128 results about "Eigendecomposition of a matrix" patented technology

The invention relates to the technical field of artificial intelligence, and provides a quantum-enhanced multi-scale network intrusion detection method, which comprises the following steps: calculating a covariance matrix for an original traffic feature matrix, and obtaining a feature value and a feature vector through feature decomposition, mapping each sample xi to a quantum Hilbert space to generate an enhanced feature matrix, executing complex field transformation on the enhanced feature matrix to generate an entangled feature tensor, and realizing dynamic feature enhancement through a multi-head attention mechanism based on a quantum probability amplitude; performing space-time attention calculation and gating fusion on the feature tensor after dynamic feature enhancement to obtain a space-time fusion feature; converting the space-time fusion features into a time sequence form, extracting behavior features through a multi-scale convolution branch, and fusing the behavior features to obtain a three-dimensional feature tensor; and calculating a mean value of the three-dimensional feature tensor in a sequence dimension, generating a two-dimensional feature matrix, and performing classification prediction, uncertainty quantification and threat grading evaluation based on a classification network, an uncertainty network and a threat grading network.

The invention relates to the technical field of welding stress detection, and discloses a vehicle welding stress detection method and device, and the method comprises the steps: applying a preset load to a welding part, carrying out the nonlinear finite element analysis and data driving mapping processing of an initial deformation field matrix, carrying out the characteristic decomposition of a stress field tensor to form a characteristic tensor, and carrying out the detection of the welding stress. Carrying out threshold segmentation on stress concentration factor distribution, inputting an abnormal region coordinate set into a damage evaluation model to carry out life and damage quantification processing, carrying out data integration and visualization processing on damage indexes, and generating a welding stress detection report; by applying the preset load, collecting the three-dimensional strain data of the welding part, constructing the initial deformation field matrix and combining nonlinear finite element analysis and data driving mapping processing, the stress field tensor is directly obtained, the error caused by deducing stress distribution only depending on surface deformation is avoided, and the accuracy of stress calculation is improved.

The invention provides a computing power resource partitioning method and device based on hypergraph clustering, equipment and a medium. The method comprises the following steps: constructing a computing power resource hypergraph in a computing power network resource side scene; determining an equipment degree matrix and a hyperedge degree matrix of the computing power resource hypergraph, and converting the computing power resource hypergraph into an equipment Laplacian matrix required by hypergraph clustering by using the equipment degree matrix and the hyperedge degree matrix; carrying out eigendecomposition on the equipment Laplacian matrix to obtain an equipment eigenvector; and in combination with the graph cutting target function, performing hypergraph clustering processing on the computing power resource hypergraph based on the feature vector of the device Laplacian matrix to obtain a target resource partitioning result corresponding to the computing power resource hypergraph. According to the method, the problems of low resource utilization rate, high scheduling complexity, poor resource dynamics, lack of effective partitioning strategies and the like existing in the existing computing power resource partitioning can be obviously improved.

The invention belongs to the technical field of image processing, and particularly relates to a composite material ultrasonic scanning image defect feature extraction method and system, and the method comprises the steps: obtaining an ultrasonic scanning image of a composite material, and carrying out the filtering processing; constructing a gradient outer product matrix of neighborhood pixel points of the pixel points and accumulating to obtain a local structure matrix; performing characteristic decomposition on the local structure matrix to obtain a characteristic value, and calculating a structure coherence factor according to the characteristic value; coupling potential energy is constructed in combination with a gray value and a structural coherence factor, and a low-gray and disordered defect signal is highlighted through nonlinear gain; and performing region segmentation by using the high-coupling potential energy points as anchor points, and extracting defect blocks. According to the method, the problem of low-contrast defect leak detection under the strong texture background is effectively solved by utilizing the essential difference between the background texture and the defect in the physical topology, and the defect feature extraction precision is remarkably improved.

The invention discloses a remote area video monitoring automatic anomaly detection method based on deep learning. The method comprises the following steps: S1, collecting remote area video data and completing preprocessing; s2, adjacent frame difference is executed in the transverse direction, channel fusion is executed in the longitudinal direction, and a composite frame set is generated; s3, performing bidirectional mask modeling and disturbance reconstruction on the composite frame set to obtain a regional feature matrix; s4, mapping the regional feature vectors into nodes, constructing a weighted adjacent matrix, and generating graph structure representation; s5, performing Laplace characteristic decomposition on the graph structure representation, and judging that the nodes are abnormal; s6, according to the abnormal node set, sub-graphs are divided according to topological neighborhoods, abnormal scores are calculated, and an abnormal region set is obtained; and S7, mapping the abnormal region set into the video data, and outputting an abnormal position, starting and ending frame numbers and an abnormal scoring result. According to the method, efficient and automatic anomaly detection of videos in remote areas is realized, the false alarm rate is remarkably reduced, and the anomaly positioning precision is improved.

The invention provides a multi-frequency distributed nested array DOA estimation method based on low-rank constraint, and the method comprises the steps: S1, building a receiving signal model of a distributed nested array, and obtaining a receiving signal based on the receiving signal model; s2, calculating a covariance matrix of the received signal, and performing vectorization processing on the covariance matrix to obtain a first virtual array corresponding to the distributed nested array; s3, extrapolating sub-arrays in the first virtual array to fill part of holes existing under the long baseline condition to obtain a second virtual array; s4, zero elements are inserted into the remaining hole positions to obtain a third virtual array, and the rank of the covariance matrix corresponding to the third virtual array is recovered; s5, establishing a matrix reconstruction model combining the low-rank constraint and the sparsity constraint, and solving to obtain a complemented covariance matrix; and S6, carrying out characteristic decomposition on the complemented covariance matrix to obtain a final DOA estimation value.

The invention discloses a coupling tensor decomposition method based on sparse low-rank constraint, which is used for characteristic decomposition of a multi-frequency dynamic function network connection tensor in resting state function magnetic resonance imaging data. According to the algorithm, on the basis of the traditional coupling canonical factorization (CCPD), an optimization model of sparse and low-rank constraint is constructed, and the sparse and low-rank constraint optimization model is constructed by the algorithm. On the spatial connectivity dimension, redundant function connection is reduced through an L1 sparse penalty term, and the spatial specificity of the key brain network is enhanced; and in time and frequency band dimensions, low-rank regularization constraint is adopted to improve discrimination of cross-subject time sequence characteristics. Generally speaking, the method can effectively extract connectivity characteristics with statistical significance and time states of different frequency bands from dynamic function network connection tensors of multiple frequency bands, thereby effectively identifying functional connection heterogeneity characteristics between schizophrenia patients and healthy control groups.

The invention is suitable for the field of geophysics, and provides a geologic structure edge detection method based on an enhanced gravity tensor eigenvalue, and the method comprises the steps: preparing data, and generating a gravity gradient tensor matrix; performing characteristic decomposition on the gravity gradient tensor matrix to obtain a gravity tensor characteristic value; performing deep decomposition on the gravity tensor characteristic value to extract amplitude information; constructing an enhanced edge detection filter F based on the amplitude information; according to the above steps, the structure edge detection calculation based on the enhanced gravity tensor characteristic value is realized, and the underground geologic structure edge is obtained. Compared with a common characteristic value detection technology, the geological structure (or target body) edge detection result obtained through the method technology can obtain a clearer and more continuous recognition effect, the positioning accuracy is high, the method can be well suitable for edge description of the geological structure (or target body) with multiple scales and different buried depth superposition, and the practicability is higher.

The invention discloses a narrowband adaptive digital beam forming and optimization design, which comprises the following steps of: providing an optimized robust adaptive beam forming method on the basis of the idea of covariance matrix correction and guide vector estimation: firstly, reconstructing an IPNCM matrix, establishing an optimal interference covariance matrix in an interference signal angle sector by applying the idea of an uncertain set, and then establishing an optimal interference covariance matrix in an interference signal angle sector; a noise covariance matrix is established by applying characteristic decomposition of a matrix, so that an IPNCM matrix is obtained, the influence of an expected signal in the covariance matrix is reduced to the maximum extent, the suppression capability of an algorithm on interference noise under the condition of low SNR is ensured, the situation that the algorithm misjudges the expected signal as interference for suppression under the condition of high SNR is avoided, and then the noise suppression capability of the algorithm is improved. A new constraint condition is added on the basis of the RCB algorithm, and it is ensured that the guiding vector converges to an expected signal instead of an interference signal, especially under the condition of low SNR. An optimized robust Capon algorithm is provided by using a Capon power spectrum, an uncertainty set idea and convex optimization, and the provided algorithm has higher robustness and can ensure effectiveness.

The invention discloses an artificial intelligence-based interactive question-answering method and system, and relates to the technical field of interactive question-answering, and the method comprises the steps: converting a user input text question into vector representation, constructing a relation graph by combining a dependency relation tree of a text entity, capturing a dependency relation in the question as an edge weight, forming an adjacent matrix, and constructing a node degree matrix; and determining a standardized graph Laplacian matrix, performing feature decomposition, determining an optimal clustering number, clustering language phrases of the text question, and determining task fragment sets of different clusters. According to the method, semantic association between language phrases is achieved by constructing the dependency relationship graph, meanwhile, syntactic relationships are fused, the integrity of text structure information is ensured, the degree of nodes serves as a quantitative index of local strength in the graph, the information directly supports subsequent Laplacian matrix normalization processing, a unified scale is provided, and the method is suitable for being applied to the field of text processing. And the imbalance problem of the data range is avoided.

The invention discloses a very high frequency broadband radiation source positioning method based on energy weighted subspace fitting, and relates to the technical field of signal processing and radiation source imaging, and the method comprises the steps: carrying out the DFT processing of a lightning very high frequency broadband signal collected by an array, and obtaining any narrowband component; then characteristic decomposition is carried out on covariance matrixes of X-axis and Y-axis sub-arrays according to array receiving signals, main characteristic vectors are taken as signal sub-spaces, in order to emphasize the signal sub-spaces with higher energy, characteristic values are used for weighting, weighted signal sub-spaces are constructed, then the signal sub-spaces are fitted by using a least square method, and the signal sub-spaces with higher energy are obtained; the method comprises the following steps: further constructing a sub-band energy weighted spatial spectrum adapted to a very high frequency broadband signal, then carrying out spectrum peak search on the weighted spatial spectrum to obtain a spatial angle of a spectrum peak, and finally calculating the direction of arrival of a radiation source according to a spatial angle relation to obtain an azimuth angle and an elevation angle. According to the invention, aiming at a lightning very-high-frequency weak radiation source positioning problem, a lightning channel can be inverted more clearly.

The invention relates to a non-stationary industrial process anomaly detection method based on slow characteristic decomposition and Kupman high-dimensional space prediction, and belongs to the technical field of industrial process time sequence anomaly detection.The method comprises the steps that characteristic decomposition is conducted on industrial time sequence data through a slow characteristic analysis method, and the industrial time sequence data are divided into fast and slow change parts; respectively mapping the fast and slow features to a linearly observable Kupman high-dimensional space based on the Kupman theory, and constructing a depth prediction model to realize step-by-step prediction of a feature sequence; in combination with the deviation between a prediction result and an actual value, constructing an SPE statistical magnitude approximately obeying weighted chi-square distribution, and setting an SPE control limit of anomaly detection according to the SPE statistical magnitude; introducing KL divergence to measure a distribution difference between a current working condition and a normal working condition, and adaptively updating an SPE control limit according to the distribution difference; and in the model online deployment stage, an SPE value is calculated in real time and compared with a control limit, so that abnormal rapid identification and dynamic early warning under a non-stable working condition are realized.

Embodiments of the invention disclose a model quantification method and apparatus, an electronic device, a storage medium and a program product. The method comprises the steps of obtaining to-be-quantized data in a target model; performing characteristic decomposition on the covariance matrix of the to-be-quantized data to obtain a characteristic vector matrix; constructing a variance equilibrium rotation matrix based on the feature vector matrix; based on the variance equilibrium rotation matrix, performing variance equilibrium transformation on the to-be-quantized data; and based on the to-be-quantized data after variance equilibrium transformation, performing quantization processing on the target model so as to deploy the quantized target model to hardware equipment, the quantized target model being capable of being called by the hardware equipment to perform corresponding task processing, thereby performing principal component analysis on the to-be-quantized data based on eigendecomposition, and obtaining the to-be-quantized data. According to the technical scheme, accurate alignment of the variances between the channels can be achieved, the variances between the channels of the to-be-quantized data tend to be consistent, numerical distribution of the quantized data is more uniform, higher quantization precision is achieved, and then the execution efficiency of hardware equipment is improved.

The invention discloses a transformer partial discharge multi-target positioning method and system based on an ultrasonic array. The method comprises the following steps: S1, receiving an ultrasonic signal and constructing a data matrix; s2, calculating a covariance matrix and performing characteristic decomposition; s3, estimating the number of partial discharge sources based on a Gerschnoid circle criterion; s4, determining the spatial orientation of the discharge source based on a multi-signal classification algorithm; and S5, calculating space coordinates through multi-platform direction-finding cross positioning: synchronously measuring the azimuth angle theta and the pitch angle of the same discharge source through at least two independently arranged ultrasonic arrays, combining the azimuth angle theta and the pitch angle with the space position coordinates of the arrays, establishing an equation set based on a triangular relationship, and solving the equation set to obtain the three-dimensional coordinates of the discharge source so as to realize multi-target positioning. The method has the advantages of high positioning precision and the like.

Provided in the present application are a channel information feedback method, a channel information receiving method, and related apparatuses. The channel information feedback method provided in the present application comprises: a terminal device receiving first indication information, wherein the first indication information is used for indicating M space-frequency units for which the terminal device feeds back channel information, M being an integer greater than or equal to 1; and then, on the basis of the first indication information, the terminal device sending the channel information of the M space-frequency units, wherein the channel information of the M space-frequency units is obtained by means of measuring a first reference signal, the first reference signal is carried on N space-frequency units, and the M space-frequency units are some or all of the N space-frequency units, with N being an integer greater than or equal to 1. Therefore, a network device can conveniently perform inference on the basis of the channel information of the M space-frequency units, so as to obtain channel information of the N space-frequency units. A terminal device directly feeds back channel information of some space-frequency units, without the need to perform eigenvalue decomposition and DFT calculation on a downlink measurement channel so as to obtain a precoding matrix. Therefore, the terminal complexity is reduced.

The invention relates to the technical field of data processing, and discloses a large matrix characteristic decomposition method, system, storage medium and device, the method carries out matrix decomposition based on House holder transformation and Givens transformation, and the problems of low efficiency and the like in the prior art are solved.

The disclosure describes a method of evaluating performance of a two-sided model used for performing CSI compression. The method includes receiving, by a network, from a UE, a compressed channel Ht, a second eigenvector EVt+1 corresponding to a channel Ht+1, and a first similarity score determined between a first eigenvector EVt and the second eigenvector EVt+1. A channel Ht is reconstructed by processing the compressed channel Ht using a decoder part of the two-sided AI / ML model. Eigen decomposition of the channel Ht is performed to obtain an eigenvector (EV)t. A second similarity score is determined between the eigenvectors (EV)t and EVt+1. Alternatively, the first similarity score is determined between an estimated channel H and a codebook based precoder W, and a second similarity score is determined between a reconstructed channel H and the codebook based precoder W.

The application discloses a kind of conformal array rotation interference amplitude-phase error correction design methods, first, construct rotating array signal receiving model, according to the radius of rotating platform and angular velocity calculation rotating guide vector, construct rotating array channel amplitude-phase error model, correction source incident signal information acquisition, record the rotation angle of precision turntable;Covariance matrix is solved to the received data and eigenvalue decomposition, construct amplitude-phase error calculation equation;Solve amplitude-phase error matrix, construct amplitude-phase error correction matrix and correct the received data;Solve adaptive weight vector, and weighted output is carried out to the corrected data.The application compared with traditional array amplitude-phase error correction scheme, system structure is simple and easy to realize, without specific algorithm can be completed to the amplitude-phase error caused by array receiving channel is accurately corrected, effectively solve the receiving channel mismatch problem, improve array anti-interference performance.

This invention discloses a method and system for separating AIS aliasing signals using a four-circular array antenna. The steps are as follows: S100: AIS signals are received through multiple antennas to obtain the spatial baseband signal matrix X, and the phase error Δφ is calculated using interval calculation. k S200: Calculate the phase compensation matrix p; S200: Compensate the baseband signal according to the phase compensation matrix to obtain the baseband signal X. comp S300: For baseband signal X comp By expanding the spatial and temporal dimensions, we obtain the spatiotemporal snapshot matrix X. st (t); S400: The covariance matrix R is obtained by estimating the spatiotemporal snapshot matrix. xx S500: For the covariance matrix R xx Eigenvalue decomposition is performed to obtain the signal subspace and noise subspace; S600: DOA estimation is performed using the MUSIC algorithm to construct the spatial spectrum P(θ); S700: The beam w at the optimal solution is obtained using the space-time constrained minimum variance algorithm; S800: The separated signal is output through the sparse separation method of protocol features. This method can significantly improve the availability of AIS self-organizing network and provide more reliable protection for ship navigation safety.

The invention belongs to the field of network addiction detection, and provides a network space addiction behavior index system construction method and device, equipment and a medium, and the method comprises the steps: obtaining an index and an addiction level of a network space addiction behavior, and determining a visual angle; performing two-point anchor point selection processing on the visual angle to obtain an anchor point set; according to the anchor point set, a similarity graph, a diversity graph and a consistency graph of the anchor points are constructed, and the similarity graph, the diversity graph and the consistency graph are used for representing the incidence relation of the view angles; performing fusion calculation on the diversity graph and the consistency graph of the anchor points to obtain a weighted similarity matrix; carrying out eigendecomposition on the weighted similarity matrix to obtain low-dimensional embedded representation, and carrying out supervised training on the low-dimensional embedded representation through a loss function to obtain a network space addiction behavior index system; and network addiction detection is carried out through the network space addiction behavior index system. According to the invention, the accuracy of user network addiction level identification is improved.

The invention discloses a weighted ESPRIT-based DOA estimation method under the assistance of a semi-passive intelligent reflection surface, and the method comprises the steps: constructing a DOA estimation model under the assistance of the semi-passive intelligent reflection surface, and obtaining a receiving signal; thirdly, calculating a covariance matrix of a received signal, decomposing the characteristics of the covariance matrix, and extracting a signal subspace and a characteristic value; secondly, constructing a Gram matrix, performing characteristic decomposition on the Gram matrix, and extracting a signal subspace and a characteristic value; then, calculating a weight parameter according to the characteristic value; and finally, weighting and fusing the signal subspace information to obtain DOA estimation. According to the method, the target can be effectively sensed under the condition that the base station is shielded, and DOA estimation is carried out. In addition, the characteristics of the intelligent reflection surface are fully utilized, an adaptive weighting mechanism is introduced, and the DOA can be estimated robustly. Simulation results show that the DOA estimation precision of the algorithm provided by the invention is superior to that of a traditional multiple signal classification method, and is close to and approaches the Cramer-Rao bound with an atomic norm minimization algorithm.

The invention provides an abnormal transaction identification method and device, equipment and a storage medium, and relates to the technical field of financial data identification. The method comprises the following steps: acquiring entities in transaction data to be identified and a transaction relationship between the entities; and generating a weighted financial association network graph according to the entities and the transaction relationship between the entities. And calculating a Laplacian matrix of the weighted financial association network diagram, and carrying out eigendecomposition on the Laplacian matrix to obtain eigenvectors corresponding to the first k minimum eigenvalues. Generating a feature matrix based on the k feature vectors; and clustering nodes in the feature matrix to obtain k node communities, and determining a risk score of each node community. And if it is determined that the target node community with the risk score exceeding the preset risk threshold exists, determining that the target node community is an abnormal object. The method is used for achieving the effect of improving the recognition capability of abnormal transaction structures of complex and irregular network communities.

An adaptive beamforming method disclosed by the present invention mainly solves the problem of the output SINR decrease caused by the array manifold mismatch in adaptive beamforming. The implementation process is as follows: using a uniform linear array to collect training data; constructing a spatial blocking matrix by means of the prior angle information of the target; preprocessing the training data with the blocking matrix and calculating the interference covariance matrix based on the minimum power criterion; using matrix projection transformation to perform eigenvalue decomposition on the interference subspace matrix; and optimizing the beamforming weight vector by combining the idea of spatial response invariance. When there is a mismatch in the array manifold, the present invention can output the target without distortion on the premise of ensuring the anti-interference ability, and can be used to realize adaptive beamforming in the presence of the angle of arrival and array calibration errors.

The application belongs to the technical field of image processing, and particularly relates to a composite material ultrasonic scanning image defect feature extraction method and system, which comprises the following steps: obtaining an ultrasonic scanning image of a composite material and performing filtering processing; constructing a gradient outer product matrix of neighborhood pixel points of a pixel point and accumulating to obtain a local structure matrix; performing feature decomposition on the local structure matrix to obtain an eigenvalue, and calculating a structure coherence factor according to the eigenvalue; combining a gray value and the structure coherence factor to construct a coupling potential, highlighting a low-gray and disordered defect signal through a nonlinear gain; and using a high coupling potential point as an anchor point to perform region segmentation and extract a defect block. The application effectively solves the problem of missing detection of low-contrast defects in a strong texture background by using the essential difference between background texture and defects in physical topology, and significantly improves the precision of defect feature extraction.

The application discloses a point cloud local curvature feature calculation method and system for structural surface identification, and particularly relates to the technical field of three-dimensional point cloud data processing and feature extraction, and is used for solving the problem that the existing method leads to local curvature feature distortion due to direct calculation based on original point cloud containing noise, and then affects the subsequent structural surface identification precision; a plurality of different neighborhood radiuses are calculated respectively for the query point in the point cloud, and a plurality of sets of eigenvalues and eigenvectors are obtained by performing feature decomposition on the covariance matrix; an optimal set is selected based on the stability and consistency of the feature structure, and then the curvature tensor of the point is calculated; the candidate local curvature eigenvalue is derived from the curvature tensor and verified and output according to the geometric flatness prior knowledge of the structural surface; through the multi-scale analysis and verification mechanism, the interference of noise and local fluctuation is effectively inhibited, and more reliable and more robust local curvature features can be calculated.

A system and method for spectral learning in persistent cognitive machines implements learning through controlled evolution of a spectral decomposition of a cognitive manifold. The spectral decomposition, comprising eigenvectors and eigenvalues, encodes long-term memory as global geometric structure rather than as stored data or network parameters. The system performs inference operations by projecting incoming data onto the cognitive manifold using a fixed spectral decomposition without modification. Geometric invariants including principal angles, spectral gap ratios, projection residuals, and curvature statistics are continuously monitored to detect structural inadequacy. When invariants exceed thresholds, a learning event modifies the spectral decomposition through eigen decomposition with warm-start initialization while enforcing mode-specific plasticity bounds that are tighter for low-frequency eigenvectors than high-frequency eigenvectors, thereby preventing catastrophic forgetting. The system operates continuously by alternating between inference using fixed spectral decompositions and learning events that modify spectral decompositions through controlled spectral evolution.

This invention relates to an improved spatially smoothed source angle estimation method and apparatus based on a coprime linear array. The method includes: acquiring a signal from a source under test using an augmented coprime array to obtain a received signal; vectorizing the covariance matrix of the received signal, sorting the resulting vectors according to the element positions of the uniform linear array, and processing them according to a data processing strategy to obtain a virtual signal received by a virtual array; uniformly dividing the virtual array into overlapping virtual subarrays based on a pre-constructed spatial smoothing rule, and calculating the spatial smoothing covariance matrix of the received signals from the virtual subarrays; performing eigenvalue decomposition on the spatial smoothing covariance matrix using a predefined spatial spectrum estimation method to obtain a spatial spectrum function; and performing spectral peak search on the spatial spectrum function to determine the source angle of the source under test. The method provided by this invention improves the array degrees of freedom and the accuracy of source angle estimation by constructing a larger number of virtual array elements.

The present invention provides a method for direct positioning and tracking of non-circular signals based on Taylor compensation at distributed monitoring stations, comprising the following steps: obtaining a received signal from a monitoring station; expanding the received signal, calculating a covariance matrix, and performing eigendecomposition on the covariance matrix; directly positioning and tracking the non-circular signal based on Taylor expansion to obtain the position of the signal source; and applying the least squares method to predict the position of the non-circular signal when the trajectories of multiple non-circular signals intersect. The present invention significantly reduces the complexity of the algorithm. In the deviation calculation, the orthogonality between the steering vector and the noise subspace of the received signal is fully utilized to calculate the deviation for tracking. The method has low algorithm complexity and good monitoring accuracy, and can still obtain good tracking and positioning results in poor environments. It avoids the data association problem when multiple signal sources exist, and considers the mutual influence problem when the signal source trajectories intersect, resolving the position ambiguity existing at that moment through the least squares method.

The invention relates to the technical field of photon calculation, in particular to a wavelength division multiplexing photon neuromorphic calculation method and system. Acquiring a central data matrix based on the electric signal matrix and the mean vectors of all the preprocessed electric signals; acquiring a covariance matrix based on the total number of the electric signals and the central data matrix; performing eigendecomposition on the covariance matrix to obtain a group of eigenvectors and eigenvalues corresponding to the eigenvectors, constructing a compression matrix, and compressing the electric signal matrix; flattening the compressed electric signal matrix in a time division multiplexing mode, loading the compressed electric signal matrix to different wavelengths of a streamline photon reserve pool calculation system in a wavelength division multiplexing mode to realize optical calculation, and obtaining an output electric signal through photoelectric conversion; and decompressing the output electric signal to obtain a processing result of the input data. Through fusion of a compression algorithm and a wavelength division multiplexing structure, the data processing efficiency and precision are effectively improved, and meanwhile, through streamline hidden layer design, the integration of the system is enhanced.

The invention relates to the technical field of radars, in particular to an inverse main lobe interference method and device based on signal subspace steering vector reconstruction, and the method comprises the steps: constructing a first covariance matrix corresponding to an interference sample in a radar array receiving signal, carrying out the eigendecomposition according to the first covariance matrix, and obtaining a first eigenvector; selecting a first feature vector corresponding to the main lobe interference, constructing a feature projection matrix, and eliminating the main lobe interference through the feature projection matrix to obtain an intermediate radar signal; constructing a second covariance matrix of the intermediate radar signal, and performing characteristic decomposition on the second covariance matrix to obtain a second characteristic vector; obtaining a signal subspace according to the second feature vector, reconstructing a steering vector of the signal subspace, and substituting the reconstructed steering vector into a preset algorithm to obtain a target angle spatial spectrum so as to determine a target angle; the method has high target direction estimation precision and has the capability of suppressing multi-lobe interference.