123 results about "Dictionary learning" patented technology

The invention discloses a soil heavy metal inversion method and system integrating satellite remote sensing and near-end sensing, and the method comprises the steps: collecting a soil sample, and measuring the soil heavy metal content and a soil visible light-near infrared spectrum; obtaining a time sequence multispectral image of a research area, calculating a spectral index, and selecting and screening bare soil pixels through a threshold value to obtain a bare soil image; performing spectrum correction on the bare soil image; obtaining a joint dictionary and a sparse coefficient through sparse representation and dictionary learning, and reconstructing a hyperspectral image of the bare soil image; converting the one-dimensional spectral data into a two-dimensional spectrogram by using continuous wavelet transform, extracting spectral features in combination with a 2D-CNN algorithm, and constructing a soil heavy metal inversion model; and using the trained inversion model to predict the soil heavy metal content of the research area based on the reconstructed hyperspectral image. According to the method, satellite remote sensing and near-end sensing are integrated to obtain a large-scale accurate soil heavy metal content distribution map, deep features are extracted in combination with a 2D-CNN algorithm, and the inversion model precision and model efficiency are improved.

Methods and apparatus are provided for encoding and subsequent decoding of video data by using a learnt video specific dictionary for implicit neural compression. An implicit neural representation comprising a head layer and a tail layer is used with approximations to the head layer parameters. The approximations are determined with combinations of atoms of a learnt video specific dictionary. In one embodiment, the head layer approximations and tail layer parameters are encoded in a bitstream. The dictionary is learnt at the decoder. In another embodiment, a dictionary is sent in the bitstream. At decoding, a reconstructed image is computed using transmitted INR parameters.

The invention discloses a foundation pit support deformation real-time monitoring method based on edge calculation, and aims to solve the problems of multi-node time synchronization and clock drift compensation caused by unavailability or poor quality of satellite navigation and precise time protocols in an underground or shielded environment. According to the method, through event dictionary learning, partition time slot scheduling and safety coding micro excitation time synchronization, matched filtering decoding, multi-scale arrival time estimation, paired time difference robust fusion, frequency priority and phase speed limit correction and network disconnection time holding, the time synchronization efficiency is improved; the technical effects that multi-source data alignment, deformation field reconstruction and graded early warning are completed on the unified virtual reference time scale, and high-precision synchronization and online drift compensation are obtained without depending on the external time service condition are achieved.

The invention relates to the technical field of data processing, in particular to a covariance inverse matrix recursive updating method for on-satellite hyperspectral anomaly detection, which comprises the following steps: acquiring a hyperspectral data stream; carrying out adaptive dimension reduction processing on the data stream and loading an initial background model; extracting local background statistics by adopting a sliding window mechanism, generating a background spectrum dictionary by utilizing online dictionary learning, calculating a reconstruction error between a current pixel and the dictionary and a Mahalanobis distance between the current pixel and a background model, and fusing to generate an abnormal score; performing abnormal confidence coefficient evaluation by combining the spatial context information and the spectral angle matching degree, and updating a spectral mean vector by using an exponential weighted moving average algorithm based on non-abnormal pixel data; and the updated model is injected into the next round of processing to form a recursive chain. According to the method, the high calculation complexity of direct inversion of a covariance matrix is avoided, real-time anomaly detection on a satellite is realized, the downloading amount of original data is remarkably reduced, and the congestion of a satellite-ground communication link is relieved.

The invention relates to a deep energy level transient spectrum data processing method based on dictionary learning. The method comprises the following steps: acquiring transient capacitance signals of a sample at different temperatures by using deep energy level transient spectrum equipment; establishing an Arrhenius equation that the defect emissivity changes along with the temperature; the defect emissivity under each group of values is calculated according to all possible values of the coefficient of the exhaustion Arrhenius equation; constructing a dictionary matrix based on the defect emissivity amplitude introducing Arrhenius constraint, and reconstructing a transient capacitance signal by using the dictionary matrix; and constructing a convex optimization problem, solving a dictionary index variable which enables a reconstruction signal of the transient capacitor to be closest to an acquisition signal, and extracting a non-zero solution in the dictionary index variable to obtain a corresponding defect emissivity amplitude spectrum. According to the method, the detection precision of the deep energy level defect can be improved by introducing the Arrhenius constraint relation.

The invention discloses a power distribution network fault signal reconstruction method, system and device based on K-SVD and orthogonal matching pursuit, and a medium, and belongs to the technical field of power systems, and the method comprises the steps: obtaining a fault signal from a monitoring device of a power distribution network, and carrying out the preprocessing of the signal; a K-SVD algorithm is used for dictionary learning, a sparse dictionary suitable for power distribution network fault signals is generated, and the dictionary is optimized through training of a signal data set; performing sparse representation on the signal by adopting an orthogonal matching pursuit algorithm, and realizing sparse decomposition of the signal by searching an optimal sparse representation coefficient; based on the sparse representation coefficient, reconstructing the signal from the sparse dictionary, calculating a reconstruction error, comparing the reconstruction error with an actual fault signal, evaluating a reconstruction effect and an error, and performing parameter optimization adjustment; and accurately judging the type of the fault according to the reconstructed signal. According to the method, the key information of the fault signal is effectively extracted in a complex noise environment, and the signal reconstruction precision is remarkably improved.

The invention belongs to the technical field of GIS fault detection, and discloses a GIS latent insulation fault diagnosis method and system under strong noise, and the method achieves the omnibearing signal capture through synchronously obtaining optical, mechanical, electrical and acoustic signals, and solves a problem of weak signal leak detection. For strong noise interference, noise and fault characteristic frequency bands are effectively separated by using multi-modal signal difference and constructing a time domain and wavelet domain cascaded double-layer dictionary. A discriminative sparse model is combined with a K-SVD algorithm to carry out end-to-end iterative optimization, sparse features with higher discriminative force are automatically mined, and the problem of deep feature mining is solved. For the problem of weak generalization ability of small samples, a classifier error term is introduced into an objective function, a joint optimization objective function fusing reconstruction and classification errors is constructed, dictionary learning and classifier training are combined into one, the generalization ability of the model is significantly enhanced, and the method is suitable for large-scale popularization and application. Therefore, timely and accurate identification of the latent insulation fault under the conditions of strong noise and small samples is realized.

The invention belongs to the technical field of image analysis and processing, and discloses a woven fabric texture reconstruction method based on deep shared convolutional dictionary learning, and the method specifically comprises the steps: (1) constructing a deep shared convolutional dictionary learning model; (2) decomposing an overall optimization problem of the constructed deep shared convolutional dictionary learning model into N independent optimization problems; (3) rewriting the N independent optimization problems into three sub-optimization problems with constraints, and converting the three sub-optimization problems into optimization problems without constraints according to an alternating direction multiplier method; (4) inputting woven fabric texture sample images in the training set, and iteratively solving an optimization problem without constraints by adopting a multi-stage training strategy; and (5) inputting the test set image into the deep shared convolutional dictionary learning model, outputting the corresponding adaptive convolutional dictionary and convolutional coding coefficient, and finally calculating to obtain a reconstructed image. According to the method, high-quality reconstruction of a complex woven fabric texture image and general representation of an untrained woven fabric texture image are realized.

The invention provides a foundation pit support structure lateral displacement prediction method based on dictionary learning fusion monitoring data. The foundation pit support structure lateral displacement prediction method comprises the steps that S1, a physical parameter database is established; s2, performing conversion to obtain weak prior distribution of a compressibility parameter ES; s3, sampling by using Latin hypercube to obtain N alternative parameter combinations of the soil body; s4, establishing a finite element model of the target foundation pit; s5, inputting a parameter combination to calculate corresponding side displacement data of the enclosure structure, and forming an over-complete dictionary D; s6, reading measured side displacement data of the enclosure structure to form a monitoring data matrix Y; S7, extracting corresponding data in the over-complete dictionary D, and forming a dimension transformation matrix A matched with the dimension of the monitoring data matrix Y; s8, obtaining an important atom set and a sparse solution vector x by using an orthogonal matching pursuit algorithm program; and S9, obtaining a side displacement prediction curve of the building envelope corrected by combining the monitoring data. The method has the advantages of high prediction precision, low calculation cost and high response speed.

This invention discloses a noise reduction method, a bearing fault diagnosis method, and a system based on dual sparse dictionary adaptive learning. The method performs wavelet decomposition on the signal to be denoised to obtain high-frequency and low-frequency signals, constructing high-frequency and low-frequency matrices. Then, threshold-adaptive DDTF dictionary learning is applied to both the high-frequency and low-frequency matrices. Specifically, an initial dictionary is set and learned to obtain an initial sparse coefficient matrix. An adaptive threshold update is then applied to the initial sparse coefficient matrix, selecting sparse coefficients from the initial sparse coefficient matrix in descending order as dynamic thresholds during the iteration process to update the initial sparse coefficients. After determining the final threshold, the final sparse coefficient matrix is ​​obtained, and the dictionary is updated again to obtain sub-signals. Finally, inverse transformation and matrix rearrangement inverse operations are performed on the obtained sub-signals to obtain the denoised signal. This invention combines wavelet decomposition and DDTF to construct a dual sparse pattern, effectively improving the sparse representation capability of a fixed basis.

This invention discloses a high-overload data analysis method and system based on multi-dimensional buffer protection, relating to the field of data acquisition and storage technology. The system consists of several functional modules, including: a transient impedance module, which acquires stress wave propagation parameters and material strain data collected in real time by the recorder node, constructs a dynamic digital twin model of the buffer structure, and outputs impact feature vectors and stress field distribution matrices; a signal reconstruction module, which dynamically configures the sensing signals based on the impact feature vectors, including the gain, bandwidth, and filtering parameters of the analog front end, conditions and performs analog-to-digital conversion on the sensing signals, and uses compressed sensing and dictionary learning algorithms to sparsely reconstruct abnormal data, generating a high-fidelity data stream and storage data integrity identifiers for each storage block; and a federated storage module, which inputs the stress field distribution matrix and storage data integrity identifiers into the edge-side reinforcement learning decision engine to calculate the failure risk probability of each storage block.

The invention relates to the technical field of image fusion, in particular to a hesitant fuzzy variable granularity dictionary learning-based multi-modal image fusion method, which comprises the following steps of: firstly, adaptively selecting division granularity according to image quality, and partitioning a source image into blocks; then extracting image block features and calculating hesitant fuzzy membership degrees of the image block features so as to quantitatively represent uncertainty information in the image; obtaining a joint over-complete dictionary and a sparse coefficient through dictionary learning, and fusing the hesitant fuzzy entropy and a granularity coefficient to construct an adaptive weight; and finally, fusing the sparse coefficient by using the weight and reconstructing a fused image. The problems of image fuzzy processing, structure multi-scale expression and insufficient adaptive feature extraction capability are effectively solved.

The present application belongs to the technical field of data analysis, and particularly relates to a high-throughput biosensing system for early warning of algal blooms and pathogens, which comprises multiple sampling units, a microfluidic chip with a multi-channel structure, and a high-throughput surface-enhanced Raman detection unit with an integrated metal nanostructure surface, and obtains the original surface-enhanced Raman spectrum of the water sample through a Raman spectrum acquisition mechanism. The system uses compressed spectrum sampling and sparse reconstruction methods to reduce the data volume and maintain the integrity of the spectrum information, constructs a pathogen Raman fingerprint dictionary through a dictionary learning method, and realizes the identification of different algal metabolites, algal toxins and pathogen-related molecular components and the concentration interval judgment thereof in combination with a sparse representation classification mechanism. Based on the type and concentration interval of the pathogens, the system further generates algal bloom risk warning and pathogen risk warning, and realizes early discovery of abnormal changes in the water body. The present application has significant advantages in stability, sensitivity, throughput and real-time performance.

The application discloses a chip packaging defect detection and positioning method and device, and the method comprises the following steps: performing PCA preprocessing on a test set and a training set of a chip packaging image; training a deep tensor dictionary learning method based on the features extracted from the training set, so as to obtain a trained deep tensor dictionary; determining a sparse coding third-order matrix based on the features extracted from the test set and the trained deep tensor dictionary; and judging whether the chip packaging image to be detected has defects and positioning the defects based on the sparse coding third-order matrix. The application describes the main features of the image through a three-dimensional dictionary, strengthens the feature expression ability of the deep tensor dictionary by combining deep learning, and approximately represents the image by combining sparse coding, so that whether the image is abnormal can be judged according to the error between the actual image and the approximate representation, and the abnormal position of the image can be positioned according to the sparse coding, thereby improving the detection speed and the detection accuracy of the chip packaging defect detection.

The invention belongs to the technical field of ocean exploration, and particularly relates to an ocean exploration wave field reconstruction method based on dictionary learning, and the method comprises the steps: carrying out the preprocessing of original wave field data, and obtaining the preprocessed wave field data; separating the preprocessed wave field data into non-overlapping block sample matrixes; converting all block sample matrixes into a sample vector set; selecting a training sample set from the sample vector set; adopting the training sample set to construct a double-sparse dictionary model; iteratively updating the double sparse dictionary model by adopting an alternate optimization mode to obtain a final sparse dictionary and a sparse coefficient matrix corresponding to the sparse dictionary; and carrying out sparse reconstruction on the actual observation wave field by adopting the final sparse dictionary and the sparse coefficient matrix corresponding to the final sparse dictionary. The method has high adaptability to noisy or under-sampled wave field data, and a continuous and high-fidelity wave field structure can still be recovered even under the condition that data acquisition conditions are limited.

The invention discloses a carpet type DDoS attack detection system based on address-behavior traffic decomposition, and the system is characterized in that an offline multi-scale dictionary learning module learns a multi-scale behavior template dictionary based on historical benign network traffic, and builds a cross-IP baseline activation standard and an anomaly detection threshold value of each behavior template; the online rapid decomposition module decomposes the real-time network traffic by using the behavior template dictionary, calculates the activation intensity of each IP address on each behavior template and generates a real-time activation matrix; and the collaborative anomaly detection module calculates abnormal activation conditions of the behavior template on a plurality of IP addresses based on the real-time activation matrix, and performs joint judgment through spatial diffusion and total intensity to identify carpet DDoS attacks and output an attacked IP address list. According to the method, efficient identification of cross-IP collaborative abnormal traffic is realized, and the problem that carpet-type DDoS attacks are difficult to detect in a traditional method is solved.

The application belongs to the technical field of image segmentation, and particularly relates to a complex scene target segmentation method and system based on image recognition, which comprises the following steps: performing superpixel adaptive division on an input image, fusing gray scale and texture features to determine a superpixel boundary, mapping the superpixel boundary into a graph node and calculating an edge weight, and constructing an undirected weighted graph; adaptively encoding a graph signal, utilizing a hybrid graph wavelet-Fourier joint transform to optimize and separate features, and obtaining purified graph frequency domain features; sparsely reconstructing features through adaptive super-complete dictionary learning, and obtaining target enhanced features; extracting topological parameters based on an improved persistent homology, constructing a topological constraint feature graph, mapping the topological constraint feature graph into an initial contour field, iteratively optimizing a level set and a contour through an adaptive partial differential equation, and obtaining a high-fidelity coarse segmentation result; extracting geometric features to construct a joint constraint model to repair an occluded area, and outputting a precise segmentation result. In the application, sparse topological modeling is adopted, weak features are strengthened, and target discrimination accuracy is improved.

The present application relates to the technical field of image reconstruction, and more particularly to a spectral super-resolution reconstruction method based on dynamic dictionary learning. It comprises: building a DLTN network architecture, including a feature encoding unit, a spatial down-sampling unit, a multi-level dictionary learning and sparse coding unit, a feature decoding unit and a global feature fusion mechanism; obtaining a hyperspectral training dataset, preprocessing the input low-resolution RGB image, and inputting the network for training; inputting the low-resolution RGB image to be processed into the trained network, and sequentially processing it through feature encoding, spatial down-sampling, multi-level sparse coding and feature enhancement, multi-depth feature global fusion and feature decoding, and outputting the reconstructed hyperspectral image. The advantage is that dynamic dictionary learning and lightweight Transformer architecture are fused, feature sparse representation is realized through the DLTSC module, and the calculation cost is reduced; while ensuring the reconstruction accuracy, the model parameter quantity and the floating point operation frequency are significantly reduced.

The application provides a super-large-scale MIMO mixed field channel estimation method based on dictionary learning, establishes a mixed field channel scene, the scene comprises a base station, at least one near-field scatterer, at least one far-field scatterer and K users; a mixed field channel model is constructed according to the mixed channel scene; and a mixed field channel recovery problem is established according to the mixed field channel model and solved. The application can well capture the channel characteristics of the mixed field channel and realize accurate estimation.

The invention belongs to the field of magnetic resonance sounding signal noise filtering, and particularly relates to a magnetic resonance sounding signal de-noising method based on backtracking generalized OMP, and the method comprises the steps: employing a magnetic resonance sounding water detector to carry out the sampling of a sparse signal at a frequency lower than the Nyquist frequency, and obtaining a noisy MRS signal; converting the noisy MRS signal into a two-dimensional matrix signal; dictionary learning is carried out on the two-dimensional matrix signals through a KSVD algorithm, and an updated dictionary is obtained after iteration; calculating a recovery matrix and measurement data by using an independently distributed Gaussian random matrix; performing iterative operation on the measurement data by using a backtracking generalized orthogonal matching pursuit algorithm, selecting an atomic sequence of maximum projection of the measurement data and a recovery matrix module, and obtaining a sparse coefficient matrix of the measurement data under the recovery matrix; and calculating a recovery signal according to the dictionary and the sparse coefficient matrix of the measurement data under the recovery matrix.

The invention relates to the technical field of mobile vehicle-mounted recognition, in particular to a mobile vehicle-mounted recognition method and system based on multi-domain dictionary learning, and the method comprises the steps: decomposing a multi-domain representation mobile vehicle-mounted signal into a dead load component, a narrowband vehicle-mounted component and a broadband vehicle-mounted component, after a dead load component dictionary, a frequency domain dictionary and a time domain dictionary are adopted for corresponding representation, an axle control equation is constructed in combination with a system matrix; constructing a mobile vehicle-mounted identification model based on the vehicle-induced bridge response and the axle control equation; solving the mobile vehicle-mounted identification model through iterative optimization of dead load component identification, sparse coding, frequency domain dictionary updating and time domain dictionary updating to obtain a dead load component coefficient, a narrowband vehicle-mounted component coefficient and a broadband vehicle-mounted component coefficient; and substituting the solved dead load component coefficient, the narrowband vehicle-mounted component coefficient and the broadband vehicle-mounted component coefficient into a mobile vehicle-mounted identification model to realize mobile vehicle-mounted identification. According to the invention, high-precision identification of the mobile vehicle can be realized.

The application relates to a fine image classification method, which is characterized by the following steps: constructing a decomposition dictionary learning model, constructing a self-adaptive local order preserving constraint model, constructing a classifier model, constructing a target function, solving the target function and constructing a classification method model. The application can not only adaptively capture the local geometric structure information of atoms and the high-order rank information features of the neighborhood of each decomposition atom in the decomposition dictionary learning process, but also can measure the contribution of multi-view data to the representation of an object through the distance between atoms, enhance the discrimination performance of the decomposition coefficients and the decomposition dictionary. Furthermore, the decomposition coefficients and the class label matrix of the fine image are used to construct a classifier model, the classifier parameters are constrained by a norm, the robustness of the classifier learning model is enhanced, the time complexity of the algorithm is reduced, and the classification performance of the adaptive local order preserving constraint multi-view decomposition dictionary learning algorithm is improved.

The application discloses an image classification method based on adaptive local ordinal preserving analytic dictionary learning and belongs to the technical field of computer vision. The method comprises the following steps: performing feature extraction on images in a data set, dividing the images into a training set and a test set, adopting a support adaptive local ordinal preserving analytic dictionary learning model to simultaneously learn an analytic dictionary and a classifier based on the training set, then calculating coding coefficients of the test set based on the learned analytic dictionary, and finally obtaining class labels of the test set through the classifier and the coding coefficients of the test set. The method introduces an adaptive ordinal local preserving term based on a discriminative convolution analytic dictionary learning model, simultaneously preserves neighborhood correlation between dictionary atoms and distance ordering information of atoms in the neighborhood in the learning process, optimizes the analytic dictionary learning model, enhances the discriminativeness of the dictionary, and improves the image classification accuracy of the model in general scenarios such as face recognition, object recognition and scene recognition.

The invention discloses a near-field millimeter wave super-resolution imaging method based on nuclear adaptive filtering, and relates to the field of super-resolution imaging. The method is based on a kernel adaptive filtering operator, through a lightweight learning method, it is ensured that estimation characteristics conforming to original data to the maximum extent are met, and a high-resolution image is efficiently predicted and reconstructed. Aiming at a subsequent image processing process of a two-dimensional near-field millimeter wave imaging system, under the condition of shortening data sampling time, a high-resolution image is reconstructed from low-resolution sampling data, and redundant hardware, a large number of image data sets and a dictionary learning process are not needed. Based on the method provided by the invention, the optimal visual effect is achieved in data sets with different shape features, and detail information in an amplified view is clearest. And compared with a typical super-resolution algorithm, the effect is better. Compared with a typical super-resolution algorithm, the method provided by the invention has the highest numerical value similarity and higher structural similarity.

Three-dimensional sensing data is transferred efficiently by applying compressed sensing using dictionary learning. Conventional image-based lossy compression improves a feature according to which power is concentrated on low-frequency components and deterioration of three-dimensional information increases. By changing the decimation rates for the depth information and the color information and using a dictionary vector created from the other restoration result for restoration, the original result is reconstructed using a small data volume.

The application relates to a fine image classification method based on deep migration dictionary learning, which is characterized by comprising the following steps: constructing a deep dictionary learning model, learning a source field dictionary and a target field dictionary to extract complex nonlinear features of fine images, then solving a target function of an output layer in each layer of the deep dictionary learning model by using a gradient descent method to obtain an encoding dictionary and an encoding coefficient of the output layer, and finally obtaining classifier parameters through the obtained encoding dictionary and the encoding coefficient to construct a classification method model. The application solves the problem that source field fine images and target field fine images belong to different distributions, thereby improving the classification performance of the fine images.

The invention discloses a method for improving face recognition accuracy in a low-light-source environment, belongs to the field of robot vision and biological feature recognition, and solves the problems of low accuracy, poor interference resistance and weak generalization in the prior art. According to the core scheme, ambient light data is collected through a spectrum sensor, an optimal narrow-band light parameter is calculated and directionally emitted, and a high-signal-to-noise-ratio face image is collected by combining a wave band matching sensor; using an improved multi-scale sparse dictionary learning algorithm to enhance details and suppress noise; and finally, extracting features through a frequency domain-spatial domain double-branch attention network to complete recognition. Through the collaborative design of physical light modulation and algorithm optimization, the recognition accuracy under the environment of less than or equal to 50 lux is improved, the adaptability and generalization are excellent, devices are mature and easy to engineer, and the method can be widely applied to service, security and protection robots and other equipment.

The invention belongs to the technical field of 6G wireless communication, and particularly relates to a high-precision sensing positioning method design in a near-field millimeter wave / sub-terahertz system. According to the method, more accurate estimation angles are obtained by utilizing block sparse features in near-field channel representation and adopting a low-complexity block sparse scheme, and a triangular geometrical relationship is strictly met by automatically adapting to the estimation angles corresponding to the two sub-arrays, so that distance information does not need to be given to a user, and the user experience is improved. And then an estimated distance is obtained by using a triangular geometrical relationship of the two sub-arrays, and a dictionary matrix representing a channel is continuously updated by using a dictionary learning scheme, so that more accurate angle and distance sensing performance is obtained, and high-precision sensing positioning is realized. Simulation results show that the near-field positioning scheme of the invention better balances low complexity, strong noise robustness, no need of user distance information and good performance, and has practical application significance in low-altitude network and agent communication.

The application relates to an improved ESPIRiT reconstruction method based on outer product effectiveness and dictionary learning, and belongs to the technical field of magnetic resonance imaging. ESPIRiT is a parallel magnetic resonance imaging technology for estimating multiple sets of sensitivity maps to realize image reconstruction by using K-space calibration information. The application is based on an ESPIRiT model, combines an SOUPDIL regular term containing an L0 norm, and proposes an improved ESPIRiT reconstruction algorithm based on SOUPDIL, named SOUPDIL-ESPIRiT, uses FISTA technology for solving, and realizes parallel magnetic resonance imaging reconstruction through two steps of dictionary learning and image updating. Experimental results show that the application can better promote image sparsity, eliminate image reconstruction noise and artifacts, significantly improve the precision of the reconstructed image, and has the ability of better retaining image texture details and edge contour information.