512 results about "Discrete wavelet transform" patented technology

The invention relates to a power distribution network state estimation method based on an adaptive space-time diagram neural network, and the method mainly comprises the following steps: S1, collecting historical and real-time measurement data of a power distribution network, and carrying out the preprocessing of the data, so as to guarantee the integrity of the data, provide high-quality input data for a model, and improve the estimation precision and stability of the model; s2, discrete wavelet transform is carried out on historical measurement data, multi-scale decomposition is achieved, and low-frequency and high-frequency components are extracted; a double-branch time sequence fusion module is constructed, global trend and local fluctuation features are respectively captured through a dynamic attention mechanism and a time convolution network, and the features are efficiently fused by means of adaptive weights. S3, in the real-time data processing process, branch measurement features are extracted through a multi-layer perceptron (MLP) and mapped to nodes of the whole network, dynamic integration of historical data and real-time data is achieved, and therefore the real-time performance and accuracy of state estimation of the power distribution network are improved.

The invention relates to the field of image data processing, and discloses a texture preserving type image denoising and enhancing method based on a generative adversarial network, which comprises the following steps: acquiring an original image signal, and calculating low-frequency sub-band data and high-frequency sub-band data by using discrete wavelet transform; calculating the gradient magnitude of the low-frequency sub-band data to generate a structural significance gradient map; establishing a reverse mapping relation based on the structure saliency gradient map, and generating a spatial self-adaptive dynamic gating threshold; performing statistical gating on the high-frequency sub-band data by using the dynamic gating threshold to generate a high-pass gain coefficient and a low-pass suppression coefficient; according to the method, cross-band modulation logic of the structure flow to the texture flow is established, so that the problem that weak texture signals are easy to lose under non-uniform illumination is solved, and non-structured noise filtering and structured micro texture restoration are realized on the premise of not depending on semantic tags.

The invention provides an image super-resolution reconstruction method based on double-domain feature fusion and implicit representation, and relates to the field of image processing and computers, and the method comprises the steps: obtaining a low-resolution remote sensing image, and carrying out the preprocessing of the low-resolution remote sensing image; performing feature extraction on the preprocessed remote sensing image through Haar discrete wavelet transform and a Transform-based pyramid structure to obtain frequency domain features and spatial domain features; performing double-domain cross attention fusion on the frequency domain features and the spatial domain features to obtain local detail features and global structure features; and through an implicit representation network, the fused local detail features and global structure features are mapped to any space coordinates, a final three-channel high-resolution image is obtained, and high-quality reconstruction of a remote sensing image of any scale is realized. According to the technical scheme of the invention, the implicit neural representation is guided to realize higher-precision image reconstruction through the cooperative expression of the frequency domain information and the spatial domain information.

The invention relates to the technical field of mode recognition and data recognition, and discloses an intelligent visual detection method for non-standard precision part surface microdefects, which comprises the following steps: acquiring surface gray level image data of a to-be-detected part, physically abandoning low-frequency components through discrete wavelet transform, and reserving high-frequency detail components to construct a frequency domain input tensor; constructing a double-flow reconstruction model containing a space domain coding network and a frequency domain coding network, and minimizing the distribution difference of the same feature between double-domain characterization through potential feature space consistency constraint joint optimization; the method comprises the following steps of: calculating a spatial domain residual image and a frequency domain residual image, combining a texture topological residual image extracted by structural tensor characteristic decomposition, and generating a comprehensive abnormal response image through weighted fusion to judge the defect, and effectively inhibiting macroscopic geometric contour interference through frequency domain decoupling and a topological check mechanism on the premise of not needing a standard geometric template. And sensitive perception and accurate identification of weak texture defects on the surface of the non-standard part are realized.

The invention provides a dense overlapping target detection method based on a wavelet enhancement sparse hybrid expert model. The method comprises the following steps: firstly, extracting multi-layer features through a backbone network to capture multi-scale spatial representation; secondly, discrete wavelet transform is introduced to each level of features, spatial features are decomposed into a frequency domain, collaborative modeling of frequency domain and spatial domain features is realized, the reservation capability of detail and texture information is improved, a lightweight dynamic hypergraph aggregation module is introduced into the deepest layer of features, a hyperedge structure is adaptively learned, and the feature fusion is realized; modeling a high-order incidence relation in a local area in an explicit manner; and thirdly, in the decoding process, candidate queries are screened and reweighted through an IoU perception query selection mechanism, and a dynamic routing mechanism of sparse hybrid experts is introduced, so that query self-adaptive specialized representation learning is realized, and the target detection precision and reliability in a complex scene are effectively improved.

The invention provides an infrared image frequency division enhancement method based on a diffusion model, and the method comprises the steps: firstly carrying out the multi-scale frequency domain decomposition of an infrared image through discrete wavelet transform, and dividing the image into a low-frequency component and a high-frequency component; in the low-frequency component processing process, a diffusion equation is constructed in combination with a thermal radiation physical model, and discretization solution is performed by using a finite difference method. The high-frequency component is converted from a space domain to a frequency domain through Fourier transform, energy distribution of the high-frequency component is analyzed, and a diffusion probability function is adjusted. In the fusion stage, according to the signal-to-noise ratio of the local area of the image, the fusion weight of the low-frequency component and the high-frequency component is dynamically adjusted, the processing results of all the local areas are integrated in a weighted fusion mode, and finally the high-quality enhanced infrared image is generated. The method has excellent performance in improving image contrast, definition and detail performance, is suitable for the fields of security monitoring, military reconnaissance, industrial detection and the like, and can effectively improve monitoring, analysis and detection effects.

The invention discloses a 3D medical image segmentation method based on a diffusion model and double-attention discrete wavelet transform, and relates to the technical field of medical image segmentation. Comprising the following steps: acquiring a 3D medical image data set, and dividing into a training set, a verification set and a test set in proportion; constructing a 3D medical image segmentation model based on a diffusion model and double-attention discrete wavelet transform; inputting the training set into a 3D medical image segmentation model for model training; inputting the verification set into the trained 3D medical image segmentation model for model verification; and inputting the test set into the verified 3D medical image segmentation model for image segmentation. According to the method, the diffusion model and the full convolutional neural network structure are organically fused, discrete wavelet transform and a double-attention mechanism are introduced, and the accuracy and robustness of 3D medical image segmentation are remarkably improved.

The invention belongs to the technical field of wastewater treatment, and provides a wastewater treatment equipment remote control system based on the Internet of Things, which aims at solving the problems that the traditional fixed DO concentration control cannot adapt to water inlet load fluctuation, the effluent is easy to be substandard or the aeration energy consumption is high, the monitoring data noise is large, and the time sequence is misplaced. According to the scheme, a multi-dimensional real-time sensing network is constructed, flow, COD, BOD, ammonia nitrogen, water temperature, distributed DO and sludge activity sensors are deployed, discrete wavelet transform is adopted for noise reduction, and a data time sequence is aligned; a three-layer load-DO-energy consumption dynamic correlation model is designed, a basic layer predicts a load trend through LSTM, a middle layer quantifies a DO demand through a microbial metabolism model, and an optimization layer outputs an optimal DO set value through a PPO algorithm; and developing a dynamic decision-making system, adaptively generating a DO interval according to a load state, and adjusting fan parameters and number in a linkage manner. According to the invention, load dynamic adaptation is realized, effluent ammonia nitrogen is guaranteed to reach the standard, aeration energy consumption is reduced, and remote control precision and equipment operation efficiency are improved.

The invention discloses a frequency domain stealth backdoor attack method based on a generative adversarial network, and belongs to the technical field of image data processing, and the method comprises the steps: obtaining a clean sample set D1 and a classification model; constructing a generative adversarial network, and constructing a loss function based on picture similarity loss, frequency domain consistency loss and adversarial loss; training the generative adversarial network by using the D1 to obtain a generative model; constructing a clean data set and a backdoor data set based on the D1, a target label t of the backdoor attack and the generative model; constructing multi-layer MMD loss; and constructing total loss based on MMD and MSE, and performing poison training on the classification model to obtain a backdoor model. According to the method, a multi-domain disturbance generation network is constructed, pixel-level disturbance is generated in a spatial domain, discrete wavelet transform is introduced into a frequency domain to constrain frequency domain characteristics of back door disturbance, poisoning samples have concealment in the spatial domain and the frequency domain, multi-layer MMD loss is introduced, the deep characteristic distribution difference between clean samples and back door samples is reduced, and high-concealment attack is achieved.

The invention discloses an electroencephalogram signal processing method and system based on motion artifact prediction. The method comprises the following steps: acquiring experimental electroencephalogram data when a testee executes a motion imagination task; discrete wavelet transform is carried out on experimental electroencephalogram data, and signals are decomposed into low-frequency components and high-frequency components through a low-pass filter and a high-pass filter; inputting the low-frequency component into a pre-constructed and trained ARIMA model, and predicting to obtain a linear artifact; inputting the high-frequency component into a pre-constructed and trained XGBoost regression model, and predicting to obtain a nonlinear artifact; combining the linear artifacts and the nonlinear artifacts to generate a complete artifact prediction signal; real electroencephalogram signals are separated through difference value calculation of the experimental electroencephalogram data and the artifact prediction signals. According to the method, the time sequence change of the motion artifacts is predicted by utilizing the ARIMA model, and the nonlinear artifact features are captured in combination with the XGBoost regression model, so that the motion artifacts can be effectively removed, and purer electroencephalogram signals can be recovered.

The invention provides an image privacy protection-oriented reversible adversarial sample generation system and method, and relates to the technical field of artificial intelligence and image privacy protection. The system comprises an input module, a semantic guidance mask generation module, a potential confrontation generation module and a reversible steganography embedding module. The input module acquires an original image, generates a binary mask in the semantic guidance mask generation module, and generates adversarial text description by adopting a visual language model; a final image potential representation is generated in a potential adversarial generation module through forward diffusion and reverse sampling of a diffusion model, and adversarial disturbance is introduced by adopting a gradient optimization algorithm to generate an adversarial sample image; and finally, discrete wavelet transform is adopted in the reversible steganography embedding module to separate out a high-frequency sub-band, and a reversible neural network is combined to obtain a secret-containing hidden image. According to the method, the three requirements of visual authenticity, visual anonymity and machine identifiability can be met at the same time, and an innovative solution is provided for privacy protection in an advanced face recognition scene.

The invention discloses a weak supervision cell nucleus segmentation method based on wavelet differential convolution and region expansion, and relates to the technical field of image processing, and the method comprises the following steps: obtaining an image sample containing a cell nucleus, and marking the position of the cell nucleus through central point annotation; a wavelet differential convolution module is designed, multi-scale features are extracted through discrete wavelet transform, and cell nucleus boundary and detail information are enhanced in combination with the differential convolution module; a region expansion module is constructed, pseudo labels are generated based on point annotation iteration, a complete cell nucleus region is expanded step by step, and the problems of noise and nucleus missing detection are reduced; building a segmentation network, adding a wavelet difference convolution module to extract detail features, and optimizing network performance by using a pseudo tag as a weak supervision signal; carrying out segmentation prediction, and outputting the accurate position and shape of the cell nucleus; therefore, high-precision segmentation is realized under a small amount of annotation information, the annotation dependence is reduced, and particularly, the effects of reducing adjacent cell nucleus boundary adhesion and small cell nucleus leak detection are remarkable.

The invention belongs to the field of photovoltaic panel hot plate image processing, and particularly relates to a photovoltaic panel defect detection method fusing multi-scale wavelets and a lightweight attention mechanism. According to the method, an RHDWT discrete wavelet transform module based on multi-scale decomposition is added in a model input stage to strengthen image edge and texture detail representation; a Mix Structure Block module is introduced into a backbone network of the YOLOv11, so that multi-scale features are fused, and the feature expression capability is improved; an LWGA lightweight global attention mechanism is introduced into a neural network connection layer to enhance the context modeling capability, and the detection effect on small target defects such as fine cracks and hot spots is improved. According to the model, through collaborative optimization in three aspects of input preprocessing, feature extraction and an attention mechanism, the precision and robustness of defect detection in a complex photovoltaic module infrared or visible light image are remarkably improved, and the model is suitable for scenes such as high-precision photovoltaic panel image detection and intelligent maintenance.

The invention discloses a parallel U-Net-based dual-domain collaborative infrared and visible light image fusion method. The method comprises the following steps: firstly, extracting initial features of a source image by using dense connection blocks; then, parallel frequency domain branches and spatial domain branches are constructed, the frequency domain branches are combined with discrete wavelet transform and fast Fourier convolution to decompose and enhance multi-scale global frequency domain features, and the spatial domain branches capture long-distance spatial dependence with linear calculation complexity by using a convolutional layer and Mama based on a selective state space model; dynamic interaction and weighted fusion of double-domain information are realized through an adaptive feature fusion module; and finally, generating a fused image through an image reconstruction module. According to the method, the problems of high calculation overhead and video domain information negligence in the prior art are solved, and infrared heat radiation maintenance and visible light texture enhancement are effectively considered.

The invention discloses a remote sensing knowledge distillation method and system based on dual-mode feature spectrum decoupling, and the method comprises the steps: inputting a remote sensing image, carrying out the parallel extraction of a full-frequency feature map through a teacher model and a student model, and decomposing the full-frequency feature map into a low-frequency component and a high-frequency component through two-dimensional discrete wavelet transform; performing adaptive spatial feature enhancement on the low-frequency component and the high-frequency component through a density-independent spatial weighting mechanism to generate explicit distillation loss; inputting each component feature map output by the teacher model into a corresponding pre-trained knowledge amplifier to generate a thermodynamic diagram, predicting and capturing a decision boundary confidence implied by the teacher model through the thermodynamic diagram, and constructing implicit knowledge constraints based on student feature responses; target detection loss, explicit distillation loss and implicit knowledge constraint are combined, student model parameters are optimized through gradient propagation, and meanwhile, teacher model and knowledge amplifier parameters are frozen to guarantee knowledge migration stability.

The invention discloses a traffic flow prediction method and system based on TEDDGN, and belongs to the technical field of traffic flow prediction, and the method comprises the steps: carrying out the convolution operation of a traffic time sequence and a wavelet function through employing discrete wavelet transform, achieving the trend-event decoupling of traffic flow data, projecting a signal to different scale spaces, and carrying out the prediction of the traffic flow. Obtaining independent trend components and event components; processing the obtained trend component and event component by adopting a multi-scale time learner; spatial feature extraction is carried out on the processed data, modeling is carried out through adaptive graph convolution and dynamic graph convolution in spatial feature extraction, and the adaptive graph convolution and the dynamic graph convolution are dynamically fused and output through a learnable gating coefficient; inputting the output spatio-temporal characteristics into an output layer to form a TEDDGN prediction model; a TEDDGN prediction model is trained; and performing traffic flow prediction by using the trained TEDDGN prediction model. According to the invention, the accuracy of traffic flow prediction is realized.

The invention provides a bearing signal expansion method and system based on a chaotic particle swarm algorithm and a generative adversarial network, and the method comprises the following steps: firstly, carrying out the preprocessing of a collected bearing acceleration signal, and carrying out the denoising through employing the discrete wavelet transform (DWT) in combination with a Bayesian soft threshold method; secondly, determining a proper signal sample length according to a Nyquist sampling theorem, slicing the signal by adopting a sliding window strategy, inputting the sliced data into a chaos particle swarm algorithm for optimization, and searching an optimized vector which is closest to the structural similarity of the sliced data; and finally, inputting slice data of a real sample into a discriminator, and superposing the optimized feature vector with random disturbance to serve as initial input of a generator. In the training process, the parameters of the generator and the discriminator are mutually confronted and updated until a preset training round is reached. And the small sample problem and the class imbalance problem in bearing fault diagnosis are effectively relieved.

The invention discloses an acceleration method, device and equipment for partial discharge detection of a switch cabinet and a storage medium, and relates to the technical field of power equipment fault detection, partial discharge signals are collected in real time through a multi-parameter sensor, multi-source signal alignment is realized by combining a hardware trigger mechanism and timestamp synchronization, noise components are separated through multi-scale wavelet transform, and the partial discharge detection of the switch cabinet is realized. The method comprises the following steps: extracting time-frequency joint features through discrete wavelet transform, reducing dimensionality by adopting a principal component analysis method, taking low-dimensional feature vectors as input limited by a Gaussian mixture model (GMM), taking feature vectors limited by the GMM as input of an SVM classifier, outputting a classification result, and performing online judgment according to the classification result. When the output is partial discharge, an alarm prompt is triggered through the interface module; and when the output is normal, real-time monitoring is continued. According to the invention, signal processing and classification are accelerated by using an FPGA parallel processing architecture, and real-time early warning of power equipment faults is realized.

According to the invention, the multi-mode time-frequency domain conversion and the deep learning technology are combined, and a multi-component mixed Raman spectrum unmixing method is developed, so that clinical in-vivo and in-situ detection and disease diagnosis of novel Raman probes, instruments and the like are facilitated. The method comprises the following steps: (1) converting a mixed Raman spectrum from a time domain to a frequency domain by using fast Fourier transform (FFT), discrete cosine transform (DCT) and discrete sine transform (DST), and extracting frequency domain features; (2) extracting multi-scale local time-frequency domain characteristics of the mixed spectrum by using short-time Fourier transform (STFT) and discrete wavelet transform (DWT); (3) carrying out spectral unmixing calculation in each mode in combination with a one-dimensional attention mechanism U-shaped neural network model; and (4) fusing various modal unmixing results by using a meta-learning method, and analyzing the weight of each modal to obtain an accurate unmixing spectrum. Compared with a traditional Raman spectrum analysis method, the Raman spectrum multi-component signal unmixing method based on multi-modal time-frequency domain transformation and deep learning can accurately separate independent Raman signals of different tissue structures and biochemical components in a complex environment in a living body, so that the unmixing accuracy of the Raman spectrum multi-component signal is improved. Therefore, convenience is provided for subsequent disease mechanism analysis and diagnosis. The method provides an innovative and potential solution for in-vivo and in-situ detection analysis and disease diagnosis of medical clinical Raman spectroscopy.

The invention belongs to the technical field of computer vision and image processing, and particularly relates to a remote sensing and unmanned aerial vehicle image defogging method, a defogging network DWTMA-Net is adopted, and the DWTMA-Net is constructed based on a U-shaped architecture and comprises a discrete wavelet block DWB, a multi-dimensional attention module MAB and a wavelet down-sampling module WDM; downsampling of the encoder part is carried out by using Haar discrete wavelet transform through a WDM expansion downsampling method, and frequency information of wavelet transform is combined with spatial information of convolution downsampling; the DWB and the MAB are sequentially arranged between the encoder part and the decoder part, the DWB decomposes features into four frequency components by using Haar discrete wavelet transform (DWT), the low-frequency features are processed by a small AOD network to extract the features, and the high-frequency features are refined by using an expansion residual block; and then spatial information is reconstructed by applying inverse wavelet transform. According to the method, feature representation is improved, and the defogging performance of the network is effectively enhanced.

The invention provides a fault detection method based on multi-dimensional anomaly detection and dual-path fine-grained positioning. The fault detection method comprises the following steps: carrying out anomaly detection based on a multi-dimensional anomaly evaluation unit and fault diagnosis based on a dual-path fine-grained feature positioning structure on a photovoltaic system; wherein the multi-dimensional anomaly evaluation unit comprises an attention unit, a gating unit and an anomaly judgment unit, and the attention unit fuses discrete wavelet transform, inverse wavelet transform and a position-space attention mechanism. According to the fault detection method disclosed by the invention, the interference of complex illumination and background noise is reduced, the robustness effect of the model in a dynamic environment is enhanced, and in addition, a breakthrough is realized in fine-grained fault positioning and classification precision; dispersive abnormal areas are accurately positioned, and high-similarity fault modes are distinguished.

The invention relates to the field of personnel re-identification, and provides an infrared-visible light personnel re-identification method and system based on wavelet representation, a wave adaptive encoder is designed, characteristics are decomposed into different frequency sub-bands through discrete wavelet transform, high-frequency and low-frequency information is processed through a high-low wave sensing unit respectively, and the personnel re-identification accuracy is improved. Adaptively adjusting the importance of different frequency band characteristics by using a wave frequency domain attention mechanism; a frequency enhancement module is introduced, modeling is carried out on global spectral characteristics by using Fourier transform, deep fusion of frequency domain and spatial domain information is realized through Einstein matrix multiplication, and the robustness of the model to different illumination conditions and imaging quality is enhanced; according to the method, similarity distribution clustering loss is enhanced, cross-modal features of the same identity are drawn close, features of different identities in the same modal are also explicitly pushed away, a more robust feature space is constructed, the problem of infrared image homogeneity is effectively solved, and the discrimination capability of the feature space is enhanced.

The invention provides a low-illumination image enhancement method based on brightness priori guidance and multi-level space-frequency domain feature fusion, and the method comprises the steps: brightness priori guidance: in a multi-level processing structure, dynamically generating brightness priori features from an input low-illumination image level by level, and guiding the global brightness distribution optimization; space-frequency domain cooperative processing: in a space domain, local brightness consistency is enhanced in combination with the brightness priori features; in the frequency domain, the amplitude is adjusted through Fourier transform to compensate the global brightness; adaptively fusing the spatial domain and frequency domain features, and outputting optimized features; dual-branch detail recovery: noise is suppressed and spatial domain texture details are recovered through a gating convolution mechanism; high-frequency components are extracted through discrete wavelet transform, and edge details are enhanced in combination with channel attention; fusing the detail features of the spatial domain and the frequency domain; and image reconstruction: reconstructing a normal illumination image based on the fusion features.

The invention relates to the cross technical field of artificial intelligence and medical image processing, in particular to a medical image segmentation system and method based on a wavelet bridge diffusion model and an efficient conditional random field. A WBDM-ECRF framework is constructed and comprises a discrete wavelet transform module, a BDM-T module, a BDM-S module and an ECRF module; decomposing the image through discrete wavelet transform, extracting a low-frequency sub-band, and enhancing the contrast ratio of a focus and normal tissues; the BDM-T takes U-Net as a backbone, integrates a FlashAttention mechanism, and optimizes a variance formula to realize efficient training; the BDM-S adopts a leapfrog sampling strategy, so that the reasoning time is greatly shortened; the ECRF introduces a multivariate potential function of a structural similarity index and smooth operation through edge expansion, and accurately optimizes edge segmentation. According to the method, the dependence of marked data is reduced, the training and reasoning efficiency is improved, the problem of fuzzy edge segmentation is solved, the Dice coefficient and intersection-union ratio performance on the ISIC data set is excellent, and reliable quantitative support is provided for disease diagnosis and treatment.

The invention discloses a generative image steganography method and device based on Stable Diffusion and discrete wavelet transform, and the method comprises the following steps: S1, sampling a potential representation of a to-be-generated image through employing a Stable Diffusion diffusion model, embedding secret information into a potential space of the diffusion model, processing the potential representation through employing discrete wavelet transform to carry out frequency domain information modulation, and carrying out the frequency domain information modulation; obtaining the potential representation after the secret information is embedded; s2, inputting the potential representation embedded with the secret information into a diffusion model, and generating a visual natural steganographic image through a diffusion inversion process; s3, inputting the received steganographic image into a diffusion model, and decrypting the received steganographic image without original model weight modification and empty prompt conditions to obtain embedded secret information; according to the method, the frequency domain embedding technology is combined with the potential diffusion model, so that high-fidelity, high-capacity and high-robustness image steganography is realized.

The invention discloses a priori knowledge and frequency domain perception-based cell micronucleus detection method, and relates to the technical field of computers. The method comprises the following steps: extracting cell micronucleus features of a to-be-detected cell image in a multi-level and multi-angle manner based on priori knowledge of cell micronucleus; the cell micronucleus characteristics comprise the volume of the cell micronucleus, the form of the cell micronucleus and the distribution area of the cell micronucleus; performing continuous channel compression on the cell micronucleus features, and extracting frequency domain information of the cell micronucleus features after channel compression through discrete wavelet transform; the frequency domain information comprises high-frequency information and low-frequency information; convolution processing is carried out on the high-frequency information and the low-frequency information, and convolution processing results are fused to obtain frequency domain sensing fusion features; analyzing the frequency domain perception fusion feature to obtain a plurality of bounding boxes in the to-be-detected image; the bounding box is the predicted position of the cell micronucleus. The method can improve the detection precision of the cell micronucleus.

The invention discloses a time prediction method and system based on double-domain feature fusion. The method and system adapt to long and short term time series prediction requirements of multiple scenes such as weather forecast, energy scheduling, traffic flow and financial exchange rate. The method comprises the steps that a multi-field data set is obtained and preprocessed, and instance normalization is carried out; performing double-domain multi-scale characteristic decomposition by adopting down-sampling and discrete wavelet transform to obtain continuous trend and high-frequency mutation details; a local unit is obtained through patch cutting and embedding, local time sequence association is mined through depth separable convolution, cross-patch global interaction is achieved in combination with a multi-layer perceptron, and local-to-global progressive fusion is completed; and constructing bidirectional attention flow enhanced cross-domain and cross-scale collaboration, and combining with standardized data training to obtain a prediction model. According to the invention, the method can improve the depiction capability of non-stable and non-linear complex time sequence data containing abrupt change and multi-period superposition, gives consideration to the adaptability of long and short term prediction, remarkably improves the accuracy of multi-field time sequence prediction, and promotes the application of the prediction technology in multiple scenes.

The invention discloses a surface roughness and size precision common prediction method based on a relative position matrix, and mainly relates to the field of roughness detection. Comprising the steps that in the turning machining process, vibration signals are collected in real time through multi-channel acceleration sensors arranged at the positions of a main shaft, a tool and a workpiece clamp, and a roughness measuring instrument and a micrometer are synchronously used for measuring the surface roughness value and size data of a machined workpiece; normalization processing is carried out on the collected vibration signals, the vibration signals are converted into two-dimensional images to be expressed based on a relative position matrix algorithm, deep features are extracted through two-dimensional discrete wavelet transform, and the spatial expression ability of the signals is enhanced. The method has the beneficial effects that the vibration signal is converted into the two-dimensional RPM graph through the RPM, and the 2D-DWT is utilized to perform decomposition reconstruction and image fusion, so that the deep sharing features in the vibration signal are effectively extracted, and the method plays a key role in improving the prediction precision of the surface roughness and the size precision.

The invention provides a multivariate time sequence prediction method and system and a medium, an input time sequence is adaptively decomposed into an approximate coefficient sequence and a detail coefficient sequence by adopting an adaptive discrete wavelet transform and inverse transform mode, and compared with the previous wavelet transform depending on a predefined wavelet basis function, the multivariate time sequence prediction method and system have the advantages that the time sequence prediction efficiency is improved; the mode is more flexible, and filtering kernels of discrete wavelet transform and inverse transform can be updated in a data driving mode in the training process according to the characteristics of the time sequence, so that the filtering kernels are more suitable for the current sequence. Wavelet transformation realized by carrying out convolution operation on a time dimension lacks extraction and modeling of correlation between global information and different variables in a multivariate time sequence; therefore, a coefficient mixing module and a wavelet domain attention channel enhancement module are provided to carry out supplementary modeling on an approximation coefficient and detail coefficient sequence obtained after wavelet transformation. The method can meet the prediction requirements of the time series in actual production such as power load.

The invention relates to an atomic clock frequency anomaly detection method, and belongs to the technical field of deep learning and the field of atomic clock anomaly detection. The method comprises the following steps: acquiring a frequency sequence set with a label generated by an atomic clock in a normal frequency sequence injection known abnormal mode; performing supervised training on the anomaly detection model by using the frequency sequence set; in the anomaly detection model, performing feature fusion, nonlinear transformation and classification on a first path of features extracted by performing multilayer decomposition on an input frequency sequence by adopting discrete wavelet transform and a second path of features extracted by adopting a CNN-Transformer network on the input frequency sequence, and then detecting an abnormal frequency sequence; and performing frequency anomaly detection on the atomic clock frequency sequence acquired in real time by using the trained anomaly detection model. According to the invention, high-precision detection of frequency anomalies with different characteristics is realized, the detection range is expanded, and the detection precision is improved.