53 results about "Linear reconstruction" patented technology

The invention discloses a task unloading method based on user experience in an edge computing network. The method comprises the following steps: a scheduler receives a task unloading request submittedby a user; a task unloading strategy model is jointly optimized based on time and energy expenditure; the model is solved and calculated based on a branch and bound algorithm of linear reconstruction; the task unloading strategy model is optimized based on time and energy expenditure. Based on the results of the calculation, the decision is made whether the task is unloaded to the edge server, and the tasks unloaded to the edge server are communicated and the computational resources are allocated. The invention combines two key indexes of optimization time delay and energy consumption, and solves the problem by branch and bound algorithm based on linear reconstruction technology, which can reduce the energy consumption of mobile intelligent device under the condition of ensuring the completion of the task, reduce the task processing time delay, and achieve the purpose of maximizing the optimization of user experience.

The invention provides a method for reconstructing face image super-resolution by utilizing position image blocks. The method comprises the following steps: dividing a low-resolution face image and face images in high-low resolution training sets into mutually overlapped image blocks; calculating the optimal value of each divided image blocks input in the low-resolution face image during the linear restoration of the position block of each sample image in the low-resolution training set; replacing the position blocks of the sample images in the low-resolution training set by the position blocks of the sample images in the high-resolution training set, which correspond to each position block of the sample images in the low-resolution training set, and compositing image blocks of high-resolution in a weighting manner; and splicing the composited image blocks of high-resolution into a whole image according to the position of the image blocks in the face image. The method which reconstructs a high-resolution image block in the same position by utilizing the image block in the same position of each sample image in a training set directly has the advantage that the manifold learning step or the feature extraction step which are common in similar algorithms are avoided, thereby greatly saving operation time, reducing complexity; and the quality of the composited high-resolution image is improved.

The invention provides an interactive image segmentation method for multiple foreground targets. The method comprises the following steps of: performing linear reconstruction on pixel colors in an image local window, and repeatedly modifying color reconstruction coefficients by using linear projection; repeatedly performing the linear reconstruction on pixel class label vectors in the image localwindow by using the modified color reconstruction coefficients, and estimating to acquire local reconstruction errors; accumulating the local reconstruction errors to acquire a global reconstruction error; building an interactive image segmentation model of the multiple foreground targets; and performing cluster analysis on the same class of pixels which are labeled by a user to acquire a clustering center; acquiring a group of polynomial functions by adopting regression estimation by taking the clustering center as a training sample; mapping the unlabelled pixels by using the polynomial functions to acquire an initial solution; solving the segmentation model; and determining the class attribution of the unlabelled pixels, and outputting a segmentation result. The interactive image segmentation method has a wide application prospect, and the problem that the multiple foreground targets are difficult to segment simultaneously in the prior art is solved.

The invention discloses a linear reconstruction method of standard number 12 lead electrocardiogram segments based on self-adaptive electrocardiosignal region segmentation. The method comprises the following steps of self-adaptive electrocardiosignal region segmentation, wherein standard number 12 lead electrocardiosignals are subjected to self-adaptive segmentation, and according to wave characteristics of the electrocardiosignals and different heartbeat stages, the standard number 12 lead electrocardiosignals are divided into a head and tail segment, an ST-T segment, an R-P segment and a QRSsegment; linear regression training and reconstruction, wherein a least squares criterion is used for conducting linear modeling and linear reconstruction on an existing electrocardio sample; sub region electrocardio deserialization, wherein due to the fact that signals to be reconstructed are subjected to linear reconstruction respectively after electrocardiosignal region segmentation, the reconstructed electrocardio sub region sequences need to be restored to common sequential electrocardio signals. The linear reconstruction method of the standard number 12 lead electrocardiogram segments based on self-adaptive electrocardiosignal region segmentation is fast and accurate.

The invention belongs to the technical field of image fusion, and discloses an infrared and visible light image fusion method and system, computer equipment and application, and the method comprises the steps: carrying out super-resolution reconstruction of an original infrared image based on a convolutional neural network to acquire a reconstructed high-resolution infrared image; anisotropicallydiffusing the visible light image and the high-resolution infrared image respectively to obtain a corresponding base layer and a detail layer; obtaining a maximum feature vector by utilizing KL transformation to fuse detail layers, and fusing a base layer by utilizing saliency adaptive extraction; and carrying out linear reconstruction on the fused detail layer and the base layer to obtain a finalfused image. A large number of experiments show that compared with four fusion algorithms, the algorithm provided by the invention has a better fusion effect no matter from subjective judgment or from objective evaluation indexes, and has richer detail information and contour textures while enabling a fused image to have higher definition.

The present invention relates to an active learning big data mark method and system. The method comprise: performing linearity reconstruction of each data point according to the anchor data set to be marked in the data set to be marked; calculating the distance between data points; taking the distances as the weight construction regular items of reconstruction parameters, wherein the distances are inversely proportional to the reconstruction parameters; constructing and obtaining a data mark model to perform corresponding processing and correction of the data mark model; and performing optimizing and solution to determine the anchor data for active learning. Because the distances are inversely proportional to the reconstruction parameters, the data mark model is sensitive to the distance among data points, and it is easier to determine whether the corresponding data points have representativeness or not in the solution and optimization process according to the size of the infinite norm value to accurately screen out the anchor data set for active learning in the data set to be marked so as to improve the big data anchor mark accuracy.

The invention discloses an anchor graph structure-based semi-supervised data classification method of double Laplacian regularization. The method mainly comprises the following steps: firstly, carrying out clustering on a data set to obtain anchor point data which can approximatively indicate the entire data set, and calculating linear reconstruction weights between sample points and adjacent anchor points thereof through an FLAE method; then respectively constructing Laplacian regularization terms on the anchor points and Laplacian regularization terms on the sample points on the basis of a weight matrix between the sample points and the anchor points, and establishing an anchor graph structure-based semi-supervised classification model of double Laplacian regularization; and finally, using a zero-gradient method to parse and solve the model to obtain category soft-labels of the anchor point data, and using feature codes of unlabeled samples to linearly combine the category soft-labels of the anchor points, and discriminating categories of the unlabeled samples. Double-Laplacian-regularization constraints established by the method can better describe graph structure information among the samples, and thus realize higher classification and discriminating ability, and the method has very good application prospects.

The invention provides a convolutional neural network (CNN) initialization method based on pre-training model filter extraction and relates to the technical field of video processing. The method extracts a filter parameter in a pre-training model by using a minimum entropy loss and a minimum reconstruction error in order to initialize a target task network model and achieve the small and medium-sized network initialization that meets practical application problems. By using the minimum entropy loss and the minimum linear reconstruction method, the method extracts the filter parameter from thepre-training model to initialize the target task network model, does not require that the target task network structure is consistent with the pre-training network structure, flexibly design the network structure of the target task according to practical applications so as to meet the memory overhead and calculation speed requirements in practical application problems.

The invention discloses a vehicle bearing fault diagnosis method based on CEEMDAN and APSO-SVM, and the method comprises the steps: collecting a vibration signal of a rolling bearing, measuring the parameters of the bearing, carrying out the decomposition of the collected vibration signal of the rolling bearing, carrying out the screening of IMF modal components after the decomposition, carrying out the linear reconstruction of the screened components, and removing invalid information; then singular entropy, power spectrum entropy and energy entropy calculation is carried out on the screened IMF modal components, and principal feature extraction is carried out on the reconstructed signals by using WPCA based on calculation results to obtain feature vectors; making the feature vectors into a training set and a test set of an SVM, adding a category label, and constructing and optimizing an SVM classifier model on the basis; and finally, performing fault diagnosis on the vehicle bearing by using the optimized SVM classifier. According to the method, time domain and frequency domain information is comprehensively considered, fault features can be accurately extracted, the problem that the SVM optimal parameters are difficult to manually select is solved, popularization in engineering application is facilitated, and practicability is high.

The present invention provides a feature selection method based on optimal reconstruction, the steps of which are: 1) represent each original feature in the data set as d-dimensional data as a data matrix X, where d>1; 2) for the above data matrix X establishes an optimal linear reconstruction model, and the optimization target of the model is the representation matrix B; 3) Transpose the above data matrix X to obtain the feature matrix F, and clear the representation matrix B; 4) Use iterative alternate optimization Solve the above-mentioned optimized linear reconstruction model after step 3) in a way to obtain the optimal representation matrix B*; 5) select the optimal k-dimensional feature subset that can represent all d-dimensional features according to the optimal representation matrix B* , where k

The invention discloses a hyperspectral image recognition method and device based on spatial spectrum group covariance features and a computer readable storage medium. The method comprises the following steps: acquiring a hyperspectral image to be identified, mapping the hyperspectral image to be identified to a Riemann space, and calculating the Riemann distance between data points in the Riemann space; and on the basis of a Riemann cutting space and a local linear measurement criterion, performing adaptive neighborhood calculation on each data point; projecting the data points in the Riemann space to a Riemann local tangent space according to the self-adaptive neighborhood information of each data point; obtaining the low-dimensional image features of the to-be-recognized hyperspectral image by performing eigenvalue decomposition on the reconstruction weight matrix obtained in the linear reconstruction process, and recognizing the surface features of the hyperspectral image based on the low-dimensional image features, so that the surface feature recognition accuracy of the hyperspectral image is effectively improved.

The invention discloses an X-ray image linear reconstruction method, which comprises the steps: obtaining an X-ray projection image, introducing a regular term based on total variation prior, and constructing a reconstruction target function; introducing super prior parameters, and constructing a layered Bayesian model; introducing a split variable by using a variable splitting method, and separating a data fidelity term and a TV regular term to obtain joint probability density distribution in a split form; defining a super-prior variable based on Jefferys prior to obtain condition distribution of each variable; iteratively updating the super-prior parameters, and solving conditional distribution of split variables containing TV regular terms; approximating the full-condition probability density distribution of the to-be-solved parameter by utilizing the low-rank property of the forward matrix, and calculating the target distribution of low-rank approximation to obtain a closed solution about the to-be-solved parameter; and calculating the mean value of the sampling samples, and estimating the to-be-solved parameters. According to the method, the problems of high calculation overhead and the like in solving the large-scale linear inverse problem can be effectively solved.

The invention discloses a single-sample face recognition method based on block linear reconstruction discriminant analysis. The method comprises the steps of firstly, partitioning each face training image, then expressing any face image block as a linear combination of k1 intra-class nearest neighbor image blocks, and meanwhile expressing any face image block as a linear combination of k2 inter-class nearest neighbor image blocks; respectively solving the intra-class representation coefficient and the inter-class representation coefficient by using a least square method, and calculating the intra-class reconstruction divergence and the inter-class reconstruction divergence of the sample; solving an optimal projection matrix by maximizing the ratio of the inter-class reconstruction divergence to the intra-class reconstruction divergence, and extracting features of the training sample set and the to-be-identified sample by using the projection matrix; and finally, constructing a discrimination criterion of the class label of the to-be-recognized face image, and judging the class label of the to-be-recognized face image. According to the method, the problem of single-sample face recognition can be effectively solved, the influence of changes of image illumination, face postures, expressions and the like on the recognition effect can be effectively avoided, and the recognition rateis increased.

The invention relates to a high-precision format simulation method for one-dimensional ice-water coupling motion, and belongs to the field of river ice disaster forecasting. The method comprises the steps of boundary condition acquisition, equation control, ice-water coupling motion unified discretization, surface flux calculation, shallow water wave velocity calculation under the ice-water coupling condition and a high-precision format. According to the method, ice and hydrodynamic equations are uniformly dispersed into a high-precision Godunov format; an HLL approximate Rieman solution is adopted to calculate interface flux, equivalent water depth is adopted to replace section average water depth, piecewise linear reconstruction is carried out on basic variables of each unit in space, and a prediction-correction method is adopted in time, so that the precision of a numerical solution is integrally improved to a second order. The high-precision format simulation method for ice-water coupling movement has high prediction precision for the water level and ice thickness of the riverway in the flow ice period, particularly an equivalent water depth method is adopted to replace an average water depth method to calculate the wave velocity of shallow water waves, the water level prediction precision can be obviously improved, and research results are of great significance for prevention and control of ice disasters of the riverway.

The invention discloses a sensor noise and fault judging method based on sparse representation. The specific method includes the following steps: 1. an overcomplete atom library containing normal signals, noise and fault samples corresponding thereto is built through historical data; 2. based on a hypothesis that linear representation of unknown samples of some category can be effectively realized in a corresponding subspace by a plurality of samples of the category, sparse representation of a mixed signal collected by a sensor is performed with a built dictionary, i.e., atoms best matched with the signal to be decomposed is found out from the overcomplete atom library and are subjected to linear reconstruction to obtain a new representation mode; 3. a reconstruction error of the sample subjected to linear reconstruction is calculated, and an error of a new sample reconstructed by use of a data set of each kind of noise and fault samples is obtained; 4. data of the same kind of fault samples and data of the same kind of noise samples are used to perform training to calculate a corresponding reconstruction error value; and 5. noise and fault judgment is realized through calculation of the reconstruction error.

The invention discloses a robust face recognition method based on secondary cooperative representation identification projection. The method comprises the steps of: screening out K types of samples closely related to training samples through first-time cooperative representation; representing a linear reconstruction training sample through secondary cooperation to obtain a reconstruction coefficient; constructing the intra-class graph and the inter-class graph of the samples through the reconstruction coefficient to describe cohesiveness and separability of the samples; then obtaining a projection matrix by maximizing the inter-class divergence and minimizing the intra-class divergence at the same time, finally extracting the features of a to-be-identified sample and all the training samples by utilizing the obtained projection matrix, and judging the class label of the to-be-identified sample according to a classification criterion. According to the method, the training samples are reconstructed through cooperative representation, the problem of recognition errors caused by illumination, shielding, human face postures and expression changes can be effectively solved, the trainingsamples can be expressed more effectively and accurately, and the high-precision requirement for human face recognition in practical application can be met.

The invention discloses a method, a device and a storage medium for extracting news webpage content, which relate to the technical field of news webpage content extraction. The method comprises the following steps: obtaining HTML code of the webpage, HTML linear reconstruction of the webpage, removing HTML noise label, filtering and dividing data set, absorbing pseudo noise paragraph, and generating text paragraph. HTML linear reconstruction of web page linearizes the tree-shaped div tags nested with each other, and the linear structure can be conveniently located as a div tag, thereby eliminating the influence of nested tags on subsequent steps; HTML noise label removal will reduce the impact of noise text on paragraph clustering; Data set filtering partitioning further reduces the effectof noise on text segments; Absorbing pseudo-noise paragraphs increases the recall rate of text paragraphs. The method overcomes the shortcomings of specific crawling of specific websites and enhancesthe generality of extracting news page content. Compared with the existing technology, the method can extract news content accurately and efficiently, and has a good effect.

The invention discloses a transmission network utilization rate evaluation method based on power system timing coupling. The method comprises the steps of: obtaining system basic technical data, system operation constraint condition data and system operation prediction data; in a condition that the existing power supply installation and a wire frame are unchanged, considering (N-1) expected faultsof a power transmission device, constructing a reasonable utilization calculation model based on time series analysis, adopting a linear auxiliary method to introduce dual secondary variables for linear reconstruction so as to improve the model solution efficiency; and for a certain typical timing operation scene, maximizing the maximum electric quantity value reached by a power transmission device to be analyzed in the typical scene, selecting a plurality of typical scenes, the operation time frame of each scene being T0, repeating the steps mentioned above, analyzing the typical scenes, constructing a reasonable utilization index calculation formula to evaluate the utilization rate of the transmission network. The transmission network utilization rate evaluation method has the high adaptability to the operation analysis of the power system with high timing coupling.

The invention discloses a light-field image compression method based on linear reconstruction. The light-field image compression method comprises the steps of decomposing a light-field image into viewing angle image arrays and then dividing into A and B sets; at an encoding end, performing compression on viewing angle images in the A set by adopting a first video encoder, transmitting a code stream to a video decoder at the encoding end and a video decoder at a decoding end, obtaining a relationship between viewing angle images in the B set and the viewing angle images in the A set by combining the viewing angle images in the B set with viewing angle images reconstructed by the video decoder at the encoding end in the A set and utilizing a linear reconstruction theory of light-field viewing angle images, and transmitting the relationship to a second video decoder at the decoding end; at the decoding end, reconstructing the B set by utilizing the linear reconstruction theory of the light-field viewing angle images and combining with decoding results of the first and second video decoders at the decoding end; and re-forming the light-field image by utilizing the reconstructed A set and B set. According to the method, encoding data at the encoding end can be greatly reduced, and are reconstructed at the decoding end with a relatively good quality.

The invention discloses a circular workpiece surface grinding method. The circular workpiece surface grinding method comprises the steps of S1, demonstrating a scanning track; S2, calibrating a robotand a vision sensor; S3, collecting data: when the robot carries the vision sensor to scan the workpiece, the robot and the vision sensor transmit data to a computer within the same time period, and the data transmitted by the vision sensor comprises currently scanned workpiece contour data; S4, obtaining workpiece edge points; S5, reconstructing a three-dimensional point cloud of the workpiece; and S6, conducting polishing. According to the circular workpiece surface grinding method, the linear structured light sensor is used for scanning to achieve nonlinear reconstruction of a curve contoursurface; and moreover, each contour line is used for calculating and positioning the edge contour points, so that precise positioning and grinding of the workpiece contour can be achieved through point cloud reconstruction to be conducive to improving the workpiece grinding quality.

The invention discloses an uncertainty-constrained linear reconstruction method for a high-energy flash X-ray image. A two-stage framework is designed, the reconstruction process of a high-resolution image is constrained through uncertainty of high-stage target parameters so as to eliminate the phenomenon that a sampled sample is unstable due to noise in an optical path image, and a variable splitting form of posterior distribution is considered in the two-stage framework so as to explore more flexible and accurate sampling steps. A high-efficiency MCMC algorithm with a first-order truncation CG optimizer is adopted to approximate posterior distribution of target parameters, the algorithm sampling efficiency is improved, and finally a refining step is adopted to remove residual noise in a sample mean value. The algorithm is high in calculation efficiency, and the image reconstruction quality can be effectively improved.