57 results about "Matrix decomposition algorithms" patented technology

The invention discloses a fire smoke image early recognition method based on multi-modal fusion. The method comprises the following steps: S1, collecting multi-modal data and carrying out time alignment; s2, constructing a visible light branch, an infrared branch and a photoacoustic branch, generating multi-modal features, and storing the multi-modal features in a shared feature pool; s3, embedding a space-time tracking module in the shared feature pool, and generating global feature representation; s4, generating a dynamic background template by adopting a low-rank matrix decomposition algorithm, and separating and removing background noise; s5, multi-modal features are fused in the shared feature pool through a lightweight self-attention mechanism, and missing features are compensated; s6, enhancing an early smoke signal through a sparse feature enhancement algorithm, and performing fire state classification and smoke diffusion path prediction by applying a multi-task classification network; and S7, updating the shared feature pool and the space-time tracking module through a dynamic feedback mechanism. According to the invention, early-stage accurate identification of fire smoke is realized by using multi-modal fusion, space-time tracking and dynamic feedback methods.

The invention relates to the technical field of optical detection, in particular to a notebook computer shell injection molding part surface quality detection method and system, and the method comprises the steps: irradiating a standard part with a preset wave band according to preset parameters through an infrared spectrometer to obtain a first spectrogram, and irradiating the injection molding part with the preset wave band according to the preset parameters to obtain a second spectrogram. Through the synergistic effect of a non-negative matrix factorization algorithm and a Transform fusion model, the problem of spectrum overlapping caused by addition of a flame retardant is effectively solved, through seven-degree polynomial and smooth second derivative processing, the carbonyl peak position measurement error is further reduced, and compared with a traditional Fourier transform infrared spectroscopy, the feature parameter extraction precision is further improved. The number of the carbonyl absorption peaks and the positions of the carbonyl absorption peaks are compared with the standard threshold values twice, so that the problem that some qualified products are mistakenly considered as non-qualified products easily due to deviation during judgment of the carbonyl content of the injection molding shell of the notebook computer is further solved.

The invention discloses a wireless audio star flash transmission noise management method and system suitable for a conference system, and the method comprises the steps: obtaining to-be-transmitted audio information, coding the to-be-transmitted audio, carrying out the preliminary separation of the to-be-transmitted audio based on a non-negative matrix factorization algorithm, and obtaining the separated audio information; performing audio classification according to the separated audio information, and classifying to-be-transmitted audios into different audio types to obtain audio classification information; acquiring environment sound audio information according to the audio classification information, forming an environment sound audio sequence, and performing noise reduction management; human voice audio analysis is carried out, the human voice audio is divided into background human voice audio and target human voice audio, noise reduction is carried out on the background human voice audio, and anomaly detection is carried out on the target human voice audio; and carrying out comfort evaluation on the target human voice audio, judging whether human noise exists or not, carrying out audio prediction, and generating noise reduction regulation and control parameters. And the audio transmission noise management quality of the conference scene and the comfort of participants are improved.

The invention relates to a three-component high-dimensional seismic data reconstruction method and device based on vector tensor reduced rank, a medium and equipment, and the method comprises the steps: firstly, placing three-component high-dimensional seismic data at three imaginary part positions of a quaternion array in a time domain through employing a quaternion tool, and setting a real part of the quaternion array to be 0 value; and realizing vector joint representation of the quaternion-based three-component seismic data. Quaternion representation can protect the nonlinear orthogonal relation of three component seismic data in the space direction, and vector field structural characteristics of seismic waves are kept. And secondly, transforming a time domain quaternion array into a frequency domain by using quaternion Fourier transform, extracting a spatial high-dimensional quaternion array corresponding to each frequency domain slice, and performing missing data restoration and reconstruction on the high-dimensional quaternion array by using a tensor reduced-rank reconstruction technology based on a quaternion parallel matrix decomposition algorithm. According to the method, the memory occupation amount is remarkably reduced, the calculation amount is small, the calculation cost is low, and the reconstruction calculation efficiency is high.

The invention relates to a slope safety coefficient and sliding surface range solving method, device and equipment. Firstly, a slope simulation model is constructed; performing grid division to obtain a plurality of unit bodies; determining the cohesive force in each unit body, the distribution condition in the internal friction angle and the displacement value of the node corresponding to each unit body through a random field; based on an autocorrelation function between adjacent unit bodies in the two-dimensional random field, generating a group of cohesive force and internal friction angle data by adopting a matrix decomposition algorithm; calculating a plurality of groups of cohesive force and internal friction angle data according to the set sample number, solving the standard deviation and variance of the slope safety coefficient corresponding to the plurality of groups of cohesive force and internal friction angle data, and calculating the range of the slope safety coefficient according to the standard deviation and variance; and calculating a plurality of groups of clusters with a set sample number according to a clustering algorithm, and determining a sliding surface range according to the clusters. According to the invention, a reliable and accurate safety coefficient and a sliding surface range can be obtained.

The invention provides a quantum circuit system based on an AES cryptosystem and an encryption method. The quantum circuit system based on the AES cryptosystem comprises a byte replacement module which adopts a pre-optimized S-box query circuit to realize non-linear byte replacement; the row shift module completes in-row byte cyclic shift through quantum bit exchange operation; the column confusion module realizes a matrix multiplication circuit based on finite field operation, and a quantum circuit of the column confusion module is generated by a depth-first matrix decomposition algorithm; and the round key adding module realizes XOR operation of a round key and a state by using a CNOT gate array, and a key required by the round key adding module is dynamically generated through a key expansion circuit. And a compact Z-shaped architecture is adopted to optimize time sequence scheduling and resource allocation of each module. On the premise of keeping the quantum bit resource usage amount equivalent to that of an original Z-shaped architecture, the quantum circuit depth is remarkably reduced, and higher fidelity and shorter execution time can be realized on NISQ equipment.

The invention discloses an interpretable task-specific cross-city migration learning method in the technical field of smart cities and cross-city migration, and the method comprises the following steps: dynamically extracting a CTS region of a source city based on a depth matrix decomposition algorithm, calculating the similarity between the source city and a target city through employing the auxiliary information of city POI and road network data, and carrying out the task-specific cross-city migration learning. According to the cross-city knowledge migration method based on the depth matrix decomposition technology, the problems that an existing cross-city knowledge migration technology is large in calculation expenditure, low in migration effect, poor in interpretability and the like are effectively solved, and the cross-city knowledge migration method based on the depth matrix decomposition technology based on the depth matrix decomposition technology has the advantages that the calculation cost is high, the migration effect is low, and the interpretability is poor. The calculation complexity is obviously reduced, and the adaptability to the target task and the migration accuracy are improved. Besides, which areas are beneficial to the target city can be clearly identified, and efficient data completion is realized under the condition of data scarcity.

The present invention discloses a method and system for on-demand intelligence distribution, comprising the following steps: (1) receiving superior telegraphic information to obtain action tasks and intelligence information; (2) extracting features: commander features, action task features, and intelligence features; (3) error processing and model optimization; and (4) outputting intelligence in order of interest based on the optimized model. The present invention preprocesses task and intelligence text using an improved linear discriminant analysis method, encodes task and intelligence data using a sparse edge denoising autoencoder, and uses a matrix decomposition algorithm to remove blank portions of the matrix for dimensionality reduction, thereby effectively improving the accuracy of the training model, while also increasing the precision of on-demand intelligence distribution and enhancing the command efficiency of commanders.

The invention discloses a fitness quality evaluation method based on myoelectricity and images, which comprises the following steps: step 1, data acquisition: the acquired data comprises original myoelectricity data with time sequence information and time sequence image data of fitness actions; step 2, preprocessing the collected data; step 3, for the preprocessed myoelectricity data, extracting myoelectricity characteristics through a sliding window by adopting a time domain intermediate frequency value model, and then decomposing a muscle activation degree matrix into a collaborative structure factor matrix and an activation coefficient matrix through a non-negative matrix decomposition algorithm; for the preprocessed time sequence image data, adopting a BlazePose algorithm to extract 33 3D articulation points in the time sequence image data and calculating the Euler angles of the articulation points so as to obtain attitude angle features; and step 4, using a support vector machine to fuse the myoelectricity features and the attitude angle features to carry out action scoring. The method can effectively improve the fitness quality of the user.

The invention relates to the technical field of embryonic development evaluation, and discloses an embryonic chromatin openness and gene expression evaluation method, which comprises the following steps: extracting an embryonic cell sample with a survival rate of more than or equal to 95% through a micromanipulation technology; acquiring chromatin open signal data by adopting an ATAC-seq technology; carrying out genome positioning by utilizing a BWA comparison tool, and constructing a three-dimensional contact matrix through a matrix decomposition algorithm in combination with Hi-C data; determining the gene expression quantity by adopting fluorescent quantitative PCR (Polymerase Chain Reaction); and constructing an XGBoost integrated learning model to fuse chromatin characteristics and gene expression data. By optimizing micromanipulation parameters and an ATAC-seq experimental process, the cell survival rate is increased to 95.2 + / -1.3%, the ATAC-seq library complexity reaches 85.3%, and Hi-C data standardization processing is combined, so that the technical problem of data deviation is solved, and the embryo sample processing efficiency and data reliability are remarkably improved.

This invention provides a method, medium, and system for acoustic early warning of thermal runaway in lithium-ion batteries based on feature reconstruction, belonging to the field of lithium-ion battery technology. The invention constructs a positive sample set by collecting safety valve opening sounds through multi-condition thermal runaway experiments, and expands the samples using data augmentation. Multi-resolution Mel spectra are extracted from the audio signals. These Mel spectra are then input into a complex-domain phase-aware separation model for joint estimation of complex-domain amplitude masking and phase residuals. Physical prior corrections are applied to the reconstruction results using a sound source separation algorithm based on wave equation time-frequency inverse scattering and a low-rank sparse time-frequency matrix decomposition algorithm based on random matrix theory. The three corrected signals are weighted and fused to obtain a corrected Mel spectra, which are finally input into a temporal convolutional network to classify and identify the safety valve opening sounds and output a thermal runaway early warning signal. This invention solves the technical problem of insufficient accuracy in thermal runaway early warning caused by acoustic feature reconstruction distortion in complex noise environments.

The invention relates to a TSN (Time Sensitive Network) timestamp recovery method based on an FPGA (Field Programmable Gate Array), which belongs to the technical field of TSNs and comprises the following steps of: S1, collecting synchronous messages of different TSN nodes in parallel by multiple ports of the FPGA, and analyzing timestamp data by taking a message serial number as an anchor point; s2, constructing an incomplete timestamp matrix and a mask matrix by adopting a time sliding window mechanism, and automatically marking missing data; s3, a random index set is generated through LFSR, and matrix sparse sampling is achieved through a bitmap judgment unit; and S4, based on an FPGA accelerated CUR matrix decomposition algorithm, solving a pseudo-inverse matrix through parallel MGS-QR decomposition, and reconstructing a complete timestamp matrix through block parallel matrix chain multiplication. The FPGA parallel architecture, ping-pong cache and fixed-point number processing are adopted, microsecond-level low-delay recovery is achieved, hardware resource occupation is greatly reduced, robustness in a severe network environment is improved, and the method is suitable for a high-precision clock synchronization scene of an industrial TSN node.

The invention discloses a convolutional non-negative pattern learning method based on maximum boundary likelihood estimation under generalized Kullback-Leibler (KL) divergence, and the method comprises the steps: researching a likelihood function of a convolutional non-negative pattern under the generalized Kullback-Leibler divergence, and carrying out the product division of an activation function from a model in an integral mode, according to the method provided by the invention, the local correlation of the columns in the observation data V can be modeled, and the used maximum likelihood estimation method has the characteristic of automatic model order selection. When an integral without an analytic solution is processed, the lower boundary of the model is continuously optimized and improved through a variational inference method, the target of maximum marginal likelihood is achieved, a final result not only keeps the simplicity of an observation data form, but also has the computational complexity obviously lower than that of a sampling algorithm represented by a Markov chain Monte Carlo algorithm, and the method has the advantages that the method is simple and easy to implement. And the calculation complexity is similar to that of a convolution non-negative matrix factorization algorithm based on multiplication iteration.

The invention provides a credit card application processing method and device, equipment, a medium and a product, relates to the field of artificial intelligence, and can be applied to the field of financial science and technology. Comprising the steps that credit card information of a target user is acquired, and the target user is a user who fills in credit card application information but does not submit the credit card application information; the credit card information of the target user is analyzed by adopting a caching technology and a data matching algorithm, intra-bank credit card information of the target user is determined, and the intra-bank credit card information is used for indicating that an intra-bank credit card is held or not held; if the intra-bank credit card information represents that the target user does not hold the intra-bank credit card, target feature information in the credit card information is extracted according to an auto-encoder algorithm; according to the target feature information and the matrix decomposition algorithm, the first service information is pushed to the target user, the first service information is information of at least one target credit card determined according to the demand of the target user, the application efficiency of the credit card is improved, and the satisfaction degree of credit card application of the user is improved.

The invention provides a belt tearing detection method based on image data processing, a medium and a system, and belongs to the technical field of image data processing.The belt tearing detection method comprises the steps that firstly, a plurality of high-resolution cameras are used for collecting belt surface images from different angles, and standardized images are obtained through automatic exposure adjustment and automatic focusing processing; next, extracting a gray feature matrix of the image, decomposing texture features and state features by using a matrix decomposition algorithm, performing clustering analysis on the state feature matrix through a parallel computing method, and extracting a feature vector representing the surface state of the belt; then, according to a preset tearing judgment equation set, the tearing risk index of each area is calculated, threshold analysis is carried out, different belt damage grades are divided, and the identification result of the belt damage area is obtained. According to the method, a processing technology of image data of the running belt is fully utilized, and the technical problem that the tearing damage risk of the running belt is difficult to efficiently and quickly obtain in the prior art is solved.

The present application relates to the field of group portrait, disclose a kind of user portrait generation method, device, computer equipment and storage medium, its method includes: obtaining user interaction history record;According to user interaction history record generates user intent data, user intent data includes each user's flow node and node intent, a flow node corresponds to a node intent;According to user intent data constructs g×c two-dimensional matrix, wherein, g is total number of users, c is total number of node intent;Two-dimensional matrix is decomposed into first matrix and second matrix by non-negative matrix decomposition algorithm, first matrix is g×p matrix, second matrix is p×c matrix, and p is self-defined dimension;According to first matrix and second matrix, construct a number of group user portraits.The user portrait generated by the present application is highly matched with the intelligent customer service system, and the reply satisfaction of the intelligent customer service system can be improved.

The application discloses a kind of early identification methods of fire smoke image based on multi-modal fusion, comprising the following steps: S1, acquisition multi-modal data and carry out time alignment;S2, construct visible light branch, infrared branch and photoacoustic branch, generate multi-modal feature and store into shared feature pool;S3, embedding space-time tracking module in shared feature pool, generate global feature representation;S4, generate dynamic background template using low-rank matrix decomposition algorithm, separate and remove background noise;S5, by light self-attention mechanism in shared feature pool fusion multi-modal feature and compensate missing features;S6, by sparse feature enhancement algorithm enhances early smoke signal, and applies multi-task classification network to fire state classification and smoke diffusion path prediction;S7, by dynamic feedback mechanism updates shared feature pool and space-time tracking module.The application utilizes multi-modal fusion, space-time tracking and dynamic feedback method, realizes the early accurate identification of fire smoke.

The present invention discloses a prediction method for drug-disease association relationships based on graph-regularized matrix factorization. The method of the present invention extracts the semantic similarity between diseases according to the directed acyclic graph of each disease, and then combines the association relationships between each disease and drugs in the existing database. Using the graph convolution method, it extracts disease features to construct a disease feature matrix, determines the cosine similarity between diseases and fuses it with the semantic similarity. After obtaining the disease association relationship based on the directed acyclic graph, it constructs a drug feature matrix according to the drug features in the existing database, combines it with the disease feature matrix, establishes an association matrix between drug molecules and diseases, and performs eigen-decomposition on the association matrix based on the matrix factorization algorithm of graph regularization and kernel method. By constructing an objective function, it optimizes the neighbor relationships of nodes in the drug similarity network and the disease similarity network, fully utilizes disease features and drug features, and realizes the accurate prediction of the association relationship between drugs and diseases.

The invention discloses a method, device, medium and product for measuring noise of an aerial helicopter, and relates to the technical field of noise measurement. The method comprises: obtaining original noise observation signals of multiple sampling points collected by each unmanned aerial vehicle; converting the original noise observation signals of each sampling point into non-negative values ​​to obtain non-negative noise observation signals of the corresponding sampling point; using a non-negative matrix decomposition algorithm, based on the non-negative noise observation signals of each sampling point, determining the target spatiotemporal correlation matrix estimation of the corresponding sampling point; based on the non-negative noise observation signals of each sampling point and the target spatiotemporal correlation matrix estimation, determining the short-time Fourier transform coefficient vector of the noise of the aerial helicopter to be measured at the corresponding sampling point; based on the short-time Fourier transform coefficient vector of the noise of the aerial helicopter to be measured at each sampling point, determining the intensity of the noise of the aerial helicopter to be measured at the corresponding sampling point. The invention realizes the measurement of the noise of the aerial helicopter.

An electronic device according to an embodiment disclosed herein includes: a battery module including a plurality of cells; the sensor circuit is used for obtaining a state value of each battery cell in the plurality of battery cells; and a processor, where the processor is capable of: obtaining, via the sensor circuitry, a value related to a state of each of the plurality of cells as input data; obtaining feature data based on a feature matrix derived from the input data by using a specified matrix decomposition algorithm; and identifying whether each of the plurality of cells is abnormal based on the feature data. The feature matrix may be a matrix in which values of a specified number of main diagonal components among a plurality of diagonal components of a diagonal matrix for input data are replaced with 0.

The invention discloses a multi-scene single-frame large-view-field scattering imaging method based on polarization coding and spatial multiplexing, and the method comprises the steps: building a reflection-type scattering imaging system and a transmission-type scattering imaging system, and respectively measuring the OME ranges of the reflection-type scattering imaging system and the transmission-type scattering imaging system; multiple spatial multiplexing speckles are collected through single exposure of a polarization camera, and a linear multiplexing model is constructed; a signal identification method based on a minimum mean square error and an N-FINDR algorithm are introduced to realize target number identification and polarization specificity speckle initial value extraction; designing a Laplacian regularization term constrained Cozishi non-negative matrix factorization algorithm, and analyzing polarization specific speckles of different OME regions by using the extracted target number and initial value iterative optimization; and reconstructing a speckle result by using a phase recovery algorithm, and hiding the target after recovering the scattering medium. According to the method, the view field range is expanded by 4.5 times in a transmission-type / reflection-type system, the fidelity of a demultiplexing result reaches 32dB, and hidden scene reconstruction without continuous acquisition is realized.

The invention relates to a decentralized recommendation method with privacy protection performance, which is designed based on a matrix decomposition algorithm and a block chain and is used for providing privacy protection in a recommendation system. By introducing the block chain technology, the single-point fault risk in traditional centralized federated learning is eliminated, and the privacy protection capability is enhanced. In the model training period, a user locally trains a model containing private parameters and public parameters, but only shares the public parameters, so that leakage of original private data and the private parameters of the user is avoided, meanwhile, quantitative analysis is performed on privacy protection performance of the method, the total probability of privacy leakage is deduced, and the user experience is improved. And a theoretical basis is provided for evaluating the privacy protection intensity of the user.

The application discloses a multi-scene single-frame large-field-of-view scattering imaging method based on polarization encoding and spatial multiplexing, and comprises the following steps: a reflection type and a transmission type scattering imaging system are built, and the OME ranges of the two systems are measured; a linear multiplexing model is constructed by collecting multiple spatial multiplexed speckles through a single exposure of a polarization camera; a signal recognition method based on minimum mean square error and an N-FINDR algorithm are introduced to realize target number recognition and polarization-specific speckle initial value extraction; a Cauchy non-negative matrix decomposition algorithm with a graph Laplace regularization term is designed, and the extracted target number and initial value are iteratively optimized to analyze the polarization-specific speckles in different OME regions; and a phase recovery algorithm is used to reconstruct the speckle results, and hidden targets in a scattering medium are recovered. The method can expand the field of view range by 4.5 times in the transmission type / reflection type system, the fidelity of the demultiplexing result reaches 32 dB, and the hidden scene reconstruction is realized without continuous collection.

The application discloses a transcription factor regulation network modeling method based on single cell transcriptome data, comprising the following steps: S1, extracting multi-phenotype single cell omics data, performing data cleaning, and integrating the cleaned data; S2, based on a biological knowledge base, analyzing the data processed in S1, and constructing a prior inter-molecular interaction relationship network; S3, based on a multi-factor non-negative matrix factorization algorithm, establishing a multi-dimensional molecular synergistic interaction relationship module according to the data processed in S1 and the prior inter-molecular interaction relationship network in S2; S4, calculating the interaction relationship module related to the phenotype of the multi-dimensional molecular synergistic interaction relationship module; and S5, visualizing and exporting the interaction relationship between the multi-dimensional molecular synergistic interaction relationship module and the phenotype. With the aid of single cell sequencing technology and prior biological knowledge, the application extracts multi-dimensional molecular synergistic regulation relationship in high-dimensional single cell omics data, and obtains a transcription factor-gene function synergistic regulation interaction mechanism closely related to phenotype changes.

The invention belongs to the technical field of image processing, and particularly relates to a dietary tracking image processing method based on non-negative matrix blind source separation, which comprises the following steps: acquiring a dietary image, and preprocessing the dietary image; inputting the preprocessed image into a Unet network for food image classification; analyzing the classified food images by adopting a non-negative matrix factorization algorithm to obtain different types of food images; converting different types of food images into an HSV color space, and performing thresholding processing on the food images converted into the HSV color space; obtaining an independent food image according to the image after thresholding processing; calculating volumes of the individual food images; determining the nutrient intake of the user according to the volume of the food image and the type and number of the food image, and realizing diet tracking of the user; according to the method, spatial constraint non-negative matrix blind source separation is introduced, the local structure and spatial correlation of the image are reserved, stacked or semi-covered food materials with similar colors / textures are effectively distinguished, and the accuracy of component separation in a mixed scene is remarkably improved.

The invention discloses a gene multi-omics data fusion method in a modal deletion scene, and relates to the field of bioinformatics, and the method comprises the steps: carrying out the preprocessing and feature screening of each kind of omics data of a sample; adopting a non-negative matrix factorization algorithm to obtain common potential representations of the samples, and constructing hyperedges based on cosine similarity between the common potential representations; extracting omics data features as hypergraph nodes by using the prototype network layer and the linear layer; similar sample information is aggregated through a hypergraph information propagation mechanism, a missing mode is interpolated in a potential space, and an existing mode and auxiliary information are fused to enhance a representation learning effect; and inputting the interpolated and enhanced feature representation into a multi-modal learning module, and carrying out multi-modal data fusion and prediction result output. According to the deletion mode completion method based on the hypergraph neural network, the deep learning technology is combined with knowledge in the gene multi-omics field, and therefore the purposes of data noise reduction, deletion mode completion, efficient prediction and the like are achieved.

The application provides a quantum circuit system and encryption method based on AES encryption system, which comprises a byte substitution module, a row shift module and a column confusion module; the byte substitution module uses a pre-optimized S-box query circuit to realize nonlinear byte substitution; the row shift module completes the cyclic shift of the bytes in the row through a quantum bit exchange operation; the column confusion module realizes a matrix multiplication circuit based on finite field operation, wherein the quantum circuit of the column confusion module is generated by a depth-first matrix decomposition algorithm; the round key addition module uses a CNOT gate array to realize the XOR operation of the round key and the state, and the required key of the round key addition module is dynamically generated through a key expansion circuit. The timing scheduling and resource allocation of each module are optimized by using a compact Z-shaped architecture. Under the premise of keeping the quantum bit resource usage equivalent to the original Z-shaped architecture, the quantum circuit depth is significantly reduced, and higher fidelity and shorter execution time can be realized on NISQ devices.

The invention provides a large language model-guided military vocational education online learning achievement prediction matrix decomposition method, and relates to the technical field of data processing, and the method comprises a language model-guided online learning achievement prediction module which converts multi-modal auxiliary information into a natural language sentence format, and obtains a language model-guided online learning achievement prediction matrix; a large language model is used for performing embedding acquisition, an embedded vector is applied to a matrix decomposition algorithm to realize learning achievement prediction, and an online learning achievement prediction module guided by the language model comprises a linguistization step, a prompting step and a knowledge migration step which are executed in sequence. According to the method, multi-modal auxiliary information can be seamlessly integrated, the multi-modal auxiliary information is introduced into the work of online learning result prediction based on recommendation by utilizing the multi-modal auxiliary information in standard matrix decomposition through the help of a language model, and a data set is collected from a JZ online education platform; the invention aims to develop a machine learning algorithm in the field of military vocational education.

The present invention discloses a shrimp call recognition method based on periodic analysis non-negative matrix decomposition, and its field is the field of aquaculture technology. The present invention introduces a periodic analysis non-negative matrix decomposition algorithm with self-learning ability, decomposes the original acoustic data into a basis matrix and a coding matrix, extracts preliminary information about the time change of each sound source, can effectively separate natural environmental noise and biological calls, and has a strong adaptive ability to different signals, realizes unsupervised noise separation, improves the accuracy and timeliness of recognition, adapts to dynamic environmental noise, and provides strong data support for shrimp pond aquaculture management; has self-learning ability, can cluster different basis matrices by introducing sparse constrained non-negative matrix decomposition technology, effectively recognize and distinguish the vocal characteristics of shrimp, can meet the needs of shrimp pond aquaculture, and does not need to retrain the model when facing various types of biological sound separation in various environments. It is efficient, fast and accurate, and has a wide range of application scenarios.

The application discloses a high-capacity vehicle clustering method and system based on an information-physical dual network, which comprises the following steps: firstly, constructing a physical network topology graph corresponding to the current network of vehicles, and calculating a vehicle mobility correlation index; then, constructing a physical network topology graph corresponding to the current network of vehicles, and calculating a link transmission quality; further, superimposing the physical network and the information network by weighted summation, calculating a correlation density under a comprehensive network, and further establishing an optimal clustering model based on the correlation density; finally, using a matrix decomposition algorithm to approximately solve, and obtaining a final clustering result. In the calculation process, the channel fading under the LOS and NLOS environments is considered, and a channel link capacity calculation method is given through a matrix mother function. The method can reflect the stability of the physical topology, and can also reflect the link transmission quality of the whole channel, so that the VC construction with a more efficient transmission structure can be realized.