71 results about "Robust principal component analysis" patented technology

The invention relates to the technical field of precision manufacturing nondestructive testing and machine vision image processing, in particular to a titanium alloy ring piece surface microcrack defect detection system based on machine vision, which comprises an image acquisition module for acquiring a to-be-processed image data set; the manifold calibration module is used for acquiring a main direction field of background textures and converting the to-be-processed image data set into a standard space image with aligned texture flow; the sparse decomposition module is used for acquiring a shear wave coefficient and decomposing the shear wave coefficient into a low-rank component matrix and a sparse component matrix; the reconstruction judgment module is used for generating a microcrack defect distribution diagram, obtaining a residual image and generating a final defect distribution diagram; the self-adaptive feedback module is used for adjusting the sparsity constraint weight in the robust principal component analysis algorithm; according to the method, the problem of signal aliasing caused by frequency overlapping of cracks and background textures is effectively solved, and the detection sensitivity under the strong texture background is remarkably improved.

The invention relates to the technical field of nondestructive testing, in particular to an automatic defect detection and segmentation technology based on eddy current pulse thermal imaging, which comprises a thermogram sequence preprocessing step, a defect signal feature extraction step and an image segmentation post-processing step, in the thermogram sequence preprocessing step, an excitation peak frame image is dynamically selected through an image entropy difference, a static background is inhibited by combining image difference operation, and a geometric coil mask is generated by utilizing edge detection; and a defect signal feature extraction step: carrying out defect matrix reconstruction on the robust principal component analysis model based on the difference image by adopting an augmented Lagrange multiplier method. And an image segmentation post-processing step: designing a dual-threshold segmentation method based on reconstruction matrix local space consistency, and screening thresholds by combining an adaptive threshold of image gray level distribution and an area screening threshold of a defect candidate region. Automatic detection and segmentation of defects in eddy current pulse thermal imaging are achieved, and the purposes of defect identification and quantitative detection are achieved.

The invention relates to the technical field of image analysis, in particular to a metal composite material surface defect intelligent detection method based on machine vision, which comprises the following steps: acquiring an original image and converting the original image into an observation matrix; decomposing the matrix into a low-rank matrix and a sparse matrix through a robust principal component analysis algorithm; utilizing a nuclear norm constraint low-rank matrix to establish a background statistical model, and utilizing a norm constraint sparse matrix; positioning a defect area in the sparse matrix and extracting a gray level co-occurrence matrix feature vector; performing connected domain analysis on the sparse matrix, calculating geometrical morphology parameters and evaluating a stress concentration coefficient; and establishing a self-adaptive decision model to carry out fusion judgment so as to judge the physical damage. According to the method, through low-rank sparse matrix decoupling and multi-dimensional statistical geometric feature fusion analysis, accurate stripping of weak defects and quantitative evaluation of physical damage attributes under a complex background are realized.

The invention discloses a dual robust principal component analysis image denoising method based on a graph, and belongs to the technical field of image processing. The method comprises the following steps: flattening an image to be processed to construct a data matrix in a column vector form; constructing a graph structure by using a K nearest neighbor method, and generating a graph Laplacian matrix based on the graph structure; jointly considering an image reconstruction error, a sparse noise item, a linear mapping error item and a graph structure regular item, and constructing an optimization model; and carrying out variable alternating optimization by adopting an augmented Lagrange multiplier method and an alternating direction solution method, and obtaining an image denoising result according to the low-rank principal component. According to the method, image structure information and a double constraint mechanism are introduced into a robust principal component analysis framework, so that the detail retention capability and the structure consistency of the image are effectively enhanced, the robustness and the visual quality of image denoising are improved, and the method is suitable for application scenes such as image processing, video monitoring and target detection under complex backgrounds.

The invention provides a wind power converter DC support capacitor health state real-time detection method, and belongs to the technical field of wind power converter DC support capacitor health state detection. Comprising the following steps: collecting a ripple voltage signal and a temperature signal of a DC bus capacitor in real time; variational mode decomposition is carried out on the ripple voltage signals, and time domain features, frequency domain features and nonlinear features of multi-mode components are extracted; meanwhile, mean value features and dynamic change rate features are extracted from the temperature signals; on the basis of robust principal component analysis, dimension reduction and feature enhancement are carried out on the extracted features, and a state feature matrix is generated; inputting the state feature matrix into a pre-trained intelligent learning model, and outputting a health state classification result of the DC support capacitor; according to the method, the health state of the capacitor is accurately reflected in real time, degradation and faults are found in time, and the method can be widely applied to real-time monitoring of the wind power converter.

The invention discloses a method for generating a cutting track of a convex riser on the surface of a casting based on machine vision, and the method comprises the steps: carrying out the self-adaptive anisotropic downsampling of an original point cloud, so as to reduce the data volume while keeping key geometric features; performing robust principal component analysis plane fitting on a point cloud downsampling result to obtain a reference plane parameter of the casting body under a preset coordinate system; differential geometry-based riser region segmentation is carried out on the point cloud downsampling result according to the reference plane parameters, and convex riser region point clouds are obtained; according to the point cloud of the convex riser region, constructing a triangular mesh curved surface about the surface of the convex riser; performing plane cutting on the triangular mesh curved surface to obtain a cutting contour line; based on a beam projection method, an initial cutting path is obtained according to the cutting contour line; and constructing a cutting path optimization energy functional so as to optimize the initial cutting path to generate an optimal cutting track of the convex riser on the surface of the casting. The problems that a traditional manual cutting mode is low in cutting efficiency, poor in precision, poor in consistency and large in potential safety hazard are solved.

The invention discloses a fluctuation signal coupling noise suppression method based on a neural network, which is a coupling noise suppression method in an auto-encoder fluctuation signal based on adjacent position constraint and first arrival pre-sample guidance, and is applied to the field of fluctuation signal data processing. In order to solve the problem that effective signals are prone to being greatly damaged when the coupled noise is suppressed in the prior art, the invention designs a method for representing the characteristics of waveform similarity of adjacent spatial positions of the coupled noise and fluctuation signals based on the weight of a noise sample before first arrival, and under a robust principal component analysis framework, on the basis of a model and data combined driving strategy. Effective signals and strong coupling noise are separated in an unsupervised mode, and damage to the effective signals is reduced.

The invention discloses a face recognition method based on adaptive sine angle robust principal component analysis, and belongs to the technical field of face recognition, and the method comprises the steps: obtaining a face image, and carrying out the preprocessing of the face image; constructing a robust principal component analysis model based on an adaptive sine angle, and training the robust principal component analysis model by using the training set; solving an optimal projection matrix by using a non-greedy iterative algorithm; mapping the original face features to a discrimination space by using the optimal projection matrix; and performing face recognition by using a nearest neighbor classifier, and performing robustness evaluation on a recognition result. According to the method, the adaptive sine angle is added, abnormal values can be effectively suppressed in the face of noise data, the influence of the abnormal values on the projection direction is reduced, and the method plays a key role in improving robustness.

The invention discloses an intelligent nursing method and system for infants and disabled old people. The method comprises the steps that vital sign data, behavior state data and an environment data set of a user are collected in real time; carrying out normalization processing on the collected data, carrying out feature extraction by adopting RPCA robust principal component analysis to obtain an intelligent nursing data set, and dividing the intelligent nursing data set into a training set, a verification set and a test set; constructing a TimeInf-LSTM time sequence anomaly detection model based on the intelligent nursing data set, and outputting a user demand analysis result; generating a personalized nursing instruction according to the user demand analysis result; a PLO aurora optimization algorithm improved based on lens imaging reverse learning and SPM chaotic mapping is adopted to optimize hyper-parameters of the TimeInf-LSTM model; according to the invention, multi-dimensional dynamic monitoring and accurate analysis of the user state are realized, and the accuracy, reliability and long-term applicability of nursing demand identification are significantly improved.

According to the multi-frame reverberation suppression method based on combination of the alternating direction multiplier method and the low-complexity high-order nonlinear cumulant, underwater reverberation in inter-frame low-rank slow-change components in reverberation is suppressed through the alternating direction multiplier method; water surface reverberation, volume reverberation and noise in inter-frame sparse fast-change components are suppressed by using low-complexity high-order nonlinear cumulants, and continuous moving target bright spots are enhanced, so that an angle-distance sonar two-dimensional image with lower reverberation intensity and clear continuous target bright spots is obtained. Compared with the reverberation suppression performance of an existing low-complexity high-order nonlinear cumulant and multi-frame accumulation after processing of an existing robust principal component analysis method, the multi-frame reverberation suppression method combining the alternative direction multiplier method and the low-complexity high-order nonlinear cumulant has a better reverberation background suppression effect.

The invention relates to the technical field of underwater target detection, and discloses a reverberation suppression method and device based on tensor robust principal component analysis. The method comprises the following steps: arranging frame signals received by a sonar receiving array according to a time sequence to obtain a distance-azimuth echo sequence; performing histogram matching processing on the distance-azimuth echo sequence, aligning the intensity distribution of each frame to a reference frame, and obtaining a distance-azimuth echo sequence after matching processing; after the distance-azimuth echo sequence after matching processing is preprocessed, low-rank component estimation initialization processing is executed; and after the initialization processing is completed, executing an adaptive threshold iterative optimization process, and iteratively solving an iterative formula by using a scale gradient descent method to obtain a low-rank component and a target component. By applying the method, the problems of high target false alarm rate and insufficient detection precision caused by reverberation interference in a complex environment can be solved.

The invention relates to the technical field of industrial process soft measurement and quality control, and discloses a paper quality prediction method based on multivariate statistical latent variable fusion and space-time transformation, which comprises the following steps: acquiring space-time sequence data of a multi-source sensor in a papermaking process, constructing a space-time diagram structure reflecting a topological relation of equipment, and preprocessing. Then, multi-view latent variables are extracted through non-negative matrix factorization, independent component analysis and robust principal component analysis, attention fusion is conducted on the latent variables through an LV fusion module, and fusion latent variables are obtained; and inputting the fusion latent variable and original node data into a multi-scale convolution auto-encoder to obtain spatial feature embedding, and inputting the spatial feature embedding and the fusion latent variable into a space-time Transform module together to realize joint modeling of space correlation and time dependence. And finally, outputting a paper quality predicted value through a linear regression module. The method can achieve the accurate prediction of the paper quality under a high-dimensional and multi-noise working condition, and is suitable for online monitoring and modeling updating.

The invention discloses an aviation aluminum plate wave field image damage monitoring method based on non-convex total variation regularization RPCA. The method comprises the following steps: acquiring Lamb wave field image data of a to-be-detected aviation aluminum plate by using a sensor array; decomposing the mapped Lamb wave field image data by using a non-convex total variation regularization robust principal component analysis algorithm, and solving sub-problems by using an alternating direction multiplier method to obtain a low-rank matrix representing background wave field information and a sparse matrix representing damage abnormal values; performing post-processing including sparse response normalization processing, adaptive threshold segmentation, connected region filtering and morphological repair on the sparse matrix to extract a damaged region; according to the extracted damage area, the damage condition of the aviation aluminum plate to be detected is evaluated and positioned. According to the invention, the accuracy and real-time performance of damage identification can be improved.

The application relates to the technical field of underwater communication and signal processing, and provides an anti-interference method for underwater power carrier communication, which comprises the following steps: acquiring a power carrier time-frequency signal received by an SRM and synchronously collecting a control instruction stream, wherein the power carrier time-frequency signal is an OFDM signal; determining a state perception weight matrix based on a relay action time window, wherein the relay action time window is determined based on the control instruction stream; determining a joint optimization problem provided with a phase distortion correction operator based on the power carrier time-frequency signal, the state perception weight matrix and a frequency domain confidence matrix in combination with a double-weighted robust principal component analysis algorithm, wherein the frequency domain confidence matrix is determined based on the high-frequency attenuation characteristics of an underwater umbilical cable; and solving the joint optimization problem based on a linearized alternating direction multiplier method to determine a target power carrier signal. The application solves the problem that reliable communication cannot be realized under the double constraints of strong interference and severe phase jitter in the related art.

This invention discloses a method, system, storage medium, and electronic device for suppressing mutual interference in spaceborne synthetic aperture radar (SAR) scattered waves based on matched filtering. The method includes the following steps: estimating relevant parameters of mutual interference based on prior orbit information and time-frequency relationships; performing range pulse compression on the interference signal in the echo data; detecting whether the range pulse contains interference using the maximum eigenvalue sequence detection method; calculating the relative signal-to-interference ratio (SIR) of the interference pulse; reconstructing and extracting the interference signal under high SIR and low SIR conditions using the eigenspace projection method and robust principal component analysis method, respectively; filtering it from the echo data; and then performing inverse range compression on the data to restore the interference-free normal echo. This invention effectively utilizes the compression gain of mutual interference, thereby more effectively extracting mutual interference signals from SAR echoes and achieving high-precision interference suppression while better protecting useful signals.

The present application relates to the technical field of marine remote sensing and water depth inversion, and discloses a water depth inversion method fusing multi-temporal information and geographic spatial perception mechanism, single-temporal remote sensing images of the same water area at different time points are acquired first to construct a multi-temporal remote sensing image sequence, then a robust principal component analysis method RPCA is used to fuse and process the multi-temporal remote sensing image sequence to obtain a fused image; then a plurality of core units are cascaded together to complete step-by-step feature extraction, and the last core unit is connected to a regression head to output continuous water depth values, thereby completing the construction of a water depth inversion network model; finally, the water depth inversion network model is trained by using the fused image combined with the geographic spatial position, and the trained water depth inversion network model is used to perform water depth inversion on the data-processed remote sensing image to be detected.

The application discloses an improved RPCA infrared small target detection method based on a local signal-to-clutter ratio. Firstly, a robust principal component analysis is performed on an infrared image containing a small target, and the image is decomposed into a background matrix and a target matrix. The background matrix captures the low rank of the image and describes the background information of the image, and the target matrix utilizes sparsity and effectively represents target information in the image. A local window is introduced in the target matrix, and a local signal-to-clutter ratio of the image in the window is calculated. The local signal-to-clutter ratio considers the difference between the target and the background, and is helpful to distinguish the real target from false alarms. By comparing the local signal-to-clutter ratio difference between the small target and the background in the infrared image, the real target can be accurately identified. The robust principal component analysis can effectively separate the target and the background in the image, thereby improving the detection rate of the small target.

This invention belongs to the field of image processing and computer vision technology, specifically disclosing an infrared small target detection method based on deep unfolded networks and learnable sparse transforms. This invention maps the single-iteration steps used in traditional robust principal component analysis (PCA) for updating the background, target, multipliers, and reconstructing the image to an end-to-end trainable model composed of multiple cascaded stages in a deep neural network. The model proposes a multi-channel attention-supervised transmission enhancement module, which drives attention selection by introducing multi-channel feature fusion and actively blocks the cross-stage propagation of errors by truncating low-response channels and retaining high-response channels. A learnable sparse transform target extraction module is introduced to achieve accurate target separation under extremely low signal-to-noise ratio (SNR) conditions. A dynamically attention-guided feature enhancement module is constructed to capture spatially changing background patterns, improving the fidelity of image reconstruction. This invention achieves high-precision, low-false-alarm-rate target detection under extremely low SNR conditions.

The application relates to the technical field of SCGD-OES online detection, and specifically provides a multi-element online quantitative method, system, equipment and medium of a high-salt matrix solution, which comprises the following steps: synchronously collecting multi-modal data; constructing a state vector according to the multi-modal data; calculating a stability index through robust principal component analysis and Mahalanobis distance; combining with working condition parameters to distinguish the working condition and the stability state; and adaptively selecting a quantitative analysis mode; under the working condition constraint, inputting spectral line characteristics into a segmented weighted least square regression model which is fused with monotonicity constraint, multi-spectral line consistency and self-absorption / quenching penalty term, compensating for the matrix effect and calculating the concentration; and finally, fusing the stability index, spectral line consistency deviation and model residual error to calculate the comprehensive confidence, and outputting the concentration and the confidence. The application improves the accuracy, robustness and result reliability of multi-element online analysis under high-salt, dynamic and nonlinear complex working conditions.

The present invention discloses a robust principal component analysis method for application scenarios with a large amount of unsupervised noise data. By adaptively assigning a weight of 0 or 1 to each sample to discard outliers and calculating the average value through the selected normal samples, the resulting projection space is more reasonable and representative for most samples. In order to maintain the local smoothness of the sample distribution, the present invention learns the affinity relationship of the samples in the subspace, maintains the local smoothness of the sample distribution, and finally proposes an iterative algorithm to solve the model. Based on robust principal component analysis and combined with adaptive outlier detection, the present invention can be applied to multiple fields such as image recognition, data compression, pattern recognition and classification, machine learning, statistics and data analysis.

This invention relates to the field of image processing technology, and more specifically, to an online detection method for surface defects of hexagonal screws based on machine vision. The method includes: acquiring an image of the hexagonal slot of the hexagonal screw, and dividing the image into multiple image blocks along a preset grid; for each image block, constructing an equilateral distance normal gradient deviation index and an equilateral distance six-fold symmetry intensity deviation index; the equilateral distance normal gradient deviation index is used to characterize the degree to which the local gradient vector deviates from the normal edge normal characteristics. This invention, by constructing an in-band joint anomaly response index, spatially adaptively adjusts the regularization parameters of the Robust Principal Component Analysis (RPCA) algorithm, achieving effective suppression of normal hexagonal step textures and high-sensitivity extraction of real minute defects, perfectly balancing the anti-interference capability and extremely high accuracy of industrial inspection.

The present invention discloses a multi-view clustering image segmentation method and system based on embedded approximate learning. The method includes the following steps: obtaining an image multi-view data set to be subjected to image segmentation; inputting the image multi-view data set into a pre-constructed clustering model, iteratively updating the parameters of the clustering model according to an optimization objective, and when the change value of the optimization objective is less than a set threshold, obtaining an optimized clustering model; inputting the image multi-view data set into the optimized clustering model, outputting a clustering result, thereby completing multi-view subspace clustering; and performing image segmentation according to the clustering result to obtain an image segmentation result. By combining self-representation learning, robust principal component analysis technology, and Grassmann manifold space approximate learning, the present invention avoids an increase in computational cost caused by eigenvalue decomposition of a similarity matrix during the optimization process, and improves the efficiency and quality of clustering.

The invention discloses an image data uncertainty quantification method based on low-rank sparse matrix factorization, which comprises the following steps: an image data preprocessing stage: collecting and matrix image data: flattening a plurality of polluted images into column vectors and arranging the column vectors in sequence to form an observation matrix; the number of rows of the observation matrix corresponds to the number of pixel points of the polluted images, the number of columns of the observation matrix corresponds to the number of the polluted images, missing marking, normalization and observation probability estimation are carried out on the observation matrix, and then an observation index set is randomly divided into a training set and a calibration set in proportion; in the training stage, a non-distributed robust principal component analysis method based on a conformal prediction framework is used for processing the training set, low-rank estimation, residual standard deviation and a weighted threshold are obtained through training, and a low-rank structure uncertainty recovery model is obtained; a test stage: performing low-rank recovery and uncertainty quantification on new polluted image data by using the low-rank structure uncertainty recovery model; the method is suitable for data recovery and uncertainty quantification in a complex environment.

The application discloses a kind of multi-channel SAR-GMTI methods based on improved robust principal component analysis, belong to signal processing technical field, including: obtaining the original echo signal of multi-channel synthetic aperture radar, and the original echo signal is imaged and handled, and image domain data is obtained;After image registration based on image domain data, each channel corresponding registration image domain data in each pixel point is sequentially taken as reference pixel, and the residual of each channel corresponding registration image domain data is compensated in combination with the auxiliary pixel in the preset range of reference pixel;Residual-compensated image domain data is detected using an improved robust principal component analysis model;Radial velocity estimation and target positioning are carried out on moving target using adaptive matched filtering algorithm.The application solves the deficiency encountered when processing local error by traditional channel equalization registration algorithm, and has good target detection capability.

The application discloses an image recognition method based on distributed machine learning and privacy protection technology, and the method is as follows: a saliency mapping attack method based on Jacobian matrix is adopted to generate an adversarial sample; reversible down-sampling and up-sampling layers are applied to a client and a server respectively, four input tensors are obtained according to color channels, and a convolution function is used to combine the input tensors to obtain a combined tensor; a Lipschitz continuous gradient function is used to solve a model obtained by a proximal gradient algorithm, robust principal component analysis and global optimization analysis to obtain a restored image; convolution and strategy are used to realize synthesis of an activation map, and linear summation is performed on the activation map to obtain a combined image; an attack image, the combined image and the restored image are recognized respectively; an adaptive weighted average method is used to sum the recognition results of the three images to obtain an image recognition result. The application can protect image privacy and improve the recognition accuracy of an image recognition model.

The invention discloses a near-infrared video heart rate detection method based on region selection and robust principal component analysis, and the method comprises the steps: capturing a face video through a camera, obtaining the feature points of a cheek region, and selecting a region of interest; pixels in a plurality of regions of interest of each frame of image are averaged and connected frame by frame, and a plurality of rPPG signals containing pulse wave information can be obtained. The rPPG signal is preprocessed, and a pre-reference signal is constructed to carry out preferential selection on original signals of different regions of interest, so that a high-quality rPPG signal is obtained. The selected signals are subjected to noise reduction processing through a robust principal component analysis algorithm, and contained pulse wave information is highlighted. And the detected heart rate is determined through frequency domain analysis of the signals. According to the method, heart rate detection is successfully completed through a common consumption-level near-infrared camera in a non-contact scene.

The application discloses a kind of building thermal insulation layer bonding surface radar data acquisition methods, belong to radar data acquisition technical field, for building thermal insulation layer bonding surface data acquisition, including based on the principle of penetration imaging radar acquisition building thermal insulation layer bonding surface data, radar echo data is handled using range migration imaging and robust principal component analysis method, using edge detection method is connected and region filling to image edge, generate building thermal insulation layer bonding surface binary image, based on the method of the application builds building thermal insulation layer bonding surface radar data acquisition experimental platform.The application is collected to building thermal insulation layer bonding surface by radar data, and radar echo and image are optimized, provide scientific basis for evaluating the structural stability and potential risk of bonding layer, improve radar data acquisition precision and image definition;Building thermal insulation layer bonding surface radar data acquisition experimental platform is designed, realizes the method of collecting building thermal insulation layer bonding surface data in laboratory environment.

This invention relates to a wavelet-based sound denoising method and system based on robust principal component analysis. The method includes: acquiring a noisy sound signal; separating the noisy sound signal using robust principal component analysis to obtain a separated noisy sound signal; filtering the separated noisy sound signal using a wavelet threshold denoising algorithm to obtain a filtered noisy sound signal; and performing low-pass filtering on the filtered noisy sound signal to obtain a denoised sound signal. This invention can improve the quality of sound signals.

The application discloses a thermal infrared small target detection method based on a non-overlapping block space-time tensor model. The method comprises the following steps: (1) using an original thermal infrared image sequence to construct a non-overlapping block space-time tensor according to a specific rule; (2) converting the infrared weak small target detection task into a tensor robust principal component analysis problem, and building a low-rank sparse tensor decomposition framework; (3) using the multi-mode expansion of the tensor and a Laplace function to obtain a non-convex low-rank estimation norm of a background tensor; (4) using a re-weighting strategy to obtain a sparsity estimation of a target tensor; (5) using a tubular sparse regularization term to measure a sparse structure component, and using a Frobenius norm to measure noise; and (6) optimizing and solving the model based on an improved ADMM algorithm, obtaining a target tensor component, reconstructing a target detection result image, and realizing infrared weak small target detection. The low-rank sparse tensor decomposition framework and the improved ADMM optimization algorithm can effectively realize the detection of infrared weak small targets.

The invention discloses a singular spectrum prior-based low-rank tensor denoising method, singular spectrum prior-based low-rank tensor denoising equipment and a medium, and relates to the technical field of data denoising processing, and the method comprises the steps: inputting a to-be-denoised data tensor and singular spectrum prior into a preset low-rank tensor denoising model, updating a previous noise tensor through the low-rank tensor denoising model, and obtaining a to-be-denoised data tensor; obtaining a current noise tensor; updating the previous low-rank tensor to obtain a current low-rank tensor; updating the previous frequency domain error matrix to obtain a current frequency domain error matrix; updating the last Lagrangian multiplier to obtain a current Lagrangian multiplier; calculating a residual error between the to-be-denoised data tensor and the current low-rank tensor; and outputting the current low-rank tensor under the condition that the residual error is determined to be smaller than the preset value. The low-rank tensor denoising method and device are used for solving the problem that in the prior art, when tensor denoising processing is carried out based on robust principal component analysis, the denoising accuracy is poor, and the accuracy and robustness of low-rank tensor denoising are improved.