33 results about "Augmented matrix" patented technology

The invention discloses an intelligent manufacturing collaborative decision-making method based on digital twinning. The method comprises the following steps: establishing a global digital twinning model and a local digital twinning model according to an intelligent manufacturing process; the method comprises the following steps of: acquiring production data in an intelligent manufacturing process in real time; determining a regulation and control index data set under sampling times; constructing a time sequence augmented matrix; constructing a dynamic weight matrix by utilizing a Pearson correlation coefficient; carrying out dimensionality reduction on the weighted time sequence augmentation matrix by utilizing the rank of the corrected singular value matrix; and standardizing the dimensionality-reduced weighted time sequence augmentation matrix, dividing the standardized matrix by using a dynamic sliding window, inputting the divided matrix into a double-branch attention network in sequence, and predicting a production decision scheme in combination with a current actual production decision scheme. Through cooperation of local self-cooperation and global scheduling, the response speed of intelligent manufacturing is improved.

The invention provides a TOF (Time of Flight) estimation method of a matrix bundle based on partitioning and random SVD (Singular Value Decomposition). The method comprises the following steps: firstly, acquiring channel state information (CSI) by using a commercial WiFi device, and constructing a CSI matrix; and secondly, performing block overlapping processing on the CSI matrix, constructing an augmented matrix for each CSI data block, and performing SVD decomposition on the augmented matrix, thereby reducing the calculation complexity of SVD decomposition. And then, the time of flight (TOF) of each CSI data block is estimated by using a matrix pencil algorithm, and the minimum TOF is direct path estimation. According to the method designed by the invention, the robustness of the system and the estimation precision of the TOF are improved under the condition that the algorithm complexity is greatly reduced.

The invention relates to the technical field of pet food spectrum detection, and discloses a pet food spectrum detection device and a detection method. Sampling at equal intervals in a preset wavelength range, and eliminating noise interference through dark current and standard whiteboard calibration; taking a natural logarithm of the reflectivity to balance a dynamic range; constructing a linear spline basis in an equidistant and segmented manner, and adaptively calculating a regularization coefficient according to the energy ratio of the signal to the basis function; constructing an augmented matrix according to a basis function and a signal inner product, and solving a spline coefficient through Gaussian elimination; reconstructing a spectrum by using the coefficient and quantifying a residual error; establishing a reference model based on the coefficient mean value and the standard deviation of the multiple qualified samples; comparing the Euclidean distance between the to-be-detected sample coefficient and the model mean value with a threshold value, and generating a residual error, a distance, a threshold value and a judgment report; empirical parameters or manual adjustment and optimization are not needed in the whole process, online adaptive detection in batches can be achieved, and noise correction, feature enhancement, over-fitting suppression and traceable rapid and accurate detection are achieved.

The present application relates to the field of artificial intelligence. Provided in the present application are a method and system for implementing a self-attention mechanism having a linear complexity, and a device and a medium. The method comprises: performing positional encoding on sequence data to be processed, and performing feature mapping to apply the positional encoding to a query matrix and a key matrix; on the basis of the query matrix and the key matrix, performing calculation to obtain a low-rank query matrix, a low-rank key matrix and a low-rank generalized inverse matrix; augmenting a value matrix to obtain an augmented matrix, and on the basis of the augmented matrix, the low-rank key matrix, the low-rank generalized inverse matrix and the low-rank query matrix, obtaining a sequence self-attention matrix; and transposing the sequence self-attention matrix, using the transposed matrix as sequence data to repeat the foregoing steps, using the newly obtained sequence self-attention matrix as a feature self-attention matrix, and on the basis of the sequence self-attention matrix and the feature self-attention matrix, obtaining a self-attention matrix having a linear complexity. The present application implements linear self-attention, better maintains the model performance and has relatively good expansibility.

The application discloses an intelligent manufacturing collaborative decision-making method based on digital twinning, which comprises the following steps: establishing a global digital twinning model and a local digital twinning model according to the process of intelligent manufacturing; collecting production data in the intelligent manufacturing process in real time, determining a regulation and control index data set under a sampling number, constructing a time series augmented matrix, constructing a dynamic weight matrix by using a Pearson correlation coefficient, adjusting the time series augmented matrix into a weighted time series augmented matrix, performing singular value decomposition, introducing production process constraints to modify the singular value matrix, and reducing the dimension of the weighted time series augmented matrix by using the rank of the modified singular value matrix; standardizing the reduced weighted time series augmented matrix, dividing the standardized matrix by using a dynamic sliding window, inputting the matrix into a double-branch attention network in sequence, combining a current actual production decision scheme, and predicting a production decision scheme. The application improves the response speed of intelligent manufacturing through collaborative local self-collaboration and global scheduling.

The application relates to a real-valued super-resolution direction estimation method of a high-order acoustic field sensor array, which uses a received signal covariance matrix to approximately calculate an estimated value of noise power, avoids the iterative operation process of noise power in the prior art, improves the calculation efficiency, and reduces the floating-point operation amount. By constructing an augmented matrix, the array receiving data matrix and the array manifold matrix with a multi-dimensional structure of an element are made into Hermitian matrices, so that the unitary transformation processing of the related parameters of the high-order acoustic field sensor array is realized. On the basis of obtaining the array receiving data matrix and the array manifold matrix in the real number domain, the array receiving data matrix and the array manifold matrix are applied to a sparse approximate minimum variance method with a variable exponential factor, and the completely real-valued sparse approximate minimum variance direction estimation with a variable exponential factor is realized.

The application discloses a sequence construction method and system. The method constructs a switching matrix based on the number of various types of elements in the sequence, establishes a unified constraint system, and obtains a numerical solution of the switching matrix through optimization. Based on the numerical solution, a graph structure is constructed and connectivity is determined to determine whether a complete sequence can be formed; when the connectivity is satisfied, a target sequence that meets the switching relationship and quantity requirements is generated based on the graph structure. The method supports segment-level structure expression, modeling of the head-tail relationship of the augmented matrix, and sub-sequence combination expression, and is suitable for multiple sequence construction tasks. Accordingly, the application also provides a sequence construction system, which includes a matrix modeling module, a constraint construction module, a solving module, a connectivity determination module, and a sequence generation module.

The application discloses a kind of three-axis magnetic flux sensor measurement optimization methods, comprising: S1. the three-axis output value of magnetic flux sensor is collected;S2. according to the three-axis output value, constructs input vector;S3. with the input vector as data basis, construct coefficient augmented matrix;S4. to coefficient augmented matrix is decomposed, obtains solution vector;S5. according to the solution vector, three-axis offset correction value and three-axis proportion correction value are calculated respectively;S6. according to the three-axis offset correction value and the three-axis proportion correction value, the three-axis output value is corrected, and the output value after correction is obtained.The application can effectively correct the output value of sensor, reduce measurement error, improve the measurement accuracy of sensor.

Embodiments of the invention disclose a constraint equation determination method and apparatus, a medium, a device and a product. The method comprises the steps of obtaining a constraint equation of an augmented matrix group; the constraint equation is represented as follows: the sum of the first product and the second product is zero, a subordinate degree-of-freedom vector in the constraint equation is determined by an independent degree-of-freedom vector, a weight coefficient matrix, a displacement matrix and a distribution matrix, and the distribution matrix is determined by the weight coefficient matrix and the displacement matrix; and determining a target weight coefficient matrix and a target displacement matrix, and substituting the target weight coefficient matrix and the target displacement matrix into the constraint equation to obtain a target constraint equation of the augmented matrix group. According to the method, the weight coefficient matrix capable of accurately reflecting the physical coupling strength between the nodes and the displacement matrix capable of representing the real deformation state can be automatically calculated, the constraint equation is generated, the defect of insufficient calculation precision caused by simple hypothesis of the constraint relation is effectively overcome, and therefore the construction effectiveness of the constraint equation can be improved.

This invention relates to the field of pet food spectral detection technology, and discloses a pet food spectral detection device and method. The method involves sampling at equal intervals within a preset wavelength range, eliminating noise interference through dark current and standard white plate calibration; taking the natural logarithm of reflectance to balance the dynamic range; constructing a linear spline basis at equal intervals, and adaptively calculating the regularization coefficient based on the energy ratio of the signal and basis functions; constructing an augmented matrix based on the inner product of the basis functions and the signal, and solving for the spline coefficients using Gaussian elimination; reconstructing the spectrum using the coefficients and quantifying the residuals; establishing a reference model based on the mean and standard deviation of multiple qualified sample coefficients; comparing the Euclidean distance between the coefficients of the sample to be tested and the model mean with a threshold, and generating residuals, distances, thresholds, and a judgment report; the entire process requires no empirical parameters or manual tuning, enabling batch online adaptive detection, achieving rapid and accurate detection with noise correction, feature enhancement, overfitting suppression, and traceability.

The present invention relates to the field of automobile fault detection technology, and specifically to an automobile fault detection method, device, equipment and storage medium. The present invention first collects data at each moment of the detected automobile in real time, uses the current moment data among the data at each moment to form a current sample matrix, then applies a dynamic principal component analysis algorithm to the current sample matrix to obtain an augmented matrix of the current sample matrix, then applies a kernel principal component analysis algorithm to the augmented matrix to obtain a kernel matrix, and then constructs a principal component matrix. Finally, the eigenvectors of the principal component matrix and the kernel matrix work together to determine the detection result of the automobile. The present invention uses data for judging the detection result to come from the detected automobile itself, inputs the objective data of the automobile data into a mathematical model composed of a dynamic principal component analysis algorithm and a kernel principal component analysis algorithm, and can quantitatively detect automobile faults based on the data output by the data model, thereby improving the detection accuracy.

The invention relates to a power system dispatching optimization method and system in an abnormal state of transformers, and the method comprises the following steps: collecting high-dimensional data of each transformer in real time, and constructing a high-dimensional data matrix; performing characteristic spectrum analysis on the high-dimensional data matrix by adopting a random matrix theory, and identifying an abnormal time period; constructing an augmented matrix in the abnormal time period to perform characteristic spectrum analysis on the transformer, and positioning an abnormal transformer; and based on the abnormal time period and the positioned abnormal transformer, a scheduling optimization model is constructed and solved by taking operation cost and carbon emission minimization as targets, and a scheduling optimization process in an abnormal state is completed. Compared with the prior art, the method has the advantages of improving the abnormal precision of the transformer, realizing stable operation of a power system and the like.

The invention discloses a measurement data fusion method, which is applied to a large-size measured part and belongs to the field of data fusion technology. The method comprises: classifying a plurality of measuring instruments according to measurable parameters; combining the numbers of the plurality of measuring instruments into an instrument matrix; measuring the measurement values ​​of each characteristic parameter of a standard part; combining the measurement values ​​of the characteristic parameters into a measurement matrix; normalizing the column vectors of the measurement matrix to obtain a normalized matrix; establishing an augmented matrix according to the instrument matrix and the normalized matrix; calculating the correlation coefficient between the column vectors of two augmented matrices, and judging whether the correlation coefficient is greater than a preset value; when the correlation coefficient is greater than the preset value, performing dimensionality reduction processing on the column vectors of the two augmented matrices; inputting the column vectors of each dimensionality reduction matrix into a data fusion model for training; and terminating the training of the data fusion model when the deviation between the output value and the standard value of each characteristic parameter of the standard part is within a preset range.

The invention discloses a photovoltaic scene generation method and system based on random matrix theory enhancement, and the method comprises the steps: constructing a training sample set based on historical data, and constructing a generative adversarial network framework comprising a generator and a discriminator; on the basis of a random matrix theory, time sequence data of a generated photovoltaic power time sequence scene and a real photovoltaic power time sequence scene are constructed into an augmented matrix, spectral distribution of the augmented matrix is calculated, the spectral distribution is converted into a correlation measurement index, and the correlation measurement index serves as an additional loss item to be introduced into a loss function of a discriminator; adopting an alternate training strategy to perform joint training on the generator and the discriminator into which the correlation loss item is introduced, and optimizing a generative adversarial network framework; and inputting the random noise vector to the generator, and generating a photovoltaic power time sequence scene. Spectral distribution analysis is carried out on a generated scene and a real scene through a random matrix theory, a correlation index is used as a discriminator loss item, and the problems of insufficient interpretability, lack of correlation evaluation and the like of the generative adversarial network in photovoltaic scene generation are solved.

The invention provides a PSR-JBLND-based DOA estimation method, apparatus and device, and a medium. The method comprises the steps of constructing a sensor array receiving signal model according to a guide vector of an information source; on a preset time sequence dimension, a coordinate delay reconstruction method is adopted, a sequence in the sensor array receiving signal model is mapped in a phase space of a preset embedding dimension, and a reconstruction track matrix is obtained; obtaining the optimal time delay during phase-space reconstruction through an average mutual information method, eliminating the time delay of a reconstruction track matrix based on the optimal time delay, and then carrying out augmented matrix construction to obtain an optimized sensor array receiving signal model matrix; matrix information geometry is introduced, Jensen-Bregman LogNorm divergence is constructed, and based on the Jensen-Bregman LogNorm divergence, two Hermitian positive definite matrixes are constructed on an optimized sensor array receiving signal model matrix manifold; and obtaining a DOA estimation result based on the divergence measurement corresponding to the two Hermitian positive definite matrixes at different angles.

The invention relates to a method for calibrating a measuring device for a plurality of products, having the following steps: measuring a set of measurement values for a plurality of products and with a plurality of measurements, which includes one or more measurement variables; providing a set of reference variables, which includes at least one reference variable to be determined by the measurement variables, compiling the measurement values for all products into an augmented matrix, determining a first set of calibration parameters with one or more weights for one or more products, compiling the one or more products within one product family, which has at least the same weight for its products, determining at least one further set of calibration parameters for sub-matrices of the augmented matrix, determining one or more sets with further weights within and outside the first product family, and compiling a plurality of products with the same weight into another product family, until all weights are determined for the calibration of the measuring device.

Embodiments of the present application disclose a laser displacement sensor calibration method and device, a storage medium and an electronic device, and relate to the technical field of sensor measurement. The method comprises: constructing a matrix equation according to auxiliary parameters and a plurality of first plane equations; deleting any row elements of an augmented matrix and returning to the step of constructing the matrix equation according to the auxiliary parameters and the plurality of first plane equations until the matrix equation meets an accuracy condition to obtain a target matrix equation; and obtaining target measurement parameters of a target sensor according to the target matrix equation to complete calibration. The present application uses auxiliary parameters to construct a matrix equation for solving, continuously attempts to delete some data in a loop process to control the error of matrix solving, and the establishment of the matrix equation is derived from the measurement parameters of the target sensor and the thickness parameters of the calibration plate. In the optimization process, the influence of the plane measurement accuracy and the sensor itself error can be effectively reduced, and the quality of the sensor calibration method is improved.

The application discloses a kind of d-POD and Bi-LSTM-based impeller machine transient flow field fast prediction method. Construct full-cycle flow field snapshot matrix;Through data acquisition code, the physical quantity of flow field area of interest is extracted from simulation data, Hankel augmented matrix is constructed and standardization is completed;The time series augmented original matrix constructed is decomposed by d-POD, and the spatial mode matrix, time coefficient matrix and mode energy matrix are obtained;Bi-LSTM model is trained;The historical time series time coefficient of the predicted working condition is outputed by the model to predict the time coefficient sequence of future time;The physical quantity data reconstructed is matched and bound with the node coordinates and topological structure information of the original flow field grid one by one, to supplement the complete grid information for the restored physical quantity, according to the CFD post-processing standard format.dat File is generated, and the visualization output and subsequent engineering application of predicted flow field are completed.The present application is used to solve the core dilemma that precision and cost are difficult to consider in the prior art.

The invention relates to the technical field of wind power prediction, and discloses a physical feature enhanced SBLS-based wind power prediction method and system. The method comprises the steps of obtaining operation data of a wind turbine generator SCADA system, regarding a constructed wind speed-power dynamic characteristic curve and a decision interval as power dynamic curve knowledge, mapping the power dynamic curve knowledge into physical constraint layer nodes parallel to characteristic mapping layer nodes, and constructing an SBLS network topology containing a physical constraint layer; training the embedded SBLS prediction model containing the physical constraint layer by using a training set, outputting and combining the feature mapping layer, the enhanced mapping layer and the physical constraint layer, and constructing an augmented matrix; analyzing and solving an output weight by using a pseudo-inverse algorithm; and applying the trained SBLS prediction model containing the physical constraint layer to a test set for wind power prediction. According to the method, physical knowledge is explicitly embedded by changing a network topology structure, so that the problem of poor physical interpretability of a standard SBLS is effectively solved, and the training speed is remarkably superior to that of a deep neural network.

The application provides a query request processing method and device, the method comprises the following steps: receiving a query request sent by a user; if it is determined that global total quantity privacy protection is set and no subdivided total quantity privacy protection is set, obtaining a first historical query result recorded for the user and a current query result corresponding to the query request; wherein the first historical query result is a query result recorded based on global total quantity privacy protection; generating a first combined query matrix based on the query vectors of the first historical query result and the current query result; obtaining a first augmented matrix by adding an augmented column of all 1s based on the first combined query matrix; determining whether the first augmented matrix is full rank, if yes, responding the current query result to the user; otherwise, responding the current query result to the user after desensitization processing. The method can improve the usability of the response data while ensuring data security.

The present invention belongs to the field of intelligent transportation, and specifically relates to a short-term traffic flow prediction method based on a spatiotemporal convolutional network; the method comprises: acquiring traffic data from a traffic checkpoint; calculating the activity of the traffic checkpoint based on the traffic data; extracting hidden features of the checkpoint nodes and fusing the traffic checkpoint activity to obtain a hidden information matrix; generating an adaptive adjacency matrix based on the hidden information matrix, inputting the matrix into an adaptive graph convolutional network module for processing to obtain initial traffic flow features; processing the initial traffic flow features using a BiGRU network to obtain traffic spatiotemporal attribute features; extracting external attribute features and generating an augmented matrix; inputting the augmented matrix into the BiGRU network and processing it using an attention mechanism to generate attention weights; predicting traffic flow based on the traffic spatiotemporal attribute features and the attention weights to obtain a traffic flow prediction result; the present invention can achieve accurate prediction of traffic flow and has broad application prospects.

The present invention relates to a power system scheduling optimization method and system under abnormal conditions of transformers. The method includes the following steps: collecting high-dimensional data of each transformer in real time to construct a high-dimensional data matrix; performing eigen-spectrum analysis on the high-dimensional data matrix using random matrix theory to identify abnormal time periods; constructing an augmented matrix during the abnormal time periods to perform eigen-spectrum analysis on the transformers to locate abnormal transformers; based on the abnormal time periods and the located abnormal transformers, constructing a scheduling optimization model with the objectives of minimizing operating costs and carbon emissions and solving it to complete the scheduling optimization process under abnormal conditions. Compared with the prior art, the present invention has the advantages of improving the accuracy of transformer abnormalities and realizing the stable operation of the power system, etc.

The invention discloses a distributed solving method and system for a large-scale linear equation set based on a clustering network, and belongs to the technical field of electric digital data processing, and the method comprises the steps: S1, constructing the large-scale linear equation set; s2, decomposing the augmented matrix, firstly dividing the augmented matrix into 1 different parts according to columns, and then subdividing each part into li sub-parts according to rows; s3, constructing a clustering multi-agent network, wherein each agent has a corresponding selection matrix and serves as a calculation unit to solve a small-scale linear equation; forming a cluster i by a group of intelligent agents; each agent ij solves a small-scale linear equation set through iterative calculation by using own information and information obtained by direct communication with adjacent agents in all clusters in the solving process; and S4, performing distributed solution on the large-scale linear equation through the clustering multi-agent network. According to the method, the large-scale linear equation set is rapidly solved through the clustering network, and the efficiency of large-scale data processing is improved.

The present invention relates to artificial intelligence, and provides a group classification method, apparatus, device and storage medium. The method constructs a group augmented matrix including a group matrix according to user feature information and a preset value, performs disassembling and augmentation processing on the group matrix to obtain an initial mapping matrix and an initial feature matrix, constructs a first mask matrix and a second mask matrix according to the missing information of the user feature information, calculates a target matrix corresponding to the group matrix and the initial feature matrix based on the first mask matrix to obtain an intermediate mapping matrix, calculates the initial mapping matrix and the group augmented matrix based on the second mask matrix to obtain an intermediate feature matrix, adjusts the intermediate feature matrix based on the relationship matrix between the group matrix, the intermediate mapping matrix and the intermediate feature matrix to obtain a target feature matrix, and classifies multiple users based on the target feature matrix, thereby improving the accuracy of the group type. In addition, the present invention also relates to blockchain technology, and the group type can be stored in the blockchain.

The invention relates to the technical field of large-scale multiple-input-multiple-output, and provides a signal processing method and device, electronic equipment, a medium and a product, and the method comprises the steps: determining a Gramer matrix of a channel matrix according to the channel matrix of a multiple-input-multiple-output (MIMO) system and the conjugate transpose of the channel matrix; decomposing an augmented matrix determined according to the conjugate transpose of the channel matrix and the Gramer matrix to obtain a decomposed augmented matrix, and eliminating the Gramer matrix in the decomposed augmented matrix into an upper triangular matrix, a lower triangular matrix corresponding to the upper triangular matrix is determined according to elimination factors recorded in the elimination process; and back substitution calculation is carried out on the decomposed augmented matrix to obtain a weighting matrix, and the weighting matrix is used for carrying out zero-forcing detection on the received signal. The adaptability of the real-time signal detection method is improved, and the problem that an existing method is difficult to deal with changeable service scenes is effectively solved.

The application relates to the technical field of large-scale multiple-input multiple-output, and provides a signal processing method and device, electronic equipment, a medium and a product.The method comprises the following steps: determining a Gram matrix of a channel matrix of a multiple-input multiple-output (MIMO) system according to the channel matrix and a conjugate transpose of the channel matrix; decomposing an augmented matrix determined according to the conjugate transpose of the channel matrix and the Gram matrix to obtain a decomposed augmented matrix, wherein the Gram matrix is eliminated as an upper triangular matrix in the decomposed augmented matrix, and a lower triangular matrix corresponding to the upper triangular matrix is determined according to an elimination factor recorded in an elimination process; and performing back substitution calculation on the decomposed augmented matrix to obtain a weighting matrix, wherein the weighting matrix is used for zero-forcing detection of a received signal.The adaptability of a real-time signal detection method is improved, and the problem that an existing method is difficult to cope with various service scenarios is effectively solved.

The present invention discloses a frequency diversity MIMO radar parameter estimation method based on the matrix beam principle, the method comprising: firstly, the receiving signal model of the frequency diversity MIMO radar is derived and the coupling relationship between the angle and the distance in the transmit-receive steering vector is analyzed; secondly, the augmented matrix beam is constructed using the echo signal obtained after matched filtering at the receiving end; then, the augmented matrix beam is subjected to matrix decomposition and transformation to obtain the diagonal phase matrix containing the angle and distance information respectively. Finally, considering the coupling and matching problem of the angle and distance parameters of the target, the same characteristic vector is used to achieve matching and realize accurate positioning of the target. Simulation results show that compared with the prior art, the present invention exhibits superior estimation performance under low snapshot number and low signal-to-noise ratio.

The present invention relates to a method and system for solving a linear programming problem, and in particular to the field of linear programming problem computing technology. The method comprises constructing a linear programming problem computing model based on Kuhn-Tucker conditions; constructing an augmented matrix according to the model; judging whether there are positive values ​​in the first m elements of the last row of the augmented matrix; if there are, determining all positive values ​​as a positive number matrix; and updating the augmented matrix according to the positive number matrix, and then returning to judge whether there are positive values; if not, obtaining a feasible solution and an objective function value according to the augmented matrix; judging whether the first n elements in the last column of the augmented matrix are all greater than or equal to 0; if yes, obtaining an optimal solution and an objective function value according to the augmented matrix; if not, calculating the second row number and the second column number according to the augmented matrix, and updating the augmented matrix according to the two numbers, and returning to judge whether they are all greater than or equal to 0. The present invention can quickly obtain the optimal solution.

The invention discloses a misalignment confirmation method based on fusion of a guide rail meter and multi-source metering data, and the method comprises the following steps: periodically collecting the forward active electric energy data of each user electric energy meter in a metering box through a guide rail electric energy meter, and transmitting a broadcast freezing instruction to all user meters before collection, so as to achieve the synchronization of a data clock; constructing an electric energy conservation equation; accumulating N periodic data to form N equations, constructing an augmented matrix, and solving an error value through a Gaussian elimination method; and calculating the actual relative error of the user table according to a relative error conversion formula by using the error value, and if the actual relative error exceeds a threshold value, determining that the user table is misaligned. According to the method, a micro closed loop is established based on the direct relation between a guide rail meter and a user side single-phase meter, a high-precision and low-dependence metering misalignment recognition mechanism can be realized by comparing the data difference of total-branch meters and eliminating line loss interference, and the method has higher practicability and popularization value.

This invention proposes an FTM-corrected non-line-of-sight ranging method based on matrix-bundled CSI multipath parameter estimation. First, Channel State Information (CSI) data from WiFi devices is collected, an augmented matrix is ​​constructed, and singular value decomposition is performed to obtain a left singular matrix and singular value vectors. Second, the average power of the CSI is calculated and transformed to the logarithmic domain to determine the maximum number of candidate paths. Based on the singular value vectors, a Minimum Description Length (MDL) criterion is constructed to adaptively estimate the number of effective paths. Then, the propagation delay of each path is obtained using the matrix-bundled algorithm, a Vandermonde matrix is ​​constructed, and the received power of each path is estimated using the least squares method. Finally, the power ratio of the direct path is calculated as a reliability factor. If it is below a threshold, the second strongest path is selected from the reflection paths with power greater than the direct path. If there is only one such path, it is selected. The average delay difference between the reflection path and the direct path is calculated using power weighting, and this delay difference is used to compensate for the FTM receiving timestamp to obtain the corrected ranging distance. This invention does not require a preset number of paths, is highly adaptive, and significantly improves the accuracy of FTM ranging in non-line-of-sight environments.