50 results about "Wave transformation" patented technology

This invention discloses a planetary gearbox fault diagnosis method and related equipment based on empirical wavelet transform. The method includes: acquiring the structural parameters of each component in the planetary gearbox to be tested, determining the fault characteristic frequency and meshing frequency of each component based on the structural parameters, calculating the number of initial frequency bands, and dividing the signal spectrum of the planetary gearbox to be tested according to the number of initial frequency bands to obtain initial frequency bands; calculating the comprehensive fault index of the initial frequency bands, and determining whether the initial frequency bands are fault frequency bands. If they are fault frequency bands, reconstructing the boundary of the initial frequency bands according to the comprehensive fault index and a preset boundary merging rule to obtain target frequency bands, and performing EWT transform on the signal spectrum according to the target frequency bands to obtain target EWT component signals; calculating the IMF kurtosis value of each target EWT component signal, selecting sensitive EWT component signals from the target EWT component signals according to the IMF kurtosis value, performing envelope spectrum analysis on the sensitive EWT component signals, and obtaining a diagnostic conclusion.

This invention discloses a visual inspection method and apparatus for glass production quality, relating to the field of glass quality inspection technology. The method includes the following steps: constructing a polarization excitation detection field; acquiring and registering a sequence of original polarization images of the glass based on the polarization excitation detection field to obtain a polarization image sequence; performing image fusion on the polarization image sequence using an image fusion algorithm based on non-subsampled contour wave transform, sparse representation, and guided filters to obtain a defect-enhanced image; inputting the defect-enhanced image into a lightweight glass defect recognition model based on improved self-attention, outputting a glass defect category, acquiring glass defect quality impact data, and classifying glass quality grades based on the glass defect category and the glass defect quality impact data, thereby achieving visual inspection of glass production quality.

This invention relates to the field of image restoration technology, specifically disclosing an image restoration method and device based on discrete wavelet transform. An image restoration network, called Multi-Wavelet Attention Memory Neural Network (MWA-MNN), is designed, consisting of three sub-networks. These sub-networks include Discrete Wavelet Attention Blocks (DWABs) and Coordinate Channel Attention Blocks (CCABs) to extract spatial and channel feature information, effectively separating degraded image layers such as rain streaks from the image to obtain a clean image. The invention leverages the expressive power of wavelet transform for high-frequency information, thereby improving image restoration performance. The sub-networks also include Supervised Selective Kernel Blocks (SSKBs) for filtering effectively extracted features in each sub-network. This invention also proposes a memristor-based cross-array MWA-MNN circuit implementation scheme. Memristors are a novel type of non-volatile memory with faster read speeds and lower power consumption. Experimental results show that this method has good performance and versatility in various image restoration tasks.

The present application relates to the technical field of biomedical signal processing and artificial intelligence, and provides a heart rate detection method based on a learnable wavelet transform and feature enhancement, comprising: acquiring radar phase signals collected by a millimeter wave radar and obtained after preprocessing; and performing a learnable wavelet transform, extracting time-frequency features, and through multi-scale decomposition on a physiological related frequency band, performing dynamic weighted fusion based on energy of each frequency band, and outputting a multi-scale fusion feature tensor; after projecting and fusing the multi-scale fusion feature tensor and the original radar phase signals, inputting the same into an LSTM-Transformer hybrid time series modeling network, and outputting a second-by-second heart rate estimation value sequence; in a training stage, a hybrid loss function is calculated by using a real heart rate label and the heart rate estimation value sequence, and an end-to-end optimization is performed on the network. Through the method, the accuracy and robustness of non-contact heart rate monitoring are improved.

This application discloses a diabetes detection system based on multi-sensor pulse characteristics, relating to the field of medical signal processing. The system includes: a pulse wave data acquisition module that uses a pressure sensor and a photoelectric sensor to acquire the user's wrist pressure pulse wave and fingertip photoplethysmography (PPG) pulse wave; a preprocessing module to obtain usable single-cycle pulse wave data; a signal fusion module that converts the pressure pulse wave signal and PPG pulse wave signal into two-dimensional image signals through continuous wavelet transform to retain time-frequency domain information, and processes the two two-dimensional images using non-subsampled shear wave transform and a pulse-coupled neural network; and a feature extraction and calculation module that extracts pulse wave image features through a ResNet network and finally classifies them using a random forest. This detection system improves the fusion effect without increasing computational load, achieving non-invasive diabetes detection using multiple sensors and improving detection accuracy.

This invention proposes a collaborative reconstruction method for multimodal images from unmanned aerial vehicles (UAVs) for all-weather perception. First, it uses an Adaptive Degradation Perceptual Block (ADPB) to perform spectral analysis on degraded visible light features to perceive the degradation type, and utilizes a learnable frequency mask to decouple features, providing a clear, high-fidelity structural prior for subsequent modal interactions. Next, it leverages the multi-scale decomposition capability of wavelet transform to decouple multi-frequency sub-band signals, and learns local contraction mappings within each sub-band through deep convolution, effectively preserving structural integrity while suppressing blur and noise. Finally, addressing the challenge of simultaneous degradation and cross-modal information fusion in multimodal images, it generates dynamic modulation factors in spatial and channel dimensions to guide adaptive complementary interaction and deep fusion of cross-modal features, strengthening the representation capabilities of global and local features. This significantly improves the reconstruction quality and multi-task generalization ability of the unified multi-task recovery model in complex scenarios.

The application discloses a new energy line protection method based on voltage reverse wave polarity difference, which firstly performs phase-mode transformation on collected three-phase voltage signals and current signals, extracts line-mode fault voltage and fault current, and then calculates voltage traveling waves propagating in the reverse direction at the protection installation position. On this basis, the wave head signals of voltage reverse waves on both sides of the line are calibrated by using wavelet transformation mode maximum, the polarity characteristics of the first voltage reverse wave with reverse polarity on both ends of the line are extracted, and the internal and external faults are distinguished according to the polarity relationship of the voltage reverse waves on both ends of the line. The scheme considers that when the near-end fault occurs, the first several traveling waves are difficult to capture or the wave head aliasing problem exists due to the influence of the sampling rate, and then the new energy line protection scheme based on the polarity difference of voltage reverse waves on both ends of the line is proposed. The scheme has simple principle and certain fault tolerance.

This invention discloses an activity pattern recognition method and system based on wavelet transform and CNN / Transformer. First, wavelet transform is used to perform multi-scale time-frequency decomposition on the original signal to enhance the perception of subtle local features and high-frequency components. Then, convolution operations are used to further extract and fuse effective features in the time-frequency domain. Finally, leveraging the self-attention mechanism of Transformer, global dependencies are captured in the enhanced feature sequence, thereby achieving efficient collaborative modeling of local information and global context. This invention overcomes the insensitivity of Transformer models to local details and high-frequency information, effectively avoiding the neglect or smoothing of crucial subtle action patterns and transiently changing high-frequency signals during recognition, significantly improving the model's recognition accuracy and enhancing the reliability and interpretability of the entire system in practical applications.

The application discloses a single-end protection method based on wavelet transform and pulse neural network, which comprises the following steps: collecting transient voltage and current signals of a single end of a direct-current distribution network in real time; performing multi-resolution decomposition on the collected transient waveforms by using wavelet transform to obtain detail coefficients and high-frequency energy spectrum under different scales; taking an input layer of the pulse neural network as an encoding layer, and converting the extracted continuous analog wavelet coefficients into discrete binary pulses; and finally inputting the generated pulse sequence into the pulse neural network to complete accurate classification of fault types and output corresponding relay protection tripping instructions. The application proposes a single-end protection method based on wavelet transform and pulse neural network for problems such as uncertain energy flow direction, complex mapping relationship between fault characteristics and fault types and the like caused by large access of distributed energy, and avoids strong dependence on accurate fault mathematical models and fixed protection thresholds.

The present invention relates to the technical field of person re-identification. Disclosed are a cross-modal person re-identification method and system based on wavelet transform. The method comprises: constructing a two-stream ResNet-50 backbone network; by means of an ICB module, fusing shallow features of the two-stream ResNet-50 backbone network, so as to obtain fused features; by means of a WEB branch module, performing feature extraction on the fused features, so as to obtain modality-shared features of high-frequency and low-frequency portions; and by means of a dual-branch center-guided loss, optimizing the modality-shared features output by the WEB branch module and the fused features output by the ICB module, so as to obtain a final identification result. In the present invention, an information compensation module is combined with wavelet transform to aggregate shallow network features in different phases, and thus valuable information lost in network feature extraction is compensated for, thereby improving the quality of a final feature vector; and a wavelet enhancement module is provided, and a dual-branch center-guided loss is used to guide a network to mine modality-invariant information in wavelet subgraphs, thereby improving the model performance.

This invention discloses a method for obstacle detection in the drivable area of ​​an all-weather unmanned emergency rescue vehicle, constructing a curvature-enhanced frequency domain decoupling network. First, a multi-scale curvature enhancement branch is proposed, combining the principal curvature representation with the HOG descriptor. Second, a frequency domain decoupling module is proposed, abandoning super-resolution methods that easily amplify noise. High and low frequencies are explicitly separated using wavelet transform, and subtle details are extracted through high-frequency differential residuals. Accurate cross-domain fusion is achieved by combining a differential-guided nonlocal feature correction mechanism, effectively suppressing noise interference in low-light and low signal-to-noise ratio environments. A feature modulation unit is designed to perform adaptive modulation and multi-scale fusion before the feature input detection head, introducing KANConv and DWConv differentially for targets of different scales. This invention simultaneously addresses both refined texture-edge feature extraction and noise suppression in harsh environments.

This invention provides a lightweight intrusion detection method, system, and storage medium based on frequency domain gating and dynamic head attention distillation. The method includes: Step 1: Constructing a teacher model, which includes a frequency domain wavelet analysis module and a dynamic head interaction attention module. The frequency domain wavelet analysis module performs multi-scale decomposition and information enhancement of network traffic features in the frequency domain through a two-level Haar wavelet transform and adaptive gating mechanism. The dynamic head interaction attention module treats the feature dimension as a token sequence and models the global dependencies between features by introducing position-aware embedding and multi-head self-attention mechanisms; Step 2: Designing a dynamic entropy adaptive topology distillation strategy, enabling the lightweight student model to inherit the discriminative features and sample relationship structure of the teacher model while compressing the feature space; Step 3: Transferring the knowledge of the teacher model to the lightweight student network. The beneficial effect of this invention is that it achieves more efficient and more complete knowledge transfer.

The application discloses an underground disease intelligent identification method based on multi-physical field feature fusion. The method uses multi-source information such as ground penetrating radar B-scan data, dielectric constant distribution and conductivity distribution, extracts time domain, frequency domain and medium parameter features through a learnable wavelet transform and a multi-network structure, realizes cross-domain feature alignment and importance weighting in an adaptive fusion module, and forms a comprehensive underground target representation. Via a detection network, the positioning and classification of diseases such as cavities, voids, loose and leakage are output. The model is trained by using a joint loss containing classification, position regression, frequency domain consistency and electromagnetic propagation constraints. The method can effectively improve the accuracy and physical credibility of underground disease identification.

This invention discloses a method and apparatus for extracting roll eccentricity signals from a hot rolling mill, relating to the field of steel production control technology. The method includes: real-time data acquisition using sensors on the hot rolling mill stand to obtain the original signals of the rolling process and roll parameters; two-stage preprocessing of the original signal based on moving average filtering and least squares methods to obtain a preprocessed signal; time-frequency and time-domain joint feature fusion based on short-time Fourier transform and principal component analysis to obtain a fused time-domain feature vector; adaptive spectral decomposition using empirical wavelet transform based on the fused time-domain feature vector to obtain empirical mode components; and signal reconstruction based on roll parameters and empirical mode components to obtain a synthesized eccentricity signal. This invention provides a roll eccentricity signal extraction method with strong adaptability, strong anti-interference ability, and high extraction accuracy.

An image processing method includes acquiring an image in which a cell structure having a stained vascular network structure is imaged, applying a wavelet transform to the image such that a contour image in which contours of the vascular network structure are extracted is generated, and repeatedly excluding object pixels from object boundaries recognized in the contour image such that a skeleton image in which skeletons having a line width of a predetermined number of pixels are extracted is generated.

The application discloses a kind of distribution box switch cabinet fault detection method and system, is specifically related to switch cabinet fault detection technical field;The application realizes the accurate purification of switch cabinet fault signal, and the multiple signals collected are adaptively decomposed into intrinsic mode components by empirical wavelet transform algorithm, combined with multi-scale entropy and multi-fractal feature parameter to construct classification feature vector, relying on labeled data set and classification logic to quantize identification information and noise dominant component and eliminate noise, effectively separate the complex interference and weak fault characteristics in the field, solve the problem of traditional signal processing modal aliasing, effective component screening is not accurate, greatly improve the purity of fault characteristic signal, lay a foundation for subsequent accurate diagnosis.

The present application relates to the technical field of data processing, and especially relates to a brake pad operation state real-time monitoring and early warning method based on data analysis, which comprises the following steps: obtaining a brake pad braking vibration signal and segmenting to obtain multiple signal sequences; calculating an instability index of each signal sequence; determining a decomposition layer number of wavelet transformation and a position of a target base function in a preset base function library based on the instability index, and selecting the target base function from the preset base function library according to the position; performing wavelet transformation on each signal sequence based on the decomposition layer number and the target base function to obtain a wavelet transformation result of each signal sequence; and calculating an abnormal value based on the wavelet transformation result of each signal sequence, and responding to an operation state abnormality when the abnormal value is greater than a set threshold value. The present application determines a decomposition layer number and a target base function suitable for different signal sequences based on an instability index, realizes adaptive analysis of the signal sequences, and improves monitoring accuracy.

This patent relates to an anomaly detection method for industrial control systems based on time-frequency transform and deep learning. It aims to effectively identify abnormal behavior in flow data by combining wavelet transform and an autoencoder. First, wavelet transform is used as a time-frequency transformation tool to perform time-frequency domain analysis on the wavelet coefficients and extract statistical features. Then, a deep learning autoencoder model is designed and trained by minimizing reconstruction error to learn the time-frequency features of normal data. Subsequently, the trained autoencoder is used to predict an anomaly-laden test set, and anomalies are detected by setting thresholds and reconstruction errors. To verify the robustness of the method, three attacks were introduced in the experiment. The performance of the anomaly detection method was further verified by comparing the coefficients extracted from the wavelet transform of the attacked normal data with those of the original normal data. Experimental results show that this method has good robustness while ensuring the accuracy of anomaly detection, demonstrating its potential application prospects in the field of flow anomaly detection in industrial control systems.

The application relates to the fields of software testing and intelligent computing, and relates to an image blind watermark embedding and complete extraction method based on strong robustness multi-domain transformation. The method comprises the following steps: extracting a YUV channel of an original image; performing two-stage Haar wavelet transformation on the YUV channel; using DCT transformation on a low-frequency information matrix; using singular value decomposition to extract characteristic values corresponding to a frequency domain matrix of the block; using a variable watermark embedding algorithm to embed watermarks in the block frequency domain matrix; performing inverse singular value decomposition, inverse discrete cosine transformation and inverse wavelet transformation on the block frequency domain matrix to obtain a watermark-carrying image; using a watermark extraction transformation algorithm to perform inverse operation on the embedded watermark of the image, and extracting each bit value of the watermark embedded in the image to be extracted; correcting the bit value distribution deviation of the watermark through a watermark value clustering algorithm, and outputting the final watermark value. The application can more secretly embed image watermarks and can recover the watermark from various image attacks, thereby improving the robustness of the image watermark.

The application discloses a phonon transport property calculation model, system and method based on wave-particle duality, which comprises a normal mode analysis module, a phonon wavelet transform module and a thermal conductivity calculation module, and unifies the particle nature and wave nature of phonons in a thermal conductivity prediction framework, wherein the normal mode analysis module is used for projecting atomic scale dynamic evolution information to a phonon mode space to obtain mode-resolved space-time evolution data; the phonon wavelet transform module is used for wavelet transforming the mode-resolved space-time evolution data to quantitatively characterize the coherent time and coherent length of phonons; and the thermal conductivity calculation module is used for simultaneously calculating the contribution of phonon particle nature lifetime and wave nature coherent time to thermal conductivity based on the mode-resolved space-time evolution data, the coherent time and the coherent length, thereby providing a phonon transport image closer to physical reality and greatly improving the research and development efficiency.

This invention discloses a robust zero-watermarking method and system for medical images based on local minimum adaptive two-dimensional empirical wavelet transform (2DEWT), polar harmonic transform (PHT), and Gabor transform. The method includes the following steps: Step 1, copyright watermark encryption process; Step 2, copyright watermark embedding process; Step 3, copyright watermark extraction process; Step 4, copyright watermark decryption process; Step 5, copyright system usage process. This invention uses 2DEWT in the extraction of medical image features and, based on the characteristics of medical images, designs an adaptive improvement strategy that directly uses the local minimum points of the frequency domain energy curve as sub-band boundaries. The local minimum points are essentially the boundary points of different feature components in the frequency domain of medical images: the boundaries between low-frequency global structures and high-frequency lesion details, and the boundaries of lesion textures in different directions (such as blood vessel orientation and lesion edge angles) all exist in the form of local minimum points. Two-dimensional empirical wavelet transform is performed on medical images to extract low-frequency subbands. Gabor transform is applied to the low-frequency subbands, and PHT is applied to the high-frequency subbands. Image features are obtained through feature fusion, and the copyright watermark is embedded by obtaining the key through XOR with the copyright watermark.

This invention discloses an image processing model, a model training method, an image processing method, and its applications. The model includes a feature extraction network module, a wavelet transform module, a high-frequency information directional sub-band attention fusion module, a feature enhancement fusion module, a multi-scale fusion neck network module, and a detection head module. This invention obtains high-frequency sub-bands in multiple directions through discrete wavelet transform and uses the high-frequency information directional sub-band attention fusion module to weight the high-frequency features in each direction, thereby enhancing the representation ability of directional high-frequency details in the image. Simultaneously, the feature enhancement fusion module fuses the backbone features and high-frequency information features at corresponding scales, and combines this with the multi-scale fusion neck network module for cross-scale feature fusion, thereby improving the multi-scale feature synergistic enhancement capability and cross-scale fusion processing capability.

This invention proposes a method for detecting underground defects based on multi-frequency features of ground-penetrating radar (GPR). This method utilizes a neural network denoising model to perform multi-scale and residual adaptive denoising on GPR signals, effectively suppressing random and structural noise while preserving weak scattering features such as cavitation, loosening, voiding, and leakage. Low-frequency structural information and high-frequency scattering texture are obtained through two-dimensional wavelet transform or wavelet packet decomposition. Multiple high-frequency subbands are then spliced ​​and learnedable enhancements are applied to obtain a high-frequency representation highlighting the defect features. Combining low-frequency features and convolutional multi-scale features, after spatial scale alignment, the data is input into a multimodal fusion module. Deep feature reconstruction is achieved through learnable weights, multi-head self-attention, and convolutional integration. Finally, the classification and regression branches output the defect category and location parameters. This invention can achieve high-precision identification and stable detection of cavitation, loosening, voiding, and leakage defects under complex noise conditions.

The application discloses a power transmission line abnormal discharge fault point positioning method, comprising the following steps: S1, installing a high-frequency current transformer at a key node to collect a transient current signal during abnormal discharge at a frequency not lower than 1MHz and to carry out denoising and standardization preprocessing; S2, parallelly applying wavelet transform and Fourier transform, the wavelet transform extracts local detail features of the signal in the time-frequency domain, the Fourier transform obtains global frequency spectrum features of the signal, and the two are fused into a high-dimensional composite feature vector; S3, inputting the obtained feature vector into a pre-trained BP neural network to realize intelligent mapping from the features to the fault mode; S4, adopting an electromagnetic time reversal method, intercepting a fault initial traveling wave signal, reversing on a time axis, replaying in an accurate electromagnetic model of the line, calculating and scanning a spatial distribution function of electromagnetic energy along the line, and obtaining an accurate fault point; the application has the advantages of intelligent identification, high positioning precision and strong anti-interference performance.

This invention proposes a method for identifying fault types in 10kV distribution network lines. First, the amplitude and phase angle features of the initial fault stage are extracted from voltage and current signals to form a preliminary trajectory. Then, wavelet transform decomposition is used to obtain millisecond-scale energy distribution to capture transient characteristic differences, and key components are selected for preliminary classification. Next, fine-grained features from multiple time windows are integrated, and support vector machines are used to calculate the similarity with sample trajectories to obtain probability estimates. Finally, transient trends and steady-state features are combined for final determination, thereby achieving accurate fault type identification. Finally, by clustering and summarizing stable feature patterns and feeding back to optimize the wavelet scale and model, a continuously adaptive identification framework is constructed, significantly improving the accuracy and reliability of fault classification and identification.

The application relates to the technical field of laser cutting and discloses a steel plate laser cutting machine fault detection method and system, the method comprising the following steps: collecting a driving power current signal in real time and filtering out a direct current component to obtain an alternating current sequence; separating a multi-scale subband component containing Gaussian noise by using wavelet transform; performing empirical mode decomposition on the subband component to extract an intrinsic mode function component set; calculating the kurtosis value, kurtosis change rate and phase coupling index based on bispectrum analysis of the component to construct a high-order statistical feature vector; classifying the feature vector by using a support vector machine to output an operation state category label; combining the label risk value and the kurtosis evolution trend slope to generate a fault early warning grade value; and verifying early fault states by using a historical state sliding window sequence and a fault confirmation confidence degree, so that the method can solve the problem of low precision of nonlinear fault feature extraction in a complex noise interference background existing in the prior art.

The application provides a transformer protection method and device based on a flow rate signal wavelet transform, and the method comprises the following steps: collecting a flow rate signal of insulating oil at a connecting pipe of a transformer; detecting whether the flow rate at a current time in the flow rate signal is greater than or equal to a preset flow rate threshold value; if the flow rate at the current time is greater than or equal to the flow rate threshold value, performing wavelet transform on the flow rate signal, and detecting whether the transformer is faulty according to the wavelet transform value of the flow rate signal; and when the transformer fault is detected, controlling the transformer to be powered off. According to the flow rate signal of the insulating oil at the connecting pipe of the transformer, whether the transformer is faulty is detected, the detection process is not interfered by signal transmission in a power system of the transformer, meanwhile, harmonic interference is not caused to the power system of the transformer, and the power system operation is not affected.