121 results about "Tikhonov regularization" patented technology

The present invention includes a method for parallel magnetic resonance imaging termed Superresolution Sensitivity Encoding (SURE-SENSE) and its application to functional and spectroscopic magnetic resonance imaging. SURE-SENSE acceleration is performed by acquiring only the central region of k-space instead of increasing the sampling distance over the complete k-space matrix and reconstruction is explicitly based on intra-voxel coil sensitivity variation. SURE-SENSE image reconstruction is formulated as a superresolution imaging problem where a collection of low resolution images acquired with multiple receiver coils are combined into a single image with higher spatial resolution using coil sensitivity maps acquired with high spatial resolution. The effective acceleration of conventional gradient encoding is given by the gain in spatial resolution Since SURE-SENSE is an ill-posed inverse problem, Tikhonov regularization is employed to control noise amplification. Unlike standard SENSE, SURE-SENSE allows acceleration along all encoding directions.

The invention discloses a random recognition method for bridge floor moving vehicle loads. The method comprises the following steps that (1) raster sensors, strain sensors and acceleration sensors are arranged on a bridge floor to obtain running speeds of moving vehicles and the dynamic strain and acceleration responses of a bridge under the action of the moving vehicle loads; (2) an accurate structural finite element model is established for a bridge structure, model condensation is carried out, and the freedom degree of the bridge model after the model condensation is made to be matched with dynamic strain signal measuring positions; (3) the deterministic components of bridge random dynamic displacement, speed and acceleration are obtained by utilizing a KL decomposition technique on the basis of actual measurement dynamic strain response samples; (4) the deterministic components random moving vehicle loads on the bridge floor are recognized by directly adopting a sample matrix inversion and a Tikhonov regularization method; (5) the statistical properties, average time history and variance, of the random moving vehicle loads are obtained.

The invention relates to the technical field of synthetic aperture radar interferometry, and specifically relates to a PSInSAR estimation method applicable to complex urban area infrastructure in windy and rainy conditions. The method includes following steps: performing data preprocessing on InSAR data; establishing a Delaunay triangulation network and an adaptive encryption network; estimating an arc segment relative parameter through a robust estimator; performing Tikhonov regularization adjustment operation; establishing a local star network; recognizing single PS points and layover PS points; calculating the single PS points in remaining pixels through the robust estimator, and detecting the layover PS points through a compressed sensing algorithm; and outputting heights and deformation results of the single PS points and the layover PS points. According to the method, combined calculation of the single PS points and the layover PS points is realized through mixed network establishment without removing the atmosphere in a global manner, a temperature non-linear deformation model is introduced, the fitting degree of the time sequence phase is improved, accurate estimation of the PS points is realized by employing the robust estimator, and the layover PS points of the complex urban environment are extracted by employing super-resolution MD-TomoSAR imaging.

The invention discloses a pattern recognition method of a supporting vector machine for fault diagnosis of a device, in particular to a device failure pattern recognition method of a reduction Tikhonov regularization supporting vector machine auto-selected by utilizing parameters. The invention comprises the following steps: (1) achieving a derivation process of Tikhonov regularization supporting models of the vector machine; (2) constructing a subsample set to reject redundant sample information by utilizing a pruning method so as to build models of the reduction Tikhonov regularization supporting vector machine; (3) taking classification accuracy as a fitness function, auto-selecting width parameters and balance parameters of a Gauss kernel function of the reduction Tikhonov regularization supporting vector machine by utilizing a heredity algorithm and by taking classification accuracy as a fitness function so as to build the models of the reduction Tikhonov regularization supporting vector machine auto-selected by utilizing parameters; and (4) verifying the pattern recognition method by utilizing failure samples of a motor device so as to indicate superiority of the method proposed in the invention.

The invention relates to a method of parallel imaging reconstruction in parallel magnetic resonance imaging reconstruction. Magnetic resonance data is acquired in parallel by an array of separate RF receiver coils. A reconstruction method based on Tikhonov regularization is presented to reduce the SNR loss due to geometric correlations in the spatial information from the array coil elements. In order to reduce the noise amplification of the reconstruction so-called “g-factor”, reference scans are utilized as a priori information of the final reconstructed image to provide regularized estimates for the reconstruction using the L-curve technique. According to the invention the method with the proposed L-curve approach was fully automatic and showed a significant reduction in average g-factors in the experimental_images.

The invention discloses a high-precision near-field acoustic holography algorithm adopting a weighted iteration equivalent source method, which is characterized in that a holographic plane H is arranged in a sound source near-field radiation area, and sound pressure PH on the holographic plane H is measured; an equivalent source plane Se is arranged at the side, which is away from the holographic plane H, of an object reconstruction plane T, and equivalent sources are arranged on the equivalent source plane Se; a relation between the sound pressure PH and each equivalent source is established by using a sound pressure transfer matrix between the equivalent source and the holographic plane H; and the source intensity Q of each equivalent source is solved by adopting a new iterative regularization algorithm with a posteriori weighted norm constraint penalty term, and then sound field data on the object reconstruction plane T is calculated by using the solved source intensity Q and the transfer matrix between the equivalent source and the object reconstruction plane T. According to the invention, the source intensity of each equivalent source is precisely solved by using the new iterative regularization algorithm with the posteriori weighted norm constraint penalty term, thereby avoiding source intensity energy leakage caused by a 2-norm penalty term in the Tikhonov regularization process. Compared with conventional equivalent source based near-field acoustic holography, a calculation result acquired by the method disclosed by the invention is more accurate.

The invention discloses a time domain identification method of random dynamic loads. The time domain identification method is characterized by comprising the following steps that (1) a strain sensor and an acceleration sensor are arranged on an engineering structure, and dynamic strain samples and acceleration response samples of the structure are obtained through measurement; (2) an accurate structure finite element model is established for the engineering structure, model condensation is conducted, and a condensed finite element model matched with the freedom degree of a dynamic strain response signal obtained through actual measurement is obtained; (3) on the basis of a dynamic strain sample set obtained through actual measurement, the deterministic component of structure random dynamic response is obtained through the KL decomposition technology; (4) deterministic components of the random dynamic loads are identified by the adoption of the direct inversion method and the Tikhonov regularization method; (5) the statistic characteristics, namely, the average time history and the variance, of the random dynamic loads are obtained.

The present invention discloses a sparse representation method for dynamic load identification of a mechanical structure for solving ill-posed natures of the dynamic load identification inverse problem and overcoming the disadvantage that the number of primary functions is required to be determined in advance by a current function approximation method. The sparse representation method comprises the following steps of (1) measuring a frequency response function between an action point of dynamic load of the mechanical structure and a response point of the mechanical structure by using a hammering method, and obtaining a transfer matrix through processing; (2) measuring a response signal generated by the dynamic load acting on the mechanical structure; (3) selecting primary functions to construct a sparse representation dictionary according to morphology of the dynamic load; (4) constructing an L1-norm-based sparse representation model of the dynamic load identification; (5) solving the sparse representation model of the dynamic load identification, and obtaining dynamic load sparse representation coefficient vectors; and (6) obtaining the identified dynamic load. Impact and harmonic load acting on the mechanical structure can be effectively identified, and compared with a traditional Tikhonov regularization method based on L2 norm, the sparse representation method has the advantages that identification precision is high and stability is high.

The invention discloses a structural damage identification method based on ridge estimation and an L curve method. The structural damage identification method comprises the steps of firstly, building a structural damage equation based on sensitivity of modal strain energy according to the sensitivity of the structural modal strain energy, then determining a basic estimation criterion of the ridge estimation according to the principle of Tikhonov regularization, then utilizing the L curve method for determining the optimal ridge parameter of the ridge estimation, working out the basic solution of the structural damage equation on the basis of the optimal ridge parameter, and finally correcting the result, so that the damage coefficient of a damage unit is obtained, and the structural damage is identified. According to the structural damage identification method based on the ridge estimation and the L curve method, by the adoption of the ridge estimation, the influence of an ill-conditioned problem on damage identification can be greatly reduced, the quasi-optimal ridge parameter is determined by the adoption of the L curve method, meanwhile, a correcting strategy for the ridge estimation is provided, and therefore the better damage estimation value is obtained.

The invention discloses a facial point detection system based on multi-task regularization and a layer-by-layer supervision neural network. The system comprises a multi-task regularization module and a layer-by-layer supervision network module. The multi-task regularization module includes a main task and a related task; and the main task and the related task study jointly to obtain a common feature space and then an additional regular term is provided by using an auxiliary tag of the related task to enhance a generalization ability of a network. The layer-by-layer supervision network module, different from the traditional convolution neural network only optimizing an objective function of an output layer, introduces a supervision objective function into each interlayer, thereby enhancing the saliency of features obtained by studying of the interlayers. Therefore, problems that overfitting occurs and the feature robustness is uncertain according to the traditional convolution neural network can be solved effectively.

The invention discloses an improved method for extracting Fourier transformation infrared spectrum characteristic variable of multi-component gas by aid of TR (Tikhonov regularization). The method includes resolving a characteristic variable extracting model into weight sum of the difference of multiple spectral line values; converting an original TR objective function into an objective function based on the model; then adding a bound term of the difference of spectral line positions in the objective function based on the model; realizing optimal functional solution by means of an LASSO (least absolute shrinkage and selection operator) arithmetic based on an Engl's criterion so as to obtain the optimal value of a regressive vector; and obtaining the characteristic variable capable of overcoming interferences caused by spectrum baseline deviation. The accuracy of online multi-component gas analysis results can be improved by the aid of the improved method. The improved method for extracting Fourier transformation infrared spectrum characteristic variable of multi-component gas by aid of TR can be used for multi-component gas quantitative spectrometric analysis application in the fields of gas logging for petroleum and natural gas exploration, quality control and fault diagnosis of products, hardware, chemical engineering and environmental protection.

The present invention discloses a robust visual classification method and system and aims to effectively achieve category prediction of a no-label sample in a training sample and rapid induction and reasonable dimension reduction of a to-be-detected sample. The method comprises: integrating an error metric based on elastic regression analysis into a label propagation model outside the training sample; by a parameter, weighing the influence of a normalization manifold regularization term, a label fitting term based on soft label l2, 1 norm regularization and an elastic regression residual term based on l2, 1 norm regularization on sample description and category identification so as to complete establishing a label propagation model; and then, iteratively optimizing the label propagation model to acquire a projection matrix for determining a category of a to-be-detected sample. Therefore, according to the robust visual classification method and system, by introducing a regression error term based on l2, 1 norm regularization and soft label l2, 1 norm regularization, robustness of the system can be effectively improved while the advantages of a label propagation classification method is carried on, so that the induction process of the to-be-detected sample is rapid and accurate.

The invention relates to a method of parallel imaging reconstruction in parallel magnetic resonance imaging reconstruction. Magnetic resonance data is acquired in parallel by an array of separate RF receiver coils. A reconstruction method based on Tikhonov regularization is presented to reduce the SNR loss due to geometric correlations in the spatial information from the array coil elements. In order to reduce the noise amplification of the reconstruction so-called “g-factor”, reference scans are utilized as a priori information of the final reconstructed image to provide regularized estimates for the reconstruction using the L-curve technique. According to the invention the method with the proposed L-curve approach was fully automatic and showed a significant reduction in average g-factors in the experimental images.

The invention discloses a bridge moving load and damage synergetic recognition method based on L1 / 2 regularization. The method includes the steps that a plurality of structural response measurement sensors such as an acceleration sensor, a displacement sensor and a strain sensor are arranged on a bridge according to the aim and accuracy requirement of bridge structure health monitoring; a finite element model of the bridge is established through beam elements; moving loads are equivalent to a linear combination of rectangular loads, and bridge damage is equivalent to element stiffness reduction; the moving loads are recognized through a Tikhonov regularization method under the condition that an element damage factor is given, a reconstructed response is calculated with the moving loads, the range of the residual L2 norm of a measured response and the reconstructed response is obtained through comparison, and an L1 / 2 penalty term is introduced to establish an objective function with the element damage factor as an optimized parameter; a constraint optimization problem is solved with a Matlab optimization toolbox, the corresponding damage factor is the structural damage recognition result when the objective function is the minimum, and correspondingly the moving loads worked out through Tikhonov regularization solving are the moving load recognition result.

The invention relates to a sparse deconvolution method for impact load identification of a mechanical structure. The method is used for solving the ill-posed nature of the impact load identification inverse problem and comprises steps as follows: 1) a frequency response function between an impact load acting point and a response measurement point of the mechanical structure is measured with a hammering method, a unit impulse response function is obtained through inverse fast fourier transform, discretization is further performed, and a transfer matrix is obtained; 2) an acceleration signal generated by impact load of the mechanical structure is measured with an acceleration sensor; 3) an L1-norm-based sparse deconvolution convex optimization model for impact load identification is established; 4) the sparse deconvolution optimization model is resolved with a primal-dual interior point method, and a sparse deconvolution solution, namely, a to-be-identified impact load, is obtained. The sparse deconvolution method is suitable for identifying the impact load acting on the mechanical structure. Compared with conventional Tikhonov regularization methods based on an L2 norm, the sparse deconvolution method has the advantages of high identification accuracy, high computation efficiency and high stability.

The invention discloses a heart electrical function imaging method based on beating heart model. The method comprises following steps: obtaining the position displacement change of heart based on double-ventricle electricity-coupled heart model; respectively constructing transfer matrixes TBH between epicardium and body surface potential of beating heart at different moments by adopting boundary element method; obtaining the body surface potential distribution PhiB of the body at different moments, calculating the potential distribution of the epicardium, and achieving heart electrical function imaging; and calculating the potential distribution of the epicardium by adopting normal TiKhonov regularization method. The heart electrical function imaging method based on beating heart model can effectively suppress the geometrical error caused by static heart model to obtain more accurate potential distribution imaging of the epicardium and accurately detect the electrical movement information of the heart, thus providing meaningful clinical information for the prevention and the treatment of heart diseases.

The invention discloses a method using an improved regularization method to restrain difference global positioning system (DGPS) integer ambiguity ill-condition. The method using the improved regularization method to restrain the DGPS integer ambiguity ill-condition comprises the steps of (1), collecting observation data of a global positioning system (GPS) carrier phase, building a DGPS carrier phase double-difference observation equation; (2) obtaining a floating point solution of DGPS integer ambiguity and a corresponding variance-covariance matrix based on a least square method according to the DGPS carrier phase double-difference observation equation; (3) and using a two-step solution to build a regularization matrix in a Tikhonov regularization algorithm, obtaining a corresponding regularization parameter according to a DFP quasi-newton method, processing the variance-covariance matrix by using the obtained Tikhonov regularization algorithm, restraining ill-condition of the DGPS integer ambiguity, and at last obtaining accurate integer ambiguity. According to the method using the improved regularization method to restrain the DGPS integer ambiguity ill-condition, the improved Tikhonov regularization algorithm is used for restraining the problem of the ill-condition in the DGPS integer ambiguity, the method using the improved regularization method to restrain the DGPS integer ambiguity is beneficial for obtaining the accurate integer ambiguity, and high-accuracy positioning and attitude measurement of DGPS can be achieved.

The invention discloses a total-variation regularization and variable splitting-based lens-less imaging rapid reconstruction method. Aiming at a thought that image reconstruction problems of lens-lessimaging systems adopt total-variation regularization and variable splitting, a to-be-solved target function is split to two sub-problems, and the sub-problems are alternately solved to obtain a finalresult. The method comprises the following steps of: firstly importing a total-variation regularization image reconstruction model according to a linear imaging mechanism in lens-less imaging; importing an auxiliary variable and splitting the to-be-solved target function into the two sub-problems by using a variable splitting method; solving the two sub-problems by using Tikhonov regularization and anisotropic total-variation (TV) regularization; and finally alternately solving the two sub-problems so as to obtain an optimum solution. According to the method, lens-less image reconstruction can be stably carried out under unsatisfactory factors, noises can be effectively removed, and detailed information such as edges and the like of reconstructed images can be kept at the same time.

The invention discloses a frequency domain load identification method based on Tikhonov regularization. The method aims at solving the problems existing in frequency response function acquisition in practical engineering application of a matrix inversion method and the morbidity problems existing in calculation. A frequency response function is acquired through a finite element method, the morbidity of mechanical equipment is evaluated through the number of matrix conditions of the frequency response function, and loads are identified based on different regularization methods when the morbidity is different. A proper mobility condition number threshold value is determined. The method has the advantages that the problem that the frequency response function of the mechanical equipment is difficult to acquire in practical engineering application of the matrix inversion method and the morbidity problems existing in calculation can be solved, the load identification accuracy in a frequency domain is improved, and high engineering application value is achieved.

The invention discloses a ventricular premature beat abnormal activation site positioning method based on ECGI (electrocardiographic imaging). According to the method, by establishing a combined collection framework of 64 lead body surface potential data and computed tomography imaging, a personalized heart-trunk model is obtained; by Tikhonov regularization, the ECG (Electrocardiograph) inverse problem is solved; an epicardial potential is reconstructed so as to accurately position a ventricular premature beat abnormal activation site. The ventricular premature beat abnormal activation site positioning method has important actual application value.

An embodiment of a method for solving the inverse problem of electrophysiology and determining a voltage distribution on a surface of a tissue may comprise receiving a plurality of voltages collected by a plurality of electrodes adjacent to the surface, discretizing the problem using a Finite Element Method (FEM) or a Boundary Element Method (BEM), introducing one or more regularization terms to an error minimization formulation, and solving, by a processor, the voltage distribution according to the plurality of voltages and according to the regularization terms. The regularization terms may comprise one or more of a Laplacian smoothness operator, a Tikhonov regularization matrix, a confidence matrix, and a linear operator that interpolates the plurality of electrode voltages to the tissue voltage distribution.

The invention relates to a radio frequency tomography method base on background learning. The radio frequency tomography method base on the background learning comprises the following steps: 1, establishing the distribution model of the received signal strength value of each link and judging whether each link is influenced according to the received signal strength of a wireless sensor network by using a mixture Gaussian background learning algorithm or a kernel density estimation background learning algorithm; 2, carrying out image reconstruction according to the distribution model of the received signal strength value of each link by using Tikhonov regularization. The mixture Gaussian background learning algorithm or the kernel density estimation background learning algorithm is applied to radio frequency tomography to estimate the distribution of the RSS measured values of each link, and a multiple target detecting and tracing function is achieved. The radio frequency tomography method base on the background learning has the advantages that higher accuracy and effectiveness can be obtained in multiple targets and time varying environments, and an offline training process is not needed.

The present invention discloses a design method of a biplane magnetic resonance imaging system gradient coil. The method comprises the steps of: performing three-dimensional continuous triangular meshdivision for a gradient coil area, according to a boundary element method, calculating a coil inductance matrix generated by the gradient coil area mesh node for a target field point, calculating thecurrent and the direction of the gradient coil plane through a coil target gradient value, an inductance constraint condition and a Tikhonov Regularization, and finally employing a stream function method to obtain actual winding distribution of the gradient coil. The design method of a biplane magnetic resonance imaging system gradient coil solves the problem that the coil inductance is large inthe gradient coil.

The invention relates to a regularization least square subspace crossing target direction finding method. According to the method, the Tikhonov regularization method and the least square subspace crossing algorithm are combined to process received signals of a shallow-sea sonar apparatus. On a basis of constructing a least square problem through the utilization of the subspace crossing principle, and then through the regularization method, optimal regularization factors are determined, so that detection and accurate positioning of shallow-sea targets are finally realized. According to the regularization least square subspace crossing target direction finding method of the invention, the Tikhonov regularization method and the least square subspace crossing algorithm are combined for form a new direction finding algorithm. Compared with routine wave beam formation algorithms, the method is better in direction resolution performance and higher in accuracy degree, and the value calculating is stable.

The invention relates to a visual scene graph generation technology, and discloses a visual scene graph generation system and method based on relation regularization, which can quickly and effectivelyjudge whether a relation exists between objects, and is beneficial to enhancing the detection effect of an object detection model. The system comprises an object detector, an object label refiner andan object relation generator. An object in the image is detected through an object detector to botain a label, an object frame feature and a joint frame feature of the object; And the labels of the objects is refined by using an object label refiner, the relationship between the objects is acquired by using an object relationship generator, and generating a final visual scene graph. The system and methods are suitable for generating the visual scene graph.

The invention belongs to the field of underwater detection and particularly relates to a method for locating identification and acoustic radiation prediction of an underwater complex-structure radiation noise source on the basis of a unit radiation approach. The method comprises the following steps of establishing a vibration sound transmission matrix, wherein a sound transmission modeling approach based on a unit radiation superposition approach is used, a radiation sound field of a piston on the surface of a regular baffle is adopted for simulating a radiation sound field of a piston on thesurface of an actual baffle, and a sound transmission matrix G of the surface of a target structure at a normal vibration velocity to the actual radiation sound field is obtained according to an analytical expression of the radiation sound field of the piston on the surface of the regular baffle; obtaining sound source distribution of the target structure, wherein a vibration sound transmission regularization matrix is adopted for solving an ill-posed problem of a noise source positioning recognition algorithm, an array is adopted for receiving data and a vibration sound transmission matrix ofvibration from the structural surface to a measurement matrix, and the vibration velocity distribution on the surface of a sound source of the target structure is obtained in combination with a Tikhonov regularization approach. The method overcomes the limitation of traditional point source spherical expansion sound transmission models and has higher accuracy and a wider application range.

The invention relates to a transformer station grounding network corrosion diagnosis method based on magnetic field inverse problem solving, which comprises the following steps: calculating a theoretical ground surface magnetic induction strength of a transformer station grounding network according to a number calculation method; inputting an AC excitation current into the transformer station grounding network, sensing a ground surface magnetic field by a detecting coil along the upper part of the grounding network, and acquiring an induction voltage signal by a data acquisition device; calculating magnetic induction strengths at n measuring points according to the acquired induction voltage signal; according to the magnetic induction strengths at the n measuring points, calculating the axial current of n conduction segments of the grounding network by means of a Tikhonov regularization algorithm; obtaining an actual calculation result for the ground surface magnetic induction strength of the grounding network; and comparing the actual ground surface magnetic induction strength of the grounding network with the ground surface magnetic induction strength which is obtained through calculation under a normal working condition. Compared with the prior art, the transformer station grounding network corrosion diagnosis method has advantages such as improving operation reliability of the transformer station grounding network.

The invention discloses a dynamic brain functional magnetic resonance imaging method and system. The method includes the steps that high-time low-space-resolution navigation data S<NAV> and high-space low-time-resolution dynamic image data S are collected in (k,t) space; the navigation data S<NAV> are decomposed through singular values to obtain a model order L and a time primary function phi[l](t), and Tikhonov regularization constraint is conducted by combining with the image data S to solve a space primary function cl(k); interpolation recovery with the high-time high-space coverage density is conducted on the (k,t) space through the model order L, the time primary function phi[l](t) and the space primary function cl(k), Fourier inversion is then conducted, and therefore a high-time-resolution and high-space-resolution dynamic brain function magnetic resonance image is obtained. By means of the method and the system, quality of the reconstructed image is improved, and meanwhile the high-time-resolution and high-space-resolution dynamic brain function magnetic resonance image is obtained.

The invention provides an improved L-curve electrical tomography reconstruction method based on curvature computing. The method is suitable for bubbly flow tomography, and includes the following steps that according to a tested scene domain, a relative boundary measured value vector b and a sensitivity matrix A which are required by reconstruction are obtained; by means of Tikhonov regularization, an L-curve is computed and drawn; whether local inflection points exist in the L-curve is judged; if the local inflection points do not exist, a regularization coefficient optimally selected is determined through an L-curve method; if the local inflection points exist, the regularization coefficient optimally selected is determined through an improved L-curve method, wherein the improved L-curve method includes the steps that the local inflection points of the curve are determined by calculating a second peak value of the curvature of the L-curve, regularization coefficients corresponding to the local inflection points serve as optimally selected coefficients, the optimally selected regularization coefficients are substituted into Tikhonov regularization for solving of an image reconstruction inverse problem, and imaging is carried out. The improved L-curve electrical tomography reconstruction method facilitates precise solving of an electrical tomography imaging inverse problem, and improves image reconstruction quality.

The invention discloses a flight disturbing magnetic field compensation method based on Tikhonov regularization. According to the method, by conducting singular value decomposition on a coefficient solution matrix K and considering the existence of random disturbing magnetic fields unrelated to changes of flight postures in the real environment, the method, compared with ridge estimation, has theadvantages that more specific information of the coefficient solution matrix is obtained, the coefficient solution precision is higher, and the compensation effect is better.