31 results about "Complex wavelet transform" patented technology

The present disclosure relates to digital watermarking systems that may use visual masks to optimize watermark embedding in digital imagery. A visual mask provides guidance for adjusting digital watermark signal strength, enabling improved trade-offs between watermark imperceptibility and robustness. Multiple embodiments generate visual masks including: (1) a Perceptual Modeling Candidate approach using contrast masking and texture classification based on standard deviation mapping; (2) a wavelet-based approach using Dual-Tree Complex Wavelet Transform for translation-invariant frequency analysis; (3) artificial intelligence approaches employing convolutional neural networks trained to optimize embedding strength while minimizing perceptual distance metrics such as LPIPS; and (4) LPIPS threshold masking that determines optimal embedding strengths by testing multiple candidate strengths. Visual masks enable content-adaptive digital watermarking that places stronger signals in textured regions while maintaining imperceptibility in flat regions, improving visibility-robustness performance compared to uniform embedding approaches.

The invention belongs to the technical field of computer vision and computer graphics, and discloses a multi-view three-dimensional reconstruction method and device based on cross-domain feature fusion and a medium, and the method comprises the steps: carrying out the cross-feature-domain coding of an initial feature token generated by a multi-view image, according to the coding, characteristics are decomposed into low-frequency components and high-frequency components through dual-tree complex wavelet transform, amplitude modulation is carried out on the high-frequency components to enhance details, and space-frequency fusion characteristics are generated; variance embedding weighted combination is carried out on the multi-view space-frequency fusion features, the combination generates a weight by calculating the variance of each feature token, adaptive weighted clustering is carried out based on the weight, and fused multi-view features are generated; and performing three-source attention decoding on the fused features, and performing cross-domain attention calculation and up-sampling by taking static embedding as query, taking the space-frequency fused features as keys and taking the fused multi-view features as values, thereby finally generating a three-dimensional voxel reconstruction result of the target object.

The invention provides a millimeter-wave radar personnel perception method based on a time-frequency domain and a deep CNN, and relates to the technical field of radar signal processing, and the method comprises the steps: carrying out the preprocessing, spectrogram conversion and enhancement of a millimeter-wave radar echo signal; spatial features are extracted by using depth separable convolution of a residual structure, time-frequency features are extracted by combining dual-tree complex wavelet transform and attention-enhanced cavity convolution, motion features are extracted through optical flow estimation and three-dimensional convolution, and the three features are adaptively fused; the method comprises the following steps: constructing a dynamic spatio-temporal reasoning network, obtaining key spatio-temporal features by using a recurrent neural tensor network, a non-local neural network and deformable convolution, extracting multi-scale features and performing adversarial feature alignment, and completing personnel target classification through a dynamic routing mechanism based on a capsule network. According to the method, clutter interference can be effectively suppressed, space, time frequency and motion information is fully utilized, accurate perception of personnel targets is realized, and the classification accuracy is improved.

The invention provides a millimeter-wave radar personnel perception method based on a time-frequency domain and a deep CNN, and relates to the technical field of radar signal processing, and the method comprises the steps: carrying out the preprocessing, spectrogram conversion and enhancement of a millimeter-wave radar echo signal; spatial features are extracted by using depth separable convolution of a residual structure, time-frequency features are extracted by combining dual-tree complex wavelet transform and attention-enhanced cavity convolution, motion features are extracted through optical flow estimation and three-dimensional convolution, and the three features are adaptively fused; the method comprises the following steps: constructing a dynamic spatio-temporal reasoning network, obtaining key spatio-temporal features by using a recurrent neural tensor network, a non-local neural network and deformable convolution, extracting multi-scale features and performing adversarial feature alignment, and completing personnel target classification through a dynamic routing mechanism based on a capsule network. According to the method, clutter interference can be effectively suppressed, space, time frequency and motion information is fully utilized, accurate perception of personnel targets is realized, and the classification accuracy is improved.

The application provides a millimeter wave radar personnel sensing method based on a time-frequency domain and a deep CNN, relates to the technical field of radar signal processing, and comprises preprocessing, spectrum conversion and enhancement of a millimeter wave radar echo signal; spatial features are extracted by using a deep separable convolution with a residual structure, time-frequency features are extracted by combining a dual-tree complex wavelet transform and an attention-enhanced hollow convolution, and motion features are extracted by optical flow estimation and three-dimensional convolution, and the three are adaptively fused; a dynamic space-time reasoning network is constructed, key space-time features are obtained by using a recurrent neural tensor network, a non-local neural network and a deformable convolution, multi-scale features are extracted and aligned in an adversarial manner, and a dynamic routing mechanism based on a capsule network is used to complete personnel target classification. The application can effectively suppress clutter interference, fully utilize spatial, time-frequency and motion information, realize accurate sensing of personnel targets, and improve classification accuracy.

The application discloses a virtual viewpoint image quality evaluation method and system based on double-tree complex wavelet transform. The method comprises the following steps: using an algorithm to perform feature matching on a reference image; performing multi-scale double-tree complex wavelet transform on the matched reference image and a virtual viewpoint image for multiple times to obtain a first wavelet subband set corresponding to the matched reference image and a second wavelet subband set corresponding to the virtual viewpoint image; calculating gradient amplitude similarity between third-level wavelet subbands of the first wavelet subband set and the second wavelet subband set, and using the gradient amplitude similarity to calculate a texture distortion score of the virtual viewpoint image; calculating structural similarity between fifth-level wavelet subbands of the first wavelet subband set and the second wavelet subband set, and using the structural similarity to calculate a structure distortion score of the virtual viewpoint image; and calculating a quality score of the virtual viewpoint image according to the texture distortion score and the structure distortion score. The application can accurately evaluate the quality of the virtual viewpoint image.

The invention discloses a common transformer side single-phase intelligent electric meter electric quantity acquisition accuracy improving system, and relates to the technical field of electric variable measurement. According to the system, sampling rate adaptive switching is realized through the working condition modal identification module and the dynamic sampling control module, full-band power accurate metering is carried out by using complex wavelet transform, online calibration and optimization are carried out by means of the parameter self-learning module, the accuracy and adaptability of an electric meter in a complex power grid environment are comprehensively improved, and the accuracy and adaptability of the electric meter in the complex power grid environment are improved. The problems that the fixed sampling rate of a traditional intelligent electric meter is not matched with the dynamic working condition, the harmonic and inter-harmonic power metering is lacked, and the working condition self-adaptive capacity is lacked are effectively solved, the electric meter achieves self-adaptive balance of precision and efficiency in a complex power grid environment, full-band accurate metering of fundamental wave, harmonic wave and inter-harmonic power is completed, and the electric energy utilization rate is improved. And the metering accuracy and stability of the electric meter in long-term operation are ensured through a self-learning mechanism.

The application discloses a motor output state evaluation method and system, comprising collecting the stator current signal of a target motor, and performing zero-phase band-pass filtering pretreatment on the stator current signal of the target motor to obtain a pretreated current signal. The application firstly retains the phase characteristics of the original signal through zero-phase band-pass filtering pretreatment, then accurately extracts an instantaneous frequency signal through complex wavelet transform, constructs a frequency-adaptive comb notch filter set relying on the motor working specification, filters out the operation signal related to the normal operation of the motor in the instantaneous frequency signal, so as to capture the residual frequency fluctuation caused by the dynamic balance fault, and then obtains an enhanced imbalance confidence index by combining the instantaneous spectrum kurtosis and the fractal dimension calculation, thereby improving the accuracy and reliability of the dynamic balance fault evaluation, and more accurately identifying the early dynamic balance abnormality of the motor, providing a reliable basis for the preventive maintenance of the motor, and effectively guaranteeing the operation safety of the transmission system.

The application discloses an infrared image non-uniformity prediction and correction method based on wavelet transform, which comprises the following steps: adopting double-density dual-tree complex wavelet transform to perform multi-scale decomposition on an input infrared image; in a high-frequency subband, detecting a blind element position based on a local variance statistical method; in a low-frequency subband, constructing an autoregressive model, predicting non-uniformity of a background region, and performing accurate processing on a prediction result; combining the prediction results of the high-frequency subband and the low-frequency subband to generate a pre-correction coefficient; and adopting the pre-correction coefficient to perform correction processing on an original infrared image to obtain a corrected image. Through the combination of an adaptive fusion strategy, space-time consistency constraint and dynamic gain and bias updating technology, the problems of non-uniformity noise, balance between details and background, device drift compensation and consistency in a dynamic scene in infrared image processing are solved, and the accuracy and stability of infrared image processing are significantly improved.

This invention discloses a ghost wave suppression method for OBS data based on horizontally extended complex wavelet transform. The method transforms land and water detection data into the complex wavelet domain using horizontally extended dual-tree complex wavelet transform. Then, the envelope energy ratio and phase ratio of the land and water detection data are calculated in different frequency bands of the complex wavelet domain to obtain the land detection amplitude and phase correction operators for each frequency band. These operators are applied to the land detection data to obtain calibrated land detection data. Subsequently, the land and water detection data are subtracted to obtain downlink wave data. Simultaneously, the above data undergoes an inverse complex wavelet transform to obtain a time-domain ghost wave model. Finally, the original land detection data and the ghost wave model are subtracted using least-squares matched subtraction to obtain the final uplink wave data. This invention, through fine calibration in the horizontally extended complex wavelet domain, can more effectively eliminate amplitude and phase differences in land and water detection data, thereby achieving higher-precision uplink and downlink wave field separation and ghost wave suppression.

The application discloses a system for improving the accuracy of power collection of a single-phase intelligent electric meter on the public power supply side, and relates to the technical field of electric variable measurement. The system realizes adaptive switching of a sampling rate through a working condition mode identification module and a dynamic sampling control module, performs accurate power measurement of a full frequency band by using complex wavelet transform, and is online calibrated and optimized with the help of a parameter self-learning module, thereby comprehensively improving the accuracy and adaptability of the electric meter in a complex power grid environment, effectively solving the problems of mismatching of a fixed sampling rate of a traditional intelligent electric meter with dynamic working conditions, missing of harmonic and interharmonic power measurement, and lack of working condition adaptability, enabling the electric meter to realize adaptive balance of precision and efficiency in a complex power grid environment, completing full-band accurate measurement of fundamental wave, harmonic and interharmonic power, and guaranteeing the measurement accuracy and stability of the electric meter in long-term operation through a self-learning mechanism.

The application belongs to the technical field of digital watermark embedding and relates to an adaptive digital watermark embedding method, device and readable storage medium. An image is subjected to double-tree complex wavelet transform decomposition to obtain a low-frequency approximation subband and a high-frequency complex coefficient subband. The subband variance, subband entropy and gradient energy of each high-frequency complex coefficient subband are calculated to obtain the joint score of each high-frequency complex coefficient subband. The amplitude graph of the high-frequency complex coefficient subband with the highest joint score is subjected to non-overlapping block division to obtain a plurality of candidate blocks. The block variance, block gradient intensity and gradient direction consistency of each candidate block are calculated to obtain the block score of each candidate block. The embedding block set is screened based on the block score, the embedding strength of each embedding block is calculated, the amplitude of all pixel positions in each embedding block is subjected to multiplicative embedding, the complex coefficient of the target embedding subband corresponding to each embedding block is reconstructed, and the watermark image is reconstructed based on the reconstructed target embedding subband, the low-frequency approximation subband and the high-frequency complex coefficient subband.

The application belongs to the technical field of computer vision and computer graphics, and discloses a multi-view three-dimensional reconstruction method based on cross-domain feature fusion, equipment and medium, the method comprises the following steps: cross-feature domain coding is carried out on the initial feature token generated by multi-angle images, the coding decomposes the feature into low-frequency and high-frequency components through dual-tree complex wavelet transform, and the amplitude modulation is carried out on the high-frequency component to enhance the details, and the space-frequency fusion feature is generated; the space-frequency fusion features of the multi-angle are weighted and merged through variance embedding, the weight is generated by calculating the variance of each feature token, and adaptive weighted clustering is carried out based on the weight to generate the fused multi-angle feature; three-source attention decoding is carried out on the fused feature, the static embedding is taken as the query, the space-frequency fusion feature is taken as the key, and the fused multi-angle feature is taken as the value, cross-domain attention calculation and up-sampling are carried out, and finally the three-dimensional voxel reconstruction result of the target object is generated.

This invention discloses a brain-computer interface (BCI) teaching demonstration system and control method, belonging to the field of BCI and artificial intelligence education technology. The system includes: a multi-channel EEG headband for synchronously acquiring EEG signals and head posture data; an edge AI hub with a built-in NPU accelerator for dynamically gating and adaptively filtering EEG signals based on head posture data, generating a two-dimensional time-frequency graph through complex Morlet wavelet transform, and using a lightweight RepEEG-Net neural network with structural reparameterization and quantization processing to analyze user intentions in real time, while simultaneously generating visualized teaching data; and a microcontroller execution chassis equipped with a real-time operating system (RTOS) for controlling the actions of the execution mechanism through multi-priority task scheduling. This invention achieves low-latency, highly interference-resistant brain-controlled demonstrations by deploying lightweight intelligent algorithms at the edge, and visualizes the algorithm processing process, significantly improving the real-time performance, robustness, and intuitiveness of the teaching demonstration.

The invention belongs to the technical field of computer vision detection, and provides a wind power blade dynamic vision detection imaging quality control method and system, and the method comprises the steps: an image quality evaluation step: initializing parameters and obtaining a blade surface image, calculating the optimal definition and the optimal information entropy, carrying out the optimization to obtain an optimal parameter combination, and obtaining a high-quality image; and an image definition evaluation step: performing dual-tree complex wavelet transform on the high-quality image, calculating logarithmic energy layer by layer according to scale factors, and adding the logarithmic energy to obtain an image definition value. According to the method, the optimal imaging parameter is automatically obtained through an algorithm, the optimal quality image is ensured to be obtained, and the problem that the image quality is uncontrollable in wind power blade detection in a complex environment is solved; a dual-tree complex wavelet transform method is introduced, so that the accuracy of definition evaluation is further improved on the basis of ensuring the operation speed of an image definition evaluation algorithm; by establishing an image quality evaluation and composite optimization algorithm, imaging parameters such as optimal focal length, focusing and exposure are determined, and an optimal image is obtained.

The invention discloses a data enhancement method based on dual-tree complex wavelet transform, and mainly solves the problems that a non-stationary signal data enhancement method in the prior art lacks multi-scale analysis capability, cannot fully retain local key features and is insufficient in model generalization capability. According to the scheme, the method comprises the following steps: 1) preprocessing an original signal to enable the original signal to meet the input requirement of dual-tree complex wavelet transform (DTCWT); 2) performing multi-scale decomposition on the signal by adopting DTCWT, and separating a low-frequency approximate component and a high-frequency detail component; 3) selecting a random zero sequence to replace the RZSR or a random noise sequence to replace the RNSR according to an enhancement requirement; 4) generating a random sequence, replacing high-frequency detail components, and meanwhile, retaining low-frequency components; and 5) reconstructing the signal through the inverse DTCWT, generating an enhanced signal, and outputting an enhanced signal. The method can effectively enhance the diversity of data, retains the global characteristics and local detail characteristics of signals, and can be used in the technical field of wireless communication.

This invention discloses a motor fault diagnosis method based on a large language model, belonging to the field of motor fault diagnosis. The method includes: first, collecting three-phase current data; then, using dual-tree complex wavelet transform to extract the amplitude and phase information of multi-layer complex wavelet coefficients as signal fault features, avoiding manual feature selection; next, constructing a multi-layer feature extraction and hierarchical weighted domain adversarial collaborative training framework that integrates grouping and Browsing optimization to extract multi-scale features and enhance domain invariance, thereby improving cross-condition generalization ability; finally, aligning the extracted signal embedding and text embedding across modalities through category-level comparative learning, and fine-tuning the large language model to recognize signal modes. This invention improves fault classification accuracy while enhancing model generalization and achieving textual semantic interpretation of diagnostic results.

The invention provides an anti-counterfeiting image texture feature extraction method and system, and the method comprises the steps: carrying out the YCbCr color space conversion of an anti-counterfeiting image, carrying out the two-dimensional dual-tree complex wavelet transformation of Y, Cb and Cr components, and obtaining a multi-scale multi-direction high-frequency sub-band coefficient matrix; energy of each high-frequency sub-band is calculated, global texture complexity parameters are determined based on the energy, a dynamic selection threshold value is generated, and the high-frequency sub-bands with the energy exceeding the threshold value are effective texture sub-bands; for the effective texture sub-band, calculating information entropy, determining the size of a sub-block, dividing the sub-block, and reserving the sub-block of which the standard deviation is greater than the mean value of the standard deviation of the sub-block in the sub-band; the effective texture sub-band reserved sub-block coefficients are scanned and spliced into a one-dimensional feature vector according to a Hilbert curve, sliding windows with odd lengths are used for traversing, a bit 1 is generated when window center coefficients are larger than a left neighborhood coefficient mean value and a right neighborhood coefficient mean value at the same time, otherwise, a bit 0 is generated, and an anti-fake image texture feature hash sequence is generated through splicing.

The invention discloses a U-KAN-based space target ISAR image component segmentation method, and relates to the technical field of ISAR image component segmentation, and the method comprises the steps: inputting an obtained to-be-detected ISAR image into a trained UD-KAN network, and achieving the component segmentation of the to-be-detected ISAR image; wherein the UD-KAN network is based on a U-Net network architecture, Tok-KAN modules are introduced into a coding path and a decoding path of the U-Net network architecture, and a down-sampling module in the coding path of the U-Net network architecture adopts an FDADM module based on dual-tree complex wavelet transform. According to the method, a UD-KAN network structure is introduced, a KAN layer can dynamically adjust the shape of an activation function according to input data, and the interpretability of the network is improved. The down-sampling module designed based on dual-tree complex wavelet transform can extract high-frequency and low-frequency components of the image at the same time, and the precision of ISAR image component segmentation is improved.

A radar image difference recognition method based on scattering center modeling and deep learning is proposed. This method utilizes a Gaussian scattering center model to parametrically model the target scattering source, extracting features such as position, amplitude, width, and direction. An automatic thresholding method based on mutual information is employed for image registration. Optimal rigid body transformation parameters are obtained through cross-threshold search and derivative-free optimization, effectively eliminating geometric errors caused by imaging offset. A difference map is generated using a logarithmic ratio operator, and high-confidence pseudo-label samples are obtained through hierarchical fuzzy C-means clustering. Simultaneously, virtual sample generation technology is combined to expand the training set. Finally, a wavelet-constrained convolutional neural network is constructed, incorporating dual-tree complex wavelet transform to enhance direction sensitivity, achieving accurate classification and diagnosis of scattering source changes. This invention significantly improves radar image registration accuracy, change detection robustness, and model generalization ability, making it suitable for applications such as stealth target performance evaluation, radar image detection, and target recognition.

The application relates to a Deepfake synthetic video detection method, system, device and medium, which comprises the following steps: dividing a data set to frame and extracting an RGB image I of a detection area of each frame; using a dual-tree complex wavelet transform to decompose the RGB image I into a low-frequency subband S L and a high-frequency subband set S H in different directions; inputting the low-frequency subband S L and the high-frequency subband set S H into a frequency domain feature extraction network respectively to obtain frequency domain features F DT‑CWT ; inputting the low-frequency subband S L and the high-frequency subband set S H into an image enhancement branch based on energy adjustment for processing to obtain space domain features F RGB ; fusing the frequency domain features F DT‑CWT and the space domain features F RGB to obtain single-frame classification features F C ; inputting the single-frame classification features F C into a preselected classifier network module to realize Deepfake video identification and output a judgment category.

A method and system for de-noising evoked potential waveforms is provided. In a first aspect, a method includes receiving an auditory brainstem response (ABR) waveform set. The ABR waveform set includes evoked potential waveforms for each ear, stimulation level, and frequency. The method includes generating a set of de-noised waveforms by applying a dual tree complex wavelet transform (DTCWT) to the set of ABR waveforms and minimizing an error between each de-noised waveform and its corresponding ABR waveform, while: limiting a total amplitude of complex wavelet coefficients to an amplitude threshold; limiting the total magnitude of the difference between the complex wavelet coefficients at adjacent stimulation levels to a level threshold; and limiting the total magnitude of the difference between the complex wavelet coefficients at adjacent frequencies to a frequency threshold.

The present application belongs to the technical field of power equipment insulation test, and relates to a transformer component level equivalent test and intelligent data cleaning method and system. The equivalent test and intelligent data cleaning method adopts an improved isolated forest model based on multi-scale feature fusion and adaptive threshold for data anomaly point detection, uses a deep learning model based on conditional generative adversarial network to complete missing data filling of high voltage test waveforms, and adopts an adaptive denoising algorithm based on dual-tree complex wavelet transform and an improved threshold function for feature extraction and denoising. The equivalent test and intelligent data cleaning system comprises a central control and scheduling module, an equivalent simulation module, a high voltage test module, a multi-parameter acquisition module and a data cleaning and intelligent analysis module, and the data cleaning and intelligent analysis module comprises a data anomaly point detection unit, a data missing filling unit and a feature extraction and denoising unit. The present application realizes automatic anomaly cleaning, missing filling and feature denoising of test data.

The application is a traffic monitoring image target detection method under adverse weather. The monitoring image to be recovered is input into a dynamic convolution module for convolution processing. After the dynamic convolution module, the image is input into a dual-tree complex wavelet enhancement module. The information enhanced by the dual-tree complex wavelet transform is input into a residual learning recovery module for residual learning to obtain a clean image. The picture output after the residual learning is sent to a YOLOv3 traffic target detection module for target detection. The application uses the dual-tree complex wavelet transform to combine the frequency characteristics and uses the dynamic convolution layer to combine the structural information to mine the robust features in the rain, snow and fog images. The residual network structure is used to refine the obtained features, and the residual learning operation is used to reconstruct the clean image. The application realizes better adaptability to the complex adverse weather scene and is used for the monitoring image recovery in the rain, snow and fog weather.

According to the invention, continuous complex wavelet transform and a deep learning technology are combined, a multi-component mixed Raman spectrum unmixing method is developed, and 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 single one-dimensional Raman spectrum into a two-dimensional matrix containing a real part and an imaginary part by utilizing continuous complex wavelet transform processing, and obtaining time-frequency domain characteristics of a multi-component mixed spectrum; (2) predicting the time-frequency domain characteristics of the target component by using a deep learning spectrum unmixing model; (3) realizing reconstruction of a target component spectrum by utilizing inverse transformation of continuous complex wavelet transformation; and (4) repeating the steps to predict the spectrums of all the target components and complete the unmixing of the original multi-component mixed Raman spectrums. Compared with a traditional Raman spectrum analysis method, the Raman spectrum multi-component signal de-mixing method based on continuous complex wavelet transform and deep learning can accurately separate independent Raman signals of different tissue structures and biochemical components in a complex environment in a living body, and has the advantages of being high in accuracy and high in accuracy. 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 visual inspection, and particularly relates to a weld joint quality inspection method and system. The method comprises the following steps: acquiring an original welding seam image, generating a feature map by combining gradient information and Laplacian information, and determining an active region after alternating sequence filtering; in the activation area, an oval solving window is arranged along the center line of the weld joint, the main axis direction is determined by the tangential direction of the center point, the proportion of the short axis to the long axis is adjusted according to local gradient features, complex wavelet transformation is carried out on pixels in each window, phases and module values are extracted, phase discontinuous points serve as feature points, and a three-dimensional point cloud is constructed; and a density clustering algorithm is adopted to analyze the point cloud, distance measurement comprehensive spatial distance and wavelet module value weighting are carried out, if a cluster with the point number exceeding a threshold exists in a clustering result, it is judged that the corresponding welding seam area is a defect area, and the clustering density threshold is determined according to the point cloud global density. According to the invention, high-sensitivity, high-reliability and low-misjudgment-rate welding seam defect accurate detection is realized.

The invention discloses a half-tone image super-resolution reconstruction method and system, and the method comprises the steps: carrying out the multi-scale feature extraction and fusion of a low-resolution continuous-tone image, enhancing the spatial domain feature through variable large kernel attention, decomposing the frequency domain feature through dual-tree complex wavelet transform, and carrying out the fusion of the multi-scale features of the low-resolution continuous-tone image. Feature fusion and resolution improvement are realized through local linear multi-head self-attention, and a high-resolution continuous feature map is obtained; and then, through a Gumbel-Softmax mechanism, carrying out derivable discretization processing and carrying out visual perception optimization so as to obtain a high-resolution halftone image. According to the method, under the combined action of spatial domain and frequency domain collaborative modeling, attention calculation of approximate linear complexity and end-to-end micro-discrete decision making, the detail authenticity, direction consistency and visual quality of a reconstructed image are effectively improved while the calculation overhead and memory occupation are remarkably reduced, and the method is suitable for being used in the field of image reconstruction. The method is suitable for edge-side low-delay and low-power-consumption real-time printing scenes.

The invention relates to the technical field of security and protection monitoring, in particular to an intelligent patrol management system based on hierarchical quality control, which comprises an acquisition preprocessing module, an intention quality control module, a privacy evidence storage module and a risk gating module. Performing complex wavelet transform on the data packet, constructing a pulse diagram, performing convolution reasoning, and outputting an intention probability and a second quality score; performing hashing, complex approximate homomorphic encryption and interval zero-knowledge proof on the event vector, and writing the gradient and the third mass score into a local evidence storage library; and finally, based on the condition in-danger value and reinforcement learning, dynamically adjusting a threshold value, triggering client prompting, robot re-checking, access control interlocking or alarming, and performing write-back on an execution hash value and a reverse gradient, thereby realizing low-missing-report, low-false-report and auditable closed-loop patrol.

The invention discloses a weak light image enhancement method and system based on dual-tree complex wavelet Retinex decomposition. The method comprises the following steps: firstly, decomposing an input image into a low-frequency illumination component and a high-frequency reflection component by using dual-tree complex wavelet transformation; for low-frequency components, a global illumination smoothing module including expansion convolution and global average pooling is designed to accurately estimate and smooth illumination. For high-frequency components, a directional attention mechanism is introduced to enhance texture and edge details. And finally, performing diffusion adjustment and refinement processing on the illumination component and the reflection component to realize self-adaptive reconstruction of the image. According to the method, the brightness, the contrast ratio and the detail visibility of the low-light image can be effectively improved.

The present application relates to a kind of partial discharge signal denoising method combined with singular value decomposition and wavelet transform, belong to signal processing field.The method includes: selecting suitable Hankel matrix is used to construct adaptive singular value decomposition model, responsible for the singular value decomposition of original noisy partial discharge signal;Calculate singular entropy increment, find the position of the maximum bending degree of singular entropy curve according to its asymptotic property;Use the position of maximum curvature as threshold, singular value greater than threshold is used to reconstruct original signal;Select suitable wavelet base function is used to construct one-dimensional two-stage double-tree complex wavelet transform, responsible for the decomposition and reconstruction of signal;Using q-shift scheme joint construction filter bank in DT-CWT interior.The present application can not only inhibit the narrow-band noise and white noise on PD signal well, but also can retain waveform mutation details, while having very low root mean square error and very high waveform similarity in index.