131 results about "Subspace clustering" patented technology

The invention discloses a significant object detection method based on sparse subspace clustering and low-order expression. The method comprises the steps of: 1, carrying out super pixel segmentation and clustering on an input image; 2, extracting the color, texture and edge characteristics of each super pixel in clusters, and constructing cluster characteristic matrixes; 3, ranking all super pixel characteristics according to the magnitude of color contrast, and constructing a dictionary; according to the dictionary, constructing a combined low-order expression model, solving the model and decomposing the characteristic matrixes of the clusters so as to obtain low-order expression coefficients, and calculating significant factors of the clusters; and 5, mapping the significant value of each cluster into the input image according the spatial position, and obtaining a significant map of the input image. According to the invention, the significant objects relatively large in size in the image can be completely and consistently detected, the noise in a background is inhibited, and the robustness of significant object detection of the image with the complex background is improved. The significant object detection method is applicable to image segmentation, object identification, image restoration and self-adaptive image compression.

Unlike traditional clustering methods that focus on grouping objects with similar values on a set of dimensions, clustering by pattern similarity finds objects that exhibit a coherent pattern of rise and fall in subspaces. Pattern-based clustering extends the concept of traditional clustering and benefits a wide range of applications, including e-Commerce target marketing, bioinformatics (large scale scientific data analysis), and automatic computing (web usage analysis), etc. However, state-of-the-art pattern-based clustering methods (e.g., the pCluster algorithm) can only handle datasets of thousands of records, which makes them inappropriate for many real-life applications. Furthermore, besides the huge data volume, many data sets are also characterized by their sequentiality, for instance, customer purchase records and network event logs are usually modeled as data sequences. Hence, it becomes important to enable pattern-based clustering methods i) to handle large datasets, and ii) to discover pattern similarity embedded in data sequences. There is presented herein a novel method that offers this capability.

The invention provides a user electricity consumption relevant factor identification and electricity consumption quantity prediction method under the environment of big data. Multiple electricity consumption modes of users are mined and existing electricity consumption behavior analysis methods are expanded by applying a mass user electricity consumption characteristic subspace clustering analysis method based on the research of the user electricity consumption characteristic evaluation index by aiming at the characteristics that the big data relevant to electricity consumption quantity prediction are various, large in size, high in dimension and high in generation speed. Meanwhile, group division is performed on the users according to different electricity consumption modes, factors relevant to user group electricity consumption quantity are identified from the aspects of regional and industry economic data, weather conditions and electricity price by utilizing mutual information matrixes, and an electricity consumption quantity big data prediction model based on a random forest algorithm is constructed so that data driving of the whole process of electricity consumption prediction is realized, adverse influence on electricity consumption quantity prediction caused by difference of the electricity consumption modes can be avoided, and thus the method has relatively high prediction precision and is suitable for big data analysis and processing.

The invention discloses a data subspace clustering method based on multiple view angles, which comprises the steps of extracting multi-view-angle characteristics in a multi-view-angle database; for the multi-view-angle database, selecting a specific linear reconstruction expression method and determining a regularization constraint method corresponding to the linear reconstruction expression method; determining reconstruction error weight of each view angle characteristic in multi-view-angle characteristics; according to the selected reconstruction expression method and the obtained reconstruction error weights of different view angle characteristics, learning to obtain a linear expression matrix for reconstructing all samples in the multi-view-angle database, wherein the linear expression matrices are used for expressing a relationship among the samples in the database and element values are used for expressing reconstruction coefficients for corresponding samples in the line to reconstruct corresponding samples in the row; correspondingly processing the linear expression matrix to obtain an affinity matrix for measuring the similarity of the samples in the multi-view-angle database; and using a spectral clustering algorithm to partition the affinity matrix to obtain multi-view-angle data subspaces.

Embodiments of the invention provide a method, system, and computer readable medium for classifying network traffic based on application signatures generated during a training phase using a modified subspace clustering scheme based on feature vectors extracted from network flows in a training set generated by a particular application and applying the signatures to a new feature vector extracted in real-time from current network data. The newly extracted feature vector is projected into the subspaces and compared with the signatures.

The invention belongs to computer visual pattern recognition and a video picture processing method. The computer visual pattern recognition and the video picture processing method comprise the steps that establishing a three-dimensional space-time sub-frame cube in a video human action reorganization model, establishing a human action characteristic space, conducting the clustering processing, updating labels, extracting the three-dimensional space-time sub-frame cube in the video human action reorganization model and the human action reorganization from monitoring video, extracting human action characteristic, confirming category of human sub-action in each video and classifying and merging on videos with sub-category labels. According to the computer visual pattern recognition and the video picture processing method, the highest identification accuracy is improved by 16.5% compared with the current international Hollywood2 human action database. Thus, the video human action reorganization method has the advantages that human action characteristic with higher distinguishing ability, adaptability, universality and invariance property can be extracted automatically, the overfitting phenomenon and the gradient diffusion problem in the neural network are lowered, and the accuracy of human action reorganization in a complex environment is improved effectively; the computer visual pattern recognition and the video picture processing method can be applied to the on-site video surveillance and video content retrieval widely.

A system to automatically enhance, tag, classify, categorize, cluster and index products described in unstructured text-based electronic documents. The system and method incorporate the use of text normalization, regular expressions, product number matching rules, text segmentation, entity detection, language models, predictive modeling, hierarchal subspace clustering, formal concept analysis, and a weighted combination of all techniques to detect and infer knowledge extracted from a digital version of raw, unstructured product text. Knowledge extracted and inferred comprises knowledge units including: main conceptual entity, entity text patterns, product language models, and conceptual hierarchies. The extracted knowledge units are utilized to store and index products in a product knowledge database and the products and knowledge units are made available to users via a user interface.

The invention provides a subspace clustering method based on high-dimensional overlapping data analysis and relates to the technical field of machine learning. The method builds a weighted mixed normsubspace representation model for a data matrix that needs clustering; obtaining an optimized coefficient matrix in the weighted mixed norm subspace representation model by using a linear alternatingdirection method; establishing a similarity matrix based on the optimized coefficient matrix; dividing the similarity matrix into subspaces by using a spectral clustering algorithm to obtain an initial clustering results; establishing an overlapping probability model of the subspaces; applying the overlapping probability model to an initial subspace division result to determine the overlapping ofthe subspaces; verifying a subspace clustering result to obtain a final clustering result. The subspace clustering method based on high-dimensional overlapping data analysis can improve the density ofthe same subspace data and the sparsity of different subspace data, and improve the accuracy of clustering.

The invention discloses a multi-view subspace clustering method based on joint subspace learning, wherein the method comprises the steps: building a target function of multi-view subspace clustering based on joint subspace learning, searching low-dimensional embedded spaces of different view data features, carrying out the conversion and fusion of original features, and reducing the impact from the noise of original data; and data self-reconstruction learning is carried out in a low-dimensional embedded space, and a low-rank sparse self-reconstruction coefficient matrix with consistent views is searched, so that a more accurate similarity relationship of data is obtained, and the robustness of the algorithm is enhanced. According to the method, the joint subspace learning and the data self-reconstruction learning are unified into an optimization framework and are solved by using an alternate optimization method, and the joint subspace learning and the data self-reconstruction learningare mutually enhanced, so that the subspace clustering performance is greatly enhanced.

The invention provides a compressed video capture and reconstruction system based on structured sparse dictionary learning. The system comprises a structured sparse dictionary learning module, a video signal sensing module and a reconstruction processing module. The structured sparse dictionary learning module firstly acquires a training set through a sub-space clustering method, then, a dictionary is acquired through a linear sub-space learning method and minimized-block-relevant block sparse dictionary learning method, the sensing module projects video signals in an image block mode, and acquired data are finally decoded and reconstructed in the reconstruction processing module. Compressed sampling is provided, the distributed progressive structure of the video sampling process is combined, the reconstruction accuracy and efficiency are improved for the special structure of a structured sparse dictionary matrix, the sampling efficiency of the video signals is greatly improved, reconstruction gains are acquired compared with other methods under different sampling compression ratios, and meanwhile the good expandability is achieved.

The present invention provides a high-dimensional data subspace clustering projection effect optimization method based on dimension reconstitution. The high-dimensional data subspace clustering projection effect optimization method based on dimension reconstitution comprises the specific steps of: 1, searching a dimension subspace, i.e. confirming a target optimization subspace of which a two-dimensional projection effect needs to be improved and selecting a subspace with an excellent clustering structure; 2, constructing a reconstructed dimension set, i.e. transferring clustering information of the subspace with the excellent clustering structure to a reconstructed dimension; 3, constructing candidate optimal dimension subspace sets, i.e. carrying out free combination on each element of the reconstructed dimension set and the target optimization dimension subspace to generate the candidate optimal subspace sets; 4, screening an optimal dimension subspace set; 5, determining an optimal dimension subspace. According to the high-dimensional data subspace clustering projection effect optimization method based on dimension reconstitution, the reconstruction concept is creatively introduced into the high-dimensional data subspace, the clustering projection effect of the target optimization subspace is improved and enhanced by the reconstructed dimension with stronger clustering information, and the problem of distortion of the clustering projection effect of the high-dimensional data subspace on the two-dimensional plane is solved.

The invention discloses an SAR image segmentation system and a segmentation method based on immune clone and projection pursuit. The system comprises an image characteristic-extracting module, an initial label-selecting submodule, a projection direction-selecting submodule and a subspace clustering submodule, wherein the image characteristic-extracting module extracts the gray co-occurrence matrixcharacteristics, the wavelet characteristics, the brushlet characteristics and the contourlet characteristics of an input image; the initial label-selecting submodule clusters the image characteristics to acquire and transmit an initial label to the projection direction-selecting submodule for calculating a linear judgment analysis projection index and acquiring an optimal projection direction; the subspace clustering submodule projects the image characteristics in the optimal projection direction, acquires and clusters an optimal subspace to acquire a clustering label, returns the clusteringlabel to the initial label-selecting submodule for iteration, acquires a final clustering label corresponding to image pixels and acquires a final image segmentation result. The invention has the advantage of high segmentation accuracy and can be applied to civil and industrial fields or as martial reconnaissance means.

The invention discloses a medical image segmentation method and device based on multi-modal subspace clustering, and the method comprises the steps: 1, obtaining an original medical image, and carrying out the preprocessing; 2, performing convolution and pooling on the original medical image preprocessed in the step 1 through a convolutional neural network, and converting the original medical image into a linear feature matrix of the original medical image; 3, constructing a model based on a self-supervision multi-modal depth subspace clustering method, and carrying out model training; performing spectral clustering on the linear feature matrix of the original medical image obtained in the step 2 by using a trained self-supervised multi-modal depth subspace clustering method model to obtain clustered medical feature data; and 4, processing the medical feature data clustered in the step 3 to pixels the same as those of the original medical image through deconvolution and up-sampling ofa convolutional neural network to obtain a segmented medical image. The method is good in complex medical image segmentation effect and high in precision.

The invention discloses a hyperspectral image low-rank representation clustering method based on bilateral weighting modulation and filtering. The method comprises the following steps: calculating a low-rank representation coefficient of a hyperspectral image; combining similarity of a spectrum and the low-rank representation coefficient, calculating a bilateral weighting matrix; using the bilateral weighting matrix to modulate the low-rank representation coefficient; performing bilateral filtering on the modulated low-rank representation coefficient; using the filtered low-rank representation coefficient to establish a similarity image; using the similarity image in spectral clustering, to obtain a final clustering result. The method makes full use of spectrum similarity of hyperspectrum and space structure information. Compared with a conventional subspace clustering method, the method is higher in clustering precision, and has higher robustness on noise. The method can be widely applied in non-supervised classification in fields of territorial resources, mineral survey, and precision agriculture.

The invention discloses a multi-view data subspace clustering method based on diversity and consistency constraints. The multi-view data subspace clustering method comprises a step 1 of obtaining the diversity constraint of multi-view data subspace clustering; a step 2 of obtaining the consistency constraint of multi-view data subspace clustering; and a step 3 of obtaining a multi-view data subspace clustering model and solving the model. According to the technical scheme, multi-view data information is fully utilized, and the image clustering performance is improved.

The invention discloses a data processing method based on subspace clustering, comprising steps of extracting characteristic points from all data which need to perform subspace clustering, performing normalization process on all extracted characteristic points to obtain a characteristic point matrix, establishing an adjacent set Omega for every characteristic point x which has gone through the normalization processing, constructing a similarity matrix W between all characteristic points according to the adjacent set of every characteristic point x, calculating a Laplacian matrix L corresponding to the similarity matrix W among all characteristic points, performing spectrum clustering segmentation on each Laplacian matrix L to obtain a category label of every characteristic points, and realizing the subspace clustering process of all the data. The data processing method based on the subspace clustering can effectively perform clustering process on the big scale data while guaranteeing high accuracy, satisfies the need for processing data in big scale, and is applicable to the data processing fields like the image processing, computer vision and image movement segmentation.

The invention discloses a convex nonnegative matrix factorization method based on subspace clustering. The method comprises the implementation steps of (1) stretching an image in an original databaseinto vectors to compose an original data matrix; (2) performing convex nonnegative matrix factorization based on spectral clustering on the original data matrix, solving by use of two optimization methods to obtain a basis matrix and an encoding matrix; and (3) performing a clustering test of a k-means clustering algorithm on the encoding matrix, counting experimental results, calculating two measure criteria of clustering precision and normalized mutual information. According to the convex nonnegative matrix factorization method based on subspace clustering, compared with an existing method,subspace structural information inside data is explored and utilized, meanwhile the local subspace constraints exerted to the algorithm enhances robustness of the algorithm and improves the clusteringeffect of the image; and the method can be widely applied to the field of data mining and data analysis.

The invention discloses a sparsity subspace clustering method for distributed implementation. The method comprises the steps of distributing data to each computing node on a cluster consisting of a plurality of computers; then selecting data of a current computer and other computers by each computing node to compute a Lasso sparse reconstruction sub-problem until the problem is converged; after all sub-problems are computed in a labor division manner by all computing nodes, summarizing computing result vectors to a main process or a management node, and performing subsequent weighted undirected graph generation and spectral clustering processes; and finally obtaining classification numbers. For relatively common ADMM serial computation, the computing speed is remarkably increased without reducing the classification accuracy.

The embodiment of the invention discloses a subspace clustering method, device and equipment based on multiple views and a computer readable storage medium. The method comprises the following steps of: estimating a rank function by taking a kernel norm and a Forbenius norm of a combined matrix as regular terms based on an extracted data characteristic matrix of multi-view data, and introducing tensor constraint to construct an optimization objective function of subspace clustering of each view matrix; Solving an optimization problem of the optimization objective function to obtain a subspace representation matrix of each view; performing Calculating to obtain an affinity matrix of the multi-view data based on the subspace representation matrix of each view; And segmenting the affinity matrix by using a spectral clustering algorithm to realize multi-view subspace clustering. According to the method and the device, high-order correlation information among multiple views is fully utilized, so that the clustering precision of the multi-view data is favorably improved; The kernel norm union and the Forbenius norm of the matrix are used as regular terms to estimate the rank function, sothat the robustness of the algorithm is improved, and the multi-view data clustering performance is improved.

The invention discloses a data mining method and system for screening terrorist attack event criminal gangs. Through Analysis of relevant data and Data cleaning, Data preprocessing is achieved throughdata conversion, data fusion is added to enrich relevant characteristics, then useful information is mined to construct a standardized data set, criminal gang division and classification are achievedthrough a subspace clustering method, and then the relevant relation between a terrorist attack event and the criminal gang is determined through a relevancy model, so that the terrorist attack earlywarning effect is achieved.

The invention discloses a novel transductive semi-supervised data classification method and system, which integrate the unsupervised subspace feature learning, discriminant clustering and adaptive semi-supervised classification into a unified framework seamlessly, and perform semi-supervised learning based on low-dimensional manifold features of original data and discriminant subspace clustering results, and can be used for high-dimensional data representation and classification. Based on the above-mentioned joint model, the graph construction and label propagation process are also seamlesslycombined, and thus, an adaptive weight coefficient matrix based on the low-dimensional manifold features and soft category labels of unlabeled data can be obtained.

The invention provides a hyperspectral image unmixing method based on subspace clustering constraint, a computer readable storage medium and electronic equipment, solving the problem that the unmixingprecision is not high due to the fact that the complexity of a hyperspectral image ground object and spatial structure characteristics are not fully considered in an existing method. The hyperspectral image unmixing method comprises the following steps: 1) inputting a hyperspectral image Y; 2) embedding a subspace clustering method into a non-negative matrix factorization framework to obtain a joint unified unmixing framework capable of fully mining a data subspace structure; 3) iteratively solving each matrix parameter in the unified framework in the step 2) to respectively obtain an end member coefficient matrix B, an abundance coefficient matrix A and a space self-expression coefficient matrix S; 4) synthesizing an end member by using the end member coefficient matrix B obtained in thestep 3.2) to obtain a demixed end member matrix M; and 5) obtaining an end member matrix M and an abundance coefficient matrix A of the hyperspectral image Y to complete demixing of the hyperspectralimage Y.

The invention discloses a semantic image segmentation method and system based on deep learning and clustering, and the method comprises the following steps: S1, carrying out the convolution and pooling of an original image through a convolution neural network, and obtaining a linear feature matrix of the original image; S2, performing subspace clustering on the linear feature matrix to obtain clustered feature data; and S3, performing deconvolution and up-sampling on the clustered feature data, and processing the clustered feature data to pixels the same as those of the original image to obtain a segmented image. According to the method, the convolutional neural network (CNN) in the deep neural network is combined with subspace clustering, and the sparse subspace is used for replacing a full connection layer in the CNN, so that the problems of complex semantic image segmentation calculation, large data volume and poor information in the prior art are solved. A subspace clustering method is introduced into the neural network, so that a large amount of marking data required by the CNN during working is reduced, and unsupervised learning of the CNN neural network is realized.

The invention provides a data fusion-oriented iterative structured multi-view subspace clustering method and device and a readable storage medium. The method comprises the steps of constructing a multi-view subspace clustering ISSMSC model; carrying out solving and target optimization on the target function, obtaining the number k and dimensions of subspaces, segmenting data points into the subspaces, and achieving multi-view subspace clustering. According to the matrix of the method, the relationship between different clusters is reduced, and the relationship in the same cluster is enhanced.Comparison of adjacency matrices demonstrates the advantages of the model. According to the method, based on the self-expression characteristic of the data, the shared information among the views is explored, and the potential supplementary information among the views is utilized. In consideration of the influence of a segmentation matrix generated in the clustering process on adjacent matrix learning, a structured l1 norm is introduced in the learning process. In addition, an effective optimization algorithm is designed to solve the problem.

The invention relates to the field of computer vision application, and provides a multi-view subspace clustering method based on block diagonal representation and view consistency, which can improve the multi-view data set clustering precision, and comprises the following steps of: constructing a multi-view data matrix according to an input multi-view data set; constructing a shared consistency representation matrix according to the multi-view data matrix to obtain an objective function; transforming the objective function and obtaining an augmented Lagrangian equation of the objective function; obtaining an updated expression of a variable of the Lagrangian equation through an augmented Lagrangian equation; initializing variables, setting iteration conditions or iteration times, performing iteration updating on the variables according to an updating expression, and outputting a shared consistency representation matrix which meets the iteration conditions or the iteration times as an optimal shared consistency representation matrix; and constructing a hypergraph through the optimal sharing consistency representation matrix, and then clustering the multi-view data set by using a spectral clustering method based on the hypergraph to obtain a clustering result of the multi-view data set.

The invention relates to a high-dimensional data visual analysis method. The method comprises the steps that local subspace difference-geodesic distance projection is built on original high-dimensional data; mapping of clustering point clusters is built; visual analysis views of a series of subspaces are built. The invention further discloses an analysis system achieving the high-dimensional data visual analysis method. Accordingly, by building mapping of the local subspace difference-geodesic distance projection and clustering point clusters and the visual analysis views of the series of subspaces, a series of interactive visual analysis operations are put forward, a reliable technological base is provided for visual subspace clustering and analysis, a user can be effectively guided and assisted to conduct effective analysis and exploration on high-dimensional data, the frequencies of tests and errors of the user are significantly reduced in the high-dimensional data processing process, the data redundancy is reduced, the interactivity of the data analysis process is enhanced, and the reliability of the result is improved.

The invention discloses a subspace clustering visual analysis method based on dimension correlation. The subspace clustering visual analysis method comprises the steps of establishing a dimension correlation measurement method based on clustering significance; establishing an effective visualization method for a subspace clustering complex structure; and establishing a visual analysis framework based on the dimension correlation. In the interactive and visualized data exploration process, the invention give a user effective guidance information to guide the user to quickly find the valuable subspace and the corresponding cluster.

This invention relates to context-dependent Chinese phone identifying and modeling method, which applies initial consonant right-dependent and final sound left dependent modeling method including: a, creating a context-dependent basic modeling unit by relating the initial consonant with the adjacent right final sound and relating the final sound with its adjacent left initial consonant, b, utilizing the state clustering method to train the model parameters to get an initial HMM, c, utilizing the sub-space clustering method to compress the HMM to generate a final model.

The invention discloses a parameter-independent aircraft flight path clustering method based on contour coefficients. The parameter-independent aircraft flight path clustering method comprises the following steps: firstly, normalizing spatial position coordinates of all tracks in a track set; then, establishing a track similarity, a track distance matrix, a degree matrix and a Laplace matrix basedon the dynamic time bending distance and a Gaussian kernel function; next, clustering the track feature subspaces cluster by cluster in a given interval by utilizing a k-means algorithm; and finally,determining the optimal cluster, the optimal cluster number and the maximum average contour coefficient by taking the average contour coefficient of the track set as an evaluation standard of clustering quality. Compared with an existing method, the parameter-independent aircraft flight path clustering method has the advantages that expert experience or domain knowledge is not needed; human intervention is eliminated; the objectivity of the clustering process and result is high; and the parameter-independent aircraft flight path clustering method is not influenced by flight path length, speed, horizontal and vertical coordinates and height coordinate value domain difference, and is suitable for various data formats; the workload of user parameter adjustment is reduced, and the time cost is saved.

The invention discloses a multi-modal image retrieval method and system for an appearance patent. Firstly, feature extraction and fusion are carried out on multiple views of an appearance patent, thenfeature extraction is carried out on a text, information of multiple modes is comprehensively considered, and finally deep visual semantic embedding is carried out, so that a good retrieval effect can be achieved in a large-scale appearance design patent database; for a tree structure in an ANN, compact coding representation is not performed on data so that efficiency is not high. Calculation ofthe Hamming distance in the hash method is not an accurate distance calculation problem. According to the invention, distance coding product quantization is provided, in the coding process, data points are coded into series connection of subspace clustering indexes, the distance between each data point and a reconstructed coded representation of the data point is coded, and an effective compact coded representation of each datum is formed; and therefore, the retrieval efficiency and accuracy are improved.