity of the Marchenko method to actual data noise, defect and amplitude distortion is remarkably improved while the advantage that the Marchenko method does not need a speed model is kept, and therefore higher-precision multiple wave suppression and more reliable primary wave recovery are achieved, and the method is suitable for popularization and application. The method is used for high-precision seismic imaging in a complex land exploration environment. The method has remarkable advantages in the aspects of constraint mechanism, error control and global feature protection.

The invention discloses a low-vibration multi-point delay digital twin drive intelligent electronic detonator blasting control method, and relates to the technical field of blasting vibration control, and the method comprises the steps: 1, constructing a frequency weight curve and a weighted energy index, and unifying a time scale and three-component measurement; 2, establishing a virtual field by using a single-hole signature library, instrument transfer function deconvolution and a boundary correction propagation operator; step 3, performing delay optimization in combination with a timing characteristic function and a partition phase, and setting a keyhole read-only and weighted threshold; fourthly, programming backward reading is conducted, gating is conducted through a consistency index and an admission ratio, only one-time fine adjustment is conducted on the non-key hole on line according to an expected energy track and deviation statistics, and block assimilation updating is conducted later. According to the scheme, quantitative suppression of sensitive frequency band energy is achieved, and construction efficiency and traceability are kept. Through unified parameter and data object through identification, database building, optimization, execution and feedback, the scheme consistency is improved, the trial explosion frequency is reduced, supervision and acceptance are facilitated, and redisk tracing is supported.

The invention discloses a Fourier finite difference deconvolution imaging method based on inclination angle adaptive noise suppression, and relates to the technical field of geophysical exploration of petroleum. The method specifically comprises the following steps: designing a scattering point speed model to solve a partial point spread function of an underground space; solving a partial point spread function by a speed disturbance method; solving point spread functions of all positions of the underground space by carrying out Gaussian filtering technology and inverse distance weighted interpolation on part of the point spread functions; using a plane wave destruction method to solve the inclination angle attribute of the work area, and constructing an inclination angle self-adaptive diagonal diffusion Fourier finite difference migration method based on the inclination angle attribute to solve a high-quality initial image suitable for an inversion system; and performing multi-dimensional deconvolution in the wave number domain to obtain a final migration imaging result. According to the method, the point spread function is successfully introduced into Fourier finite difference imaging, and the resolution and deep amplitude of a conventional Fourier finite difference imaging method can be effectively improved.

According to the infrared image enhancement method based on style migration and multi-level fusion, a style migration preprocessing step comprises five levels of coding and decoding modules, multi-scale feature fusion is realized through cross-level dense jump connection, an encoder compresses spatial dimensions step by step, and a decoder recovers resolution through deconvolution and cancels a batch normalization layer. Information loss of small target features in the standardization process is avoided, and a triple mixing loss function is adopted for training; in the super-resolution reconstruction step, curvelet decomposition is carried out on a style migration result, curve features such as an arc line are accurately captured by adopting a 36-direction wedge-shaped basis function, a local variance threshold mechanism is adopted for high-frequency sub-band fusion, and a style migration feature weight is given to a high-variance region; low-frequency sub-bands are dynamically weighted according to regional contrast, a residual dense block enhanced SRGAN architecture is adopted in a reconstruction stage, and training stability is improved in cooperation with a spectrum normalization discriminator. Through the mode, the technical problem of effectively improving the quality of the low-quality infrared image is solved.

The invention relates to the technical field of biological radar signal processing, in particular to an FMCW radar-based heart beat interval estimation method, which comprises the following steps of: S1, converting a chest vibration echo phase time sequence obtained by irradiating a chest area of a monitored object by an FMCW radar into an acceleration time sequence by using a second-order time derivative; s2, dynamically mapping the center frequency and the wavelet order in a preset frequency analysis interval, constructing a self-adaptive wavelet dictionary, then executing multi-order wavelet time domain convolution operation on the acceleration time sequence by using the dictionary, and aggregating time-frequency energy distribution to generate a time-frequency energy diagram; and S3, performing a deconvolution operation reconstruction strategy based on an energy-guided multi-stage time-frequency feature screening technology and wavelet function conjugation, generating an approximate time domain signal of heart beat vibration, and completing heart beat information inversion. According to the method, the accuracy and robustness of IBI extraction can be effectively improved under the condition of low signal-to-noise ratio, so that stable monitoring of light and moderate HRV (heart rate variability) is supported.