266 results about "Singular value decomposition method" patented technology

The invention relates to a fault diagnosis method of a rolling bearing under variable working conditions. The method comprises the steps of: extracting a characteristic vector of a signal by means of Hilbert-Huang conversion, carrying out EMD (Empirical Mode Decomposition) to a signal to obtain a plurality of IMFs (Intrinsic Mode Function), and selecting useful IMFs to carry out Hilbert conversion respectively to obtain analytic signals Hi(t) and taking envelopes respectively to form a characteristic vector w; then, carrying out singular value decomposition for w by a singular value decomposition method, and using a singular value matrix as the final fault characteristic vector; using the singular value matrixes in four modes of the rolling bearing as input and four matrixes corresponding to the four modes as output to train an Elman neural network. When the rolling bearing fails, whether the rolling bearing is faulted or not can be judged and direction of faults can be accurately positioned through analysis and judgment of a signal at t moment by the trained Elman neural network. According to the invention, classification of fault modes of the rolling bearing under variable working conditions can be effectively finished, and higher precision is maintained.

A compressed sensing magnetic resonance imaging reconstruction method relates to image processing. Provided is a compressed sensing magnetic resonance imaging reconstruction method which can improve the subjective visual effect of reconstructed images. The method comprises the following steps: a target function of MRI image reconstruction is constructed under a low-rank matrix formed based on non-local similar blocks; the problem of optimization solution of the target function is transformed into the problem of low-rank matrix de-noising and target function reconstruction solution by a variable replacing method according to the constructed target function; for the problem of low-rank matrix de-noising, the clustered low-rank matrix is decomposed by a singular value decomposition method, and soft threshold processing is carried out on feature values obtained through decomposition to obtain a de-noised low-rank matrix; and the obtained low-rank matrix is substituted into the target function, and a final reconstructed MR image is obtained by a least square algorithm after approximate optimization.

Method of automated inference using generalized incremental singular value decomposition by following the gradient of a cost function incorporating an error function of the implicit matrix defined by a current estimate of a partial SVD and the observed values. Accommodates sparse or overfilled matrices and correction for nonlinearities. Any number of observed values can be supplemented or updated, and predictions made at at any time. Any number of features can be used. Useful wherever there are two classes of objects (or the same class twice) for which there is a particular value implicitly associated with each pairing of such objects, for example including preference, filtering, rating, and recommender systems, image and other data recognition, compression, and restoration systems, warning systems, autonomous navigation, expert systems, machine learning, language processing, semantic analysis, artificial intelligence, knowledge systems, behavior prediction, economic and financial modeling, and modeling of natural systems.

The invention discloses an LCD foreign body defect detection method. The LCD foreign body defect detection method comprises the following steps of image resizing, namely, image resizing is carried out according to a bilinear interpolation algorithm, label detection, namely, the position of a label is detected and the label is eliminated to obtain an image OI, (3)image reconstruction, namely, the image OI is reconstructed through a singular value decomposition method to obtain a reconstructed image CI and the reconstructed image CI is subtracted from the image OI to obtain a difference image DI, (4), point-shaped foreign body defect detection, namely, thresholding defect cutting is conducted on the obtained difference image DI, outline number statistics is carried out through outline detection, total defect area is calculated, and whether point-shaped foreign body defects exist or not is judged, (5) region-shaped foreign body defect detection, namely, the difference image DI is divided into num windows, region foreign body defect detection is conducted on each window, and average gray value and variance statistics of each window is carried out, (6) line-shaped foreign body defect detection, namely, edge detection is conducted on the difference image DI according to an edge detection algorithm to judge whether line-shaped foreign body defects exist or not.

The invention discloses a method for improving the channel capacity of an indoor distributed multi-input and multi-output system, relates to the wireless communication field, and provides a method which can obtain the best performance of system capacity and improve the channel capacity of the indoor distributed multi-input and multi-output system. A transmitting end builds an angular domain module of the distributed multi-input and multi-output system based on angular domain information fed back by a receiving end; channel information matrix is reconstructed aimed at different indoor environments based on the built angular domain module of the distributed multi-input and multi-output system; transmitting antennas with large channel gains and larger arrival angle differences in indoor distributed antennas are selected to send data based on the channel information matrix; channel information matrix selected by the antennas is decomposed through a singular value decomposition method, and the selected transmitting antennas are subjected to power distribution through a power water filling algorithm; and signals to be sent of each user in each channel are sent by using the indoor distributed multi-input and multi-output system, so that the channel capacity is improved.

The invention relates to a tool abrasion state identification method based on multi-feature fusion. A plurality of methods are utilized for extracting feature information of vibration acceleration signals and cutting force signals, optimization is conducted on a feature set based on singular value decomposition, and the tool abrasion state identification precision is improved. According to the tool abrasion state identification method, mainly, the cutting force signals and the vibration signals are fused for feature extraction, the feature set is optimized through the singular value decomposition method, then, the feature set is input into a least squares support vector machine tool abrasion state identification model optimized on the basis of the genetic algorithm for identification, andthe tool abrasion state is output. According to the tool abrasion state identification method based on multi-feature fusion, the cutting force signals and the vibration acceleration signals are fusedfor feature extraction to serve as input, and the tool abrasion state identification precision is improved.

A small weak moving target tracking method based on sparse representation comprises the following steps: acquiring the location of an infrared image target based on a detection algorithm, and constructing an initial training sample and an initial particle set; adopting a K-means singular value decomposition method K_SVD to learn the training sample and construct an adaptive morphological ingredient over-complete dictionary of the image, then, constructing an adaptive online classification over-complete dictionary, and carrying out real-time online updating; and finally, establishing a small weak target sparse representation observation model in a particle filter tracking framework, estimating the location of the target based on the size of sparse representation residual of a particle target image block and a particle background image block in the adaptive online classification dictionary, and keeping stable target tracking in subsequent frames through repeated iteration. The method of the invention not only overcomes the defect that it is difficult for an offline structure dictionary to sparsely represent a dynamically changing image signal and improves the difference between a signal and a background in representation sparseness, but also effectively improves the capability of infrared weak small target motion detection tracking.

A method for recognizing a transformer partial discharge pattern based on a singular value decomposition (SVD) algorithm includes a training model and a classification recognizing process, comprising: firstly setting up an experimental environment having artificial defects, collecting at least one datum sample, and calculating statistical feature parameters of each datum sample to form a datum sample matrix; performing singular value decomposition on the datum sample matrix and determining an order of an optimal retention matrix by judging whether a feature of a retention matrix is clear, so as to obtain a type feature description matrix and a class-center description vector group after dimensionality reduction; preprocessing samples to be recognized to obtain a sample vector, and performing linear transformation on the sample vector utilizing a type space description matrix.

The invention relates to a human body behavior recognition method based on global characteristics and sparse representation classification. The method comprises the following steps: performing Gaussian kernel convolutional filtering preprocessing on a video frame, and extracting a moving foreground pixel by using a differential method; sampling a pixel value according to a time space dimension ofa parameter, determining a moving area, adjusting the size of the video frame, performing primary dimension reduction, splicing video frames in rows to form a vector group, and acquiring characteristic vectors; splicing the characteristic vectors in rows to form a characteristic matrix, performing secondary dimension reduction, calculating a primary characteristic dictionary of the characteristicmatrix, initializing the dictionary, after dictionary initialization, performing dictionary learning by using a class accordant K-time matrix singular value decomposition method, calculating an inputsignal sparse code according to the dictionary, inputting the code into a classifier, and outputting a behavior type; and counting dictionary learning parameters, and performing behavior recognition in real time. By adopting the method, dictionaries and linear classifiers with both reconstitution functions and classification functions are acquired, human body behavior recognition efficiency is improved, and the method is applicable to scientific fields such as security monitoring, video search based on contents and virtual reality.

A singular value decomposition method according to the present invention is a method for performing a singular value decomposition on an arbitrary matrix A using a computer, the method including the steps of: performing an upper bidiagonalization on the matrix A so as to obtain an upper bidiagonal matrix B of the matrix A; obtaining at least one singular value σ of the matrix B as singular values of the matrix A; and obtaining a singular vector of the matrix A for the σ. The step of obtaining a singular vector of the matrix A includes a step of performing a Twisted decomposition on a matrix BTB−σ2I (where I is a unit matrix) by using a Miura inverse transformation, an sdLVvs transformation, an rdLVvs transformation and a Miura transformation so as to diagonalize a matrix BTB.

The invention discloses a method of realizing high-resolution imaging of a forwarding looking radar. The method uses the least squares to address the problem that a matrix is not a square matrix of a radiation pattern, uses the truncated singular value decomposition method to suppress amplification of noise, resolves the estimation amount of an object scattering function, and realizes high-resolution imaging of forward looking scanning radar direction. According to the application, the method is fully aware of the problem of possible information loss in the presence of an object on an edge in traditional tsvd processing method which decomposes on the basis of a matrix truncated singular value and then resolves an inversion. According to the application, the method addresses the problem that a matrix is not a square matrix of a radiation pattern, and by using least squares, obviates the need for processing an original matrix, resolves an error minimum resolution by 2-nom minimization, saves edge information, and resolves the estimation amount of the object scattering function; the method analyzes and solves too loud noise in the process of resolving the estimation amount, and uses the truncated singular value decomposition method to suppress the amplification of the noise, such that the method realizes high-resolution imaging of the forward looking scanning radar.

The invention proposes an underwater 3D reconstruction device and method based on the multilayer refractive model of light field, belonging to the research field of computer vision, and the typical application is the high-precision 3D reconstruction in underwater. This device adopts multilayer refractive model of light field to calibrate binocular stereo vision. The center point correspondence algorithm of laser stripe adopts the underwater laser stripe matching algorithm based on the multilayer refractive model of light field and the singular value decomposition method based on the corresponding points under the multilayer refractive model of light field is adopted to calculate the 3D points. The whole system device is composed of two high-speed camera packaging storehouses and laser scanning computing unit packaging storehouses. The packaging storehouse adopts high-voltage resistance design. The high-speed camera packaging storehouse and the laser scanning system and the computing unit packaging storehouse are connected by high-voltage resistance tubes. The high-pressure resistant tube contains a high speed USB3.0 cables connected to the camera and the computing unit. The high speed camera package storehouse is symmetrically distributed at both ends of the device, and the laser scanning computing unit package storehouse is in the middle of the device. The invention can be applied to the high-precision three-dimensional reconstruction in the deep-water area, and the deepest depth can reach 4000 meters below water.

The invention discloses a multi-harmonic-source harmonic contribution quantitative analysis method based on a total-least square method. The multi-harmonic-source harmonic contribution quantitative analysis method includes: selecting harmonic voltage as the index for evaluating harmonic contribution, and determining the joint action of the harmonic source of each node; sectioning measured data according to time, and calculating the harmonic contribution in each time section according to the projection of the harmonic voltage of one harmonic source at one node on the total harmonic voltage at the node; using the total-least square method to solve harmonic impedance and the harmonic voltage, namely background harmonic voltage, of other harmonic sources at the node, and using a singular value decomposition method to obtain the solution of the total-least square method to obtain the quantitative data of each harmonic source at each node at each harmonic time. The multi-harmonic-source harmonic contribution quantitative analysis method has the advantages that the total-least square method is applied to the harmonic quantitative analysis field, the defect that a traditional method only considers variable harmonic voltage measuring errors and does not consider the independent variable harmonic current measuring errors is overcome, and obtained results are accurate.

The invention provides a sparse singular value decomposition scanning radar forward-looking imaging method, which is used for the radar imaging technology field. A radar forward-looking echo signal is analyzed and scanned, and an azimuth echo model of a target scattering coefficient and antenna measuring matrix operation is constructed, and then an azimuth high resolution problem is converted into a matrix inversion problem. A target function is constructed by adopting the sparse singular value decomposition method provided by the invention. The solution of the echo azimuth signal is realized by adopting an iterative strategy. A limitation of conventional TSVD methods of contradiction between resolution improvement and noise suppression is broken through, and sensitivity of an imaging result to noises is reduced, and therefore the azimuth resolution improvement is realized.

The technical proposal of the invention comprises a GRNN-based identification method. The method is used for realizing the following steps: an electrocardiogram signal sample data set including a plurality of periodic beats of a plurality of users is acquired, and wavelet transform is adopted to remove noise in the electrocardiogram signal; the denoised signals are segmented to construct the morphological features of ECG signals, and the training set beat feature databases and the test set beat feature databases are constructed respectively; singular value decomposition is used to remove redundant features from ECG signals; dimension reduction of the heart beat feature data set is carried out by linear discriminant analysis; the generalized regression neural network classifier is trained and the identity information of the individual is output according to the principle of multi-beat voting. The method has the advantages that the redundant information in the electrocardiogram signal iseffectively removed, the accuracy rate of the subsequent identification is improved, and the speed of the identification is greatly improved due to the reduction of the heart beat characteristic dimension by the LDA and the use of the GRNN neural network as a classifier.

The invention discloses a resource allocation method for a cognitive system based on space multiplexing. In the method, the F norm of a channel matrix is used to select a cognitive user so as to obtain a multi-user diversity gain of a cognitive network, a double-selection mode is used to lower the complexity of the algorithm, transmitted signals of the cognitive users are projected to a null space of an interference channel so as to avoid the interference of the cognitive user on an authorized user, the channel matrix of the cognitive user is converted into a concurrent independent channel by using a singular-value decomposition method, and the system capacity is enlarged by using a water-filling power allocation method. In a word, by using the multi-user diversity gain of the cognitive network, the method simplifies the user selection method and improves the system performance under the condition of acceptable computation complexity.

The invention relates to a structural load identification method based on regularization and singular value decomposition. The structural load identification method comprises the following steps: (1) an acceleration sensor collecting acceleration response time domain signals of actual working conditions of response points, and obtaining acceleration response frequency domain signals through Fourier transformation; (2) an exciter applying exciting action on various exciting points, the acceleration sensor collecting the acceleration response time domain signals of the response points at the same time, obtaining the acceleration response frequency domain signals and excitation frequency domain signals through the Fourier transformation, and calculating frequency-response functions between the exciting points and the response points; (3) calculating the conditional number of a multi-method matrix of the frequency-response functions; (4) judging whether the conditional number is larger than preset value, identifying identification load through a Tikhonove regularization method if the conditional number is larger than the preset value, and identifying identification load through a singular value decomposition method. Compared with the prior art, the structural load identification method has the advantages of being high in load identification precision, good in application performance, and the like.

The invention discloses a fault prediction method based on air data. The method comprises the following steps of: establishing an air database for storing air data; establishing a knowledge base for storing a rule for fault analysis; removing noise of the air data by using a singular value decomposition method; performing chaotic characteristic judgment on the air data from which the noise is removed by using a maximum Lyapunov index method; realizing prediction of the air data by using a prediction algorithm; and performing fault analysis according to a predicted value and knowledge in the knowledge base, and finally outputting a predicted result. Due to the adoption of the fault prediction method, the influence of noise on prediction accuracy is reduced, selection of a more suitable prediction model is facilitated, high prediction accuracy is achieved, and the serviceable range of the method is widened.

The invention is an edge- and texture- data synthesized pyramid image fusion method, firstly establishing a pyramid structure based on edge and texture features, considering the linear relations between binomial Gauss filter and texture extraction filter as well as edge extraction filter, adopting singular value decomposition to find the corresponding coefficient of the edge and texture images, then utilizing these features of each scale image to express each layer of the decomposed image, and finally make fusion by the fusion policy based on similarity measure and remarkability measure. It largely improves the quality of the image by a large margin and has an important significance and practical value in the follow-up treatment and image display of various application systems.

The embodiment of the invention discloses a method for providing a pre-encoding matrix, a method for providing a decoding matrix, and a base station. The method for providing the pre-encoding matrix comprises the following steps: the base station obtains an up channel matrix corresponding to each piece of user equipment; the base station combines the obtained up channel matrixes into a channel matrix; the base station decomposes the channel matrix by using a block diagonal singular value decomposition method to obtain a unitary matrix, wherein the diagonal line of the unitary matrix comprises block diagonal matrixes with the quantity same as that of the user equipment, and the block diagonal matrixes on the diagonal line correspond to the user equipment one by one; and the base station supplies each block diagonal matrix or one line of elements in each block diagonal matrix to the corresponding user equipment. By applying the technical scheme of the embodiment, the channel matrix is decomposed by using the block diagonal singular value decomposition method, and each piece of user equipment can respectively carry out pre-encoding processing or decoding processing by adopting the obtained diagonal matrix information, so that the performance of an MU-MIMO (multi-user multiple input multiple output) system can be improved.

The invention provides a method for modal analysis and structural impact monitoring based on compressed sampling and belongs to the technical field of structural health monitoring. According to the invention, a compressed sampling technology is used to collect impact load signals; then, modal analysis is conducted on compressed sampling data of the collected impact load signals by a singular value decomposition method, so that modal parameters of the structure are acquired; and finally, parameters including sizes and positions of impact loads are acquired through a signal reconstruction method based on the compressed sampling data. According to the invention, the problem about poor accuracy existing in structural modal parameter acquisition of a complicated structure based on modeling as well as the problem about overlarge data size of multiple sensors could be solved. The method has the advantages that a sensor network arranged on a to-be-monitored structure is used for impact monitoring; the structure can be monitored rapidly and accurately in real time; energy consumption is low; the size of stored and transmitted data is small; data loss is reduced during storage and transmission, etc.

The invention provides a robot hand-eye system calibration method, device and equipment and a storage medium. The method comprises the steps that center coordinates of a calibration part on a calibrator under a mechanical arm tool coordinate system are acquired; center coordinates of the calibration part under a scanner coordinate system are acquired; the center-of-gravity translation vector of the center coordinates of the calibration part under the mechanical arm tool coordinate system and the center-of-gravity translation vector of the center coordinates of the calibration part under the scanner coordinate system are calculated respectively; a transformation matrix between the scanner coordinate system and the mechanical arm tool coordinate system is calculated by means of the center-of-gravity translation vectors according to a singular value decomposition method; and according to the transformation matrix, each point coordinate in point cloud data is converted into a point coordinate under the mechanical arm tool coordinate system. According to the technical scheme, complex operation of camera pose calibration through a two-dimensional image is avoided, the imaging complexityof a robot hand-eye system during calibration is simplified, and the calibration precision and the working efficiency are improved.

The invention belongs to the technical field of damage identification and relates to a structure damage identification experimental method based on PVDF (polyvinylidene fluoride) piezoelectric-film sensors and strain modes. The damage identification experimental method comprises the following steps: collecting corresponding signals of all piezoelectric-film sensors after applying ambient excitation, and carrying out pre-treatment on the signals; obtaining an approximate impulse response matrix of reference points and answering points by using a natural excitation technique; constructing a generalized Hankel matrix by using the obtained impulse response matrix, and obtaining a system matrix with minimum order by using a singular value decomposition method; carrying out eigenvalue decomposition on the obtained system matrix so as to obtain structure modal parameters and modal shapes; respectively obtaining damaged or non-damaged structure modal parameters and determining damage positionsand carrying out damage identification. The method provided by the invention has the advantages that the error control is relatively good, the operability is strong, the sensitivity in damage identification can be improved, and modal analyzing and computing efficiency is improved.

The invention discloses a bearing fault diagnosis method based on variation modal decomposition and wavelet singular decomposition. The method includes: subjecting de-noising preprocessing on an acquired signal by using an optimal wavelet packet basis and a singular value decomposition method; after the signal is de-noised, extracting the intrinsic mode function of the signal by using a variationmodal decomposition algorithm; calculating the standard deviation of the obtained intrinsic mode function to construct a feature vector and using the obtained feature vector to diagnose a bearing fault.

The invention discloses an objective video quality evaluation method based on space domain and time domain structural similarities. Structural similarity between a distorted video and a reference video is evaluated from two aspects of the space domain and the time domain, space domain evaluation comprises luminance and chrominance structural similarities, the time domain evaluation is presented by using interframe residual structural similarities and calculation is simple and quick. A quaternion matrix is then used for building a structural similarity evaluation model. A singular value decomposition method is then adopted for solving the structural similarity quaternion matrix and the final objective quality evaluation value is obtained. The method has the advantages that complexity of calculation is low, and correlation between the objective evaluation result and subjective perception can be effectively improved.

The invention discloses a new image fusion method based on shift invariance shearlet transformation (SIST). The method includes the implementation steps of firstly, conducting multi-scale multi-direction decomposition on a source image through SIST to obtain low-frequency sub-band coefficients and high-frequency sub-band coefficients; secondly, reflecting the outline information of the image through low-frequency sub-bands, obtaining local structure descriptors for identifying the image resolution through a local structure tensor singular value decomposition method, making the local structure descriptors serve as movement measure functions in the fusion strategy, and adopting large strategy fusion. High-frequency sub-bands display detail information of the image. The invention provides a new edge strength measurement method, and a multi-strategy fusion rule based on the sigmoid function and edge strength measurement is made and used for high-frequency sub-band fusion. Finally, reverse SIST is conducted on the obtained fusion coefficients to obtain a final fusion image. The method overcomes the defect that edge distortion is easily caused by a traditional image fusion method, and more edge and detail information is reserved through the fused image.

The invention relates to a method for combined inversion of intercept and gradient based on compressional waves and converted shear waves. The method comprises the following steps of obtaining a compressional wave and converted shear wave angle domain gather, and meanwhile, obtaining the compressional wave and shear wave speeds and well logging interpretation result; extracting the sub-waves from the compressional wave and converted shear wave angle domain gather to obtain an angle domain reflection coefficient profile, and utilizing the obtained compressional wave and shear wave speeds to solve a shear wave and compressional wave ratio profile; utilizing the induced converted shear wave reflection coefficient formula and the compressional wave reflection parameter approximate formula using the intercept and gradient as parameters, and adopting a singular value decomposition method to realize combined inversion so as to obtain the intercept and gradient; utilizing the well logging interpretation result to calibrate the inverted intercept and gradient, establishing a fluid identification mode, and carrying out the fluid detection on the undrilled area. The method has the advantages that the received multi-component earthquake wave field information is sufficiently utilized, the multiple-solution property of the inversion process is reduced, the stability of the inversion process is improved, and a reliable method is provided for the oil and gas prediction.

The invention relates to the field of digital image processing technology, and provides a method for image classification through supplementing a text characteristic for settling a problem of classification result deterioration caused by image characteristic discriminating property reduction because of text characteristic information loss in a bag of visual words. The image classification method based on the supplemented text characteristic comprises the following steps of (1), respectively extracting an SIFT characteristic, a color characteristic, and a shape-and-boundary characteristic of an image in a data set; (2), randomly selecting 200000 characteristics from the supplemented text characteristics and obtaining a dictionary B through a K-singular value decomposition method (K-SVD); (3), obtaining a sparse characteristic vector C through the studied dictionary B; (4), performing spatial maximal value gathering on the sparse characteristic; and (5), transmitting an image characteristic vector F into a linear SVM classifier, thereby performing training and testing respectively for obtaining an image classification result. The image classification method is mainly applied for digital image processing.

The invention discloses a navigation registration method in the brain neurosurgery department based on a thermodynamic diagram and facial key points, and belongs to the technical field of digital medical treatment. According to the method, an n-layer piling sand glass nerve network for extracting the key points based on the thermodynamic diagram is trained by utilizing human face key point data; then a CT three-dimensional image or MRI three-dimensional image of the head of the patient is acquired, and a facial three-dimensional grid key point cloud corresponding to the image is acquired and serves as a target point cloud; a binocular camera is utilized for capturing facial images of the patient in real time in an operation, and a binocular facial key point cloud of the patient is obtainedand serves as a source point cloud; finally, a singular value decomposition method is utilized for registration for the target point cloud and the source point cloud. The method is convenient to implement, low in cost, accurate in registration result and saves time, and additional wounds cannot be caused to the patient.

The invention discloses a signal separation method based on particle swarm optimization. The method includes the steps: (1) inputting observation signals; (2) building an independent component analysis algorithm optimization model by taking a difference value between simple products of a minimum separated signal joint probability and a marginal probability as an optimization target; (3) estimating the number of source signals according to a singular value decomposition method and determining the number of optimization variables according to the number of the source signals; (4) calculating correlation coefficients of the observation signals, and determining the value range of the optimization variables; (5) optimizing separation matrixes by the aid of a particle swarm optimization algorithm; (6) taking particles with optimal fitness in the last generation population after optimization as an optimal separation matrix and multiplying the optimal separation matrix with the mixed signals to obtain the optimal separation signals. The method based on independent component analysis of the particle swarm optimization has universal applicability and effectively solves various blind source separation problems.