95 results about "Cluster labeling" patented technology

The embodiment of the invention discloses a tag clustering method and a tag clustering system, wherein the method comprises the steps of; establishing characteristic vectors of every tag to be clustered; calculating a cosine included angle of two characteristic vectors in Euclidean space to obtain the similarity between every two tags to be clustered; and clustering the tags to be clustered by using K-Means algorithm according to the similarity between the tags to be clustered. The tag clustering system comprises: a characteristic vector establishing module which is used for establishing the characteristic vectors of every tag to be clustered, a similarity calculating module which is used for calculating the cosine included angle of two characteristic vectors in Euclidean space to obtain the similarity between every two tags to be clustered, and a clustering module which is used for clustering the tags to be clustered by using the K-Means algorithm according to the similarity between the tags to be clustered. The technical scheme can overcome the defect of inaccurate similarity calculation of tags in the current collaborative tag system, settle the problems of disordered tag organization and fuzzy tag semantics, and enhance the accuracy of tag clustering effectively.

Document clustering methods, document cluster label disambiguation methods, document clustering apparatuses, and articles of manufacture are described. In one aspect, a document clustering method includes providing a document set comprising a plurality of documents, providing a cluster comprising a subset of the documents of the document set, using a plurality of terms of the documents, providing a cluster label indicative of subject matter content of the documents of the cluster, wherein the cluster label comprises a plurality of word senses, and selecting one of the word senses of the cluster label.

While deep learning is used for classification, multi-modal information with rich samples and classification contribution variability of each modality are considered, and the problem of insufficient samples is solved by using a semi supervision method. Data of different modalities of a hyperspectral image is sent into a deep neural network, the semi supervision method is used and a large number ofunlabeled samples are utilized, and the deep neural network based on self-encoding is used for feature learning. All labeled and unlabeled data are sent into the self-encoding deep neural network tocarry out learning, similar networks are designed for different modalities, a respective initialization parameter is obtained through self-encoding reconstruction, and hidden attributive classification of labeled samples is obtained through a clustering method. For the unlabeled data, a deep characteristic is calculated through a multi-target deep network, then a similar marked sample is searchedbased on a clustering label, and finally, labels of the unlabeled samples are predicted according to the label information of the labeled samples.

The invention discloses a Chinese Web document online clustering method based on common substrings. As known to all, search engines are important in application of information searching and positioning with sharp increase of information on the internet. Web document clustering can automatically classify return results of the search engines according to different themes so as to assist users to reduce query range and fast position needed information. The Web document online clustering is characterized in that non-numerical and non-structured characteristics of Web documents are required to be met on the one hand, and clustering time is required to meet online search requirements of users on the other hand. According to the two characteristics, the invention provides the Chinese Web document online clustering method based on common substrings, and the method comprises steps as follows: (1) firstly, preprocessing the first n query results returned by the search engines so as to realize deleting and replacing operation of non-Chinese characters in the return results of the search engines, (2) extracting common substrings in the Web documents by utilizing GSA, (3) presenting a weighting calculation formula referring to TF*IDF according to the common substrings which are extracted and then building a document characteristic vector model, (4) computing pairwise similarity of the Web documents on the basis of the model to acquire a similarity matrix, (5) adopting an improved hierarchical clustering algorithm to achieve clustering of the Web documents on the basis of the matrix, and (6) executing clustering description and label extraction. The Chinese Web document online clustering method based on common substrings has obvious advantages on performance, clustering label generation and clustering time effects.

The invention discloses a retrieval result clustering method facing to a Web2.0 label picture shared space, which comprises the following steps: excavating a vocabulary relationship and an associated relationship between labels; obtaining an expanded querying label set by a query label according to the vocabulary relationship between the labels; obtaining a candidate image set relevant to query by the expanded query label set; selecting front K most relevant labels according to the relevance measurement of the labels in the query label set and the candidate image set; automatically dividing the K labels into an optimal clustering result according to the association between the K labels by a clustering algorithm based on a picture division from top to bottom; and correspondingly clustering the candidate image set according to clustering labels. Aiming at the problem of inconformity of label expression, the effective query expansion is realized, and the image clustering method based on most relevant label set clustering solves the problem of diversity of label semanteme. Compared with a traditional method, the invention leads a user to rapidly and effectively retrieve and browse a picture in the Web2.0 label picture shared space.

The invention discloses a video portrait file processing method and a system thereof, which are applied in the technical field of video surveillance and solve the technical problem of inaccurate portrait clustering analysis. The technical scheme essentials are a video portrait file processing method, comprising the following steps: collecting image information to obtain image data set; Acquiring the existing archives, identifying the portrait information in the image data set, comparing the portrait information with the portrait in the existing archives, and obtaining the contrast result as portrait matching degree; Comparing portrait matching degree and threshold, when portrait matching degree is greater than or equal to threshold, portrait information is classified into existing portraitarchives; when portrait matching degree is less than threshold, portrait archives are created, and portrait information is classified into new portrait archives. The technical effect is to establisha clear portrait archives and portrait relationship topology, so that portrait archives group according to the characteristics of clustering labeling, easy to analyze.

The invention discloses a multi-agent genetic clustering algorithm-based image segmentation method, which mainly solves the problems that the prior art is sensitive to an initial clustering center, has low convergence rate and is easily trapped in a local extremum. In the method, image clustering segmentation is converted into a global optimization problem. The method comprises the following steps of: firstly, extracting two-dimensional gray scale information of a neighborhood median and a neighborhood mean of pixel points of an image to be segmented to construct a new two-dimensional histogram; secondly, combining a multi-agent genetic algorithm (MAGA) with a fuzzy C-mean (FCM) clustering algorithm and obtaining an optimal clustering center and a membership degree matrix by using the global optimization capability of the MAGA; and finally, outputting clustering tags according to the maximum membership degree principle so as to realize image segmentation. The method has high anti-noise capability and high convergence rate, can improve the image segmentation quality and the stability of a segmentation result and can be used for extracting and identifying image targets.

The invention relates to a query recommendation method and apparatus. The method comprises the steps of obtaining a historical query statement input by a preset user; analyzing the historical query statement to obtain statement information of the historical query statement; according to the statement information, determining data table item information queried by the historical query statement, and according to the data table item information, determining a cluster tag corresponding to the preset user; according to cluster tags corresponding to a plurality of historical query statements, determining a recommendation query object corresponding to the preset user; and generating a recommendation query statement corresponding to the recommendation query object. During actual application, before the user inputs the query statement next time, the query statement can be displayed through a popup window or a pull-down menu, and correspondingly, the user can finish statement input only by clicking, so that each character does not need to be tapped on a keyboard, the input efficiency of the query statement is improved, and correspondingly the data query efficiency is enhanced.

The invention discloses a fuzzy C-means clustering method of a minimum variance optimization initial cluster center, and belongs to the technical field of data mining and pattern recognition. The method comprises the steps of clustering a distance relation between an input data set and sample points; using a clustering analysis method for clustering analysis of a target data set to obtain a clustering label; and evaluating the performance of the clustering label obtained after the clustering analysis and an original label according to an evaluation index. The invention aims to solve the problem that a clustering effect of the fuzzy C-means is greatly affected by an initialized clustering center thereof and an optimal solution cannot be guaranteed. Based on the FCM algorithm, the selection of the initial clustering centers is first performed by selecting K sample points with the least variance on different regions as the initial cluster centers by taking the sample variance as heuristic information and by the sample field radius, and the algorithm does not need the setting of any parameters.

The invention discloses a searching result clustering method and apparatus for solving the problem that the searching result conforming to own requirement according to clustering labels is difficult to find by the searching result clustering method provided by the prior art, wherein, the method provided by the invention comprises: selecting the regulated numbers of phrases in the searching results; clustering the searching results containing the phrase aiming at the selected each phrase, and taking the phrase as the clustering label.

The embodiment of the invention provides a method and a device for model training, wherein the method comprises the following steps: obtaining sample data to be trained in a designated classificationcategory; carrying out feature extraction on the sample data to be trained to obtain feature information corresponding to the designated classification category; clustering feature information corresponding to the designated classification category to obtain a plurality of clustering labels; performing data equalization processing on sample data corresponding to the clustering label; taking the sample data after the data equalization processing as the target sample data; Using the target sample data, the designated model is trained. The invention can refine the labels in the existing classification categories by the above-mentioned unsupervised method, Realize the sample balance within the category, provide balanced sample data for the model, according to the balanced sample data model training can be optimized model, using the optimized model for data forecasting can get more accurate forecasting results, improve the accuracy of model forecasting.

The invention discloses a learning robustness multi-visual angle clustering method based on nonnegative dictionaries, comprising: extracting characteristics of a data set including a plurality of sub-spaces under different visual angles; embedding characteristic learning in dictionary learning, and jointly learning a meaning projection matrix and nonnegative characteristic projection; adding consistency constraints and local geometrical constraints to learn common cluster labels shared by multiple visual angles, learning each visual angle meaning projection matrix, parameter expression matrix and multi-visual angle shared meaning projection matrix under a plurality of constrains, and completing multi-visual angle clustering. The method can explore common meaning labels shared by multiple visual angles, add consistency constrains, reduce difference between a single clustering label and a common meaning label, and meanwhile add the local geometrical constraints, allowing data with a similar structure to be divided in a same cluster with a greater probability.

The present invention provides an image identification method, device and system. The method comprises: obtaining an image to be identified; and obtaining an image identification model, wherein the image identification model is generated by performing machine learning of a training sample. The training sample comprises: training images, cluster tag of the training images and dichotomy tag of the training images, wherein clustering tags of the training images are obtained after clusterings of the training images. The image identification model is employed to perform identification of the imageto be identified and determine the dichotomy scoring of the image to be identified. The image identification method can improve an identification effect of a dichotomy image.

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 figure labeling method. According to the method, figure characteristics of labeled figures in labeled figure pictures are acquired through skin color filtering in combination with face detection; according to the acquired figure characteristics of the labeled figures, corresponding figures are identified by utilizing figure characteristic similarities to carry out cluster labeling of the corresponding figures in unlabeled figure pictures. The invention further provides a terminal. Through the figure labeling method, labeling accuracy and efficiency are high.

The invention discloses a similarity based semantic Web service clustering labeling method. The method is characterized by comprising two parts of realizing semantic Web service similarity calculation and realizing a semantic Web service clustering labeling algorithm. During the semantic Web service similarity calculation, in combination with results of input / output (I / O) parameter mixed similarity calculation and service description keyword similarity calculation, a calculation result of semantic Web service similarity is comprehensively obtained, and the difference and similarity between service functions are reflected; and I / O parameters can directly describe functions of corresponding service modules and serve as measurement standards for calculating the semantic Web service similarity from a functional perspective. According to the method, the accuracy of similarity calculation can be improved and the performance of a service discovery system is further improved.

The invention discloses an intelligent adaptive equalizer and an equalization demodulation method based on machine learning. The method includes the following steps: the data of an acquired signal ispreprocessed, and the energy of input data is normalized. Then, the cluster groups in the data are clustered by using a Gaussian kernel function and a distance function under the condition of withoutany other prior knowledge, and the cluster groups in the clustered data are labeled by using the nearest neighbor algorithm in order to realize useful informationization of modulation signals. The discrete noise points outside the cluster groups are not clustered, and the cluster halos without clustering labels are marked by using the weighted K-nearest neighbor algorithm. Finally, the data is allintegrated to obtain an overall label, and the error rate of the system is estimated by comparing the overall label with pre-stored labels. By using the method of the invention, the real clustering center can be identified without any other prior knowledge, regardless of the shape and size. The computational complexity is reduced, and the accuracy of the classification result is significantly improved. The method can adapt to most of the modulation formats in the current communication system.

The invention discloses an image segmentation method based on an iteration self-organization and multi-agent inheritance clustering algorithm. The method mainly solves the problems that a segmentation result depends on initial parameters excessively, and the phenomenon of local optimum occurs easily in the prior art. The method comprises the segmentation steps that 1) gray information of an image to be segmented is extracted; 2) the algorithm thought of the iteration self-organization algorithm ISODATA is used for the image to be segmented to obtain the optimal clustering number; 3) according to the optimal clustering number, a multi-agent algorithm frame is utilized for clustering the image to be segmented to obtain an optimal clustering label; 4) according to the optimal clustering label, image pixels of the image to be segmented are classified to achieve image segmentation. According to the method, the clustering number does not need to be determined definitely, the convergence effect is good, the global optimum value can be obtained easily, the quality of image segmentation can be improved, the stability of the segmentation result is enhanced, and the method can be used for extraction and identification of image targets and other follow-up processing.

The invention discloses an image division method based on an inter-class maximized PCM (Pulse Code Modulation) clustering algorithm. The method comprises the following steps: carrying out classified labeling on pixel points of an input image according to a gray value; obtaining a clustering label when a clustering analysis method is used for dividing a target image; and carrying out performance evaluation on a label obtained by image division and an original label according to an evaluation index by a clustering method. The novel inter-class maximized PCM clustering algorithm considers the inter-class penalty, and parameters are adjusted and adjusted to enlarge the distance between class centers, so that the optimal classification of the pixel points in the image is realized.

The invention relates to a load prediction method and device, computer equipment and a storage medium. The load prediction method comprises the following steps that: computer equipment obtains load data of different transformer areas, performs clustering analysis on the load data to obtain a load data sequence corresponding to each clustering label, and inputs the load data sequence correspondingto each clustering label into a preset load prediction model to obtain a load prediction result. In the load prediction method, as the fluctuation range of the load data of different transformer areasis large, and the load data volume is large and the type is complex, the computer equipment acquires load data of different transformer areas and then carries out clustering analysis on the load dataand places the same type of data in the same load data sequence, so that the fluctuation range of the load data of the same sequence is reduced, and the prediction result of each load data sequence is more accurate when the load prediction is carried out on each load data sequence, and the accuracy of the total load prediction result is further improved.

Disclosures of the present invention describe a moving object detection system and method, wherein a pre-processer module, a feature extraction module, an image optical flow estimation module, a feature points grouping module, and a moving object determination module are provided in a controlling and processing module of the system by a form of library, variables, or operands. Moreover, a feature difference calculation unit, a matrix establishing unit and a corner feature point acquiring unit are provided in the feature extraction module, and that is helpful for enhancing computing speed of the controlling and processing device in verifying corner feature points from image frames. Therefore, after the corner feature points are applied with a cluster labeling process, the moving object determination module can achieve motion detection of at least one object locating in a monitoring area by determining whether corner feature point groups move or not.

The invention discloses an image segmentation method based on superpixels and immune sparse spectral clustering. The method mainly aims to solve the problems that existing image segmentation methods are low in segmentation precision and poor in robustness. The method comprises the steps that first, superpixel division is performed on a texture image, and texture features of the texture image are extracted to serve as a feature dataset; second, an optimal similarity matrix of the feature dataset is found in combination with an immune cloning algorithm and sparse representation in the spectral clustering process; and last, the original image is marked according to clustering tags in combination with the superpixels, and segmentation of the texture image is realized. According to the method,superpixel blocks of the image are extracted to serve as the feature dataset, the image segmentation method based on immune sparse spectral clustering is used to divide the feature dataset, and a moreaccurate segmentation result is obtained.

The embodiment of the invention provides an object clustering method and device, a computer readable medium and electronic equipment. The object clustering method comprises the following steps: obtaining an interest label of each object, generating a label sequence corresponding to each object based on the interest label, clustering the objects according to the label sequences corresponding to theobjects; and obtaining an object group corresponding to the same clustering label, and under the condition that the object number of the object group is smaller than an object number threshold value,combining the objects in the object group into the object group associated with the clustering label to obtain a target group of which the object number is greater than or equal to the object numberthreshold value. According to the technical scheme provided by the embodiment of the invention, the target group has the clustering label capable of accurately representing the object preference information; each target group has a balanced scale, so that the clustering groups are correspondingly processed according to the clustering labels of the clustering groups, the accuracy and balance of object clustering are improved, and the processing accuracy of the clustering groups is further improved.

The invention relates to a semi-supervised tourist portrait data clustering method based on density peaks and gravitation influences, which comprises the following steps: calculating density values and distance values of all points of tourist portrait data through a density peak algorithm, and finding out all possible clustering center points; Calculating the distance between the tourist portraitseed points and possible clustering center points by using the provided tourist portrait seed points, voting and screening out accurate clustering center points, and pasting clustering labels on the corresponding clustering center points by using the seed label information; Randomly selecting a seed data subset with a certain proportion from all the seed data, and calculating the gravitation influence between the seed data subset and each unlabeled data point by introducing the idea of the universal gravitation law, thereby clustering all the unlabeled data and attaching corresponding clusterlabels to the unlabeled data; And randomly selecting a seed data subset for multiple times to attach a corresponding decision-making cluster label to the unlabeled data, and voting to select cluster label information of each piece of unlabeled data. The method is good in clustering effect and high in accuracy.

The invention discloses an oversampling method based on angle and direction clustering, and the method comprises the steps: obtaining an unbalanced data set, carrying out the clustering of the unbalanced data set through employing a clustering algorithm, generating a clustering label, an angle variance, and a sorting neighbor set for each sample in the unbalanced data set, and carrying out filtering processing on each sample of which the clustering label is noise so as to obtain a filtered sample; and calculating a first oversampling weight, a second oversampling weight and an optimal interpolation neighbor set of each minority class sample in the unbalanced data set according to the clustering label, the angle variance and the sorting neighbor set of each sample in the unbalanced data set, and calculating the oversampling weight of each cluster in the unbalanced data set and the number of new samples needing to be synthesized by each cluster according to the first oversampling weights of all minority class samples. The technical problem that the importance of boundary samples in classification is ignored in an existing oversampling method can be solved.

The invention provides a training data set generation method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a classified source data set and an unclassified target data set; extracting a first feature vector set of the source data set and a second feature vector set of the target data set through a feature extractor; determining a class center feature vector corresponding to the source data set according to the first feature vector set, and determining a clustering label of the target data set and an average feature vector in a clustering cluster according to the second feature vector set; iteratively optimizing the feature extractor, so that the overall difference between the feature vectors of samples in the source data set and the feature vectors of a class center and the overall difference between the feature vectors of the elements in a clustering cluster and average feature vectors in the clustering cluster are made to be minimum; and obtaining a training data set according to the clustering label of the target data set and the elements in the clustering cluster. According to the method, the workload of manual labeling can be reduced, the manual labeling cost is reduced, and the labeling precision is improved.

The invention discloses a video processing method and system, a picture processing method and system, equipment and a medium, and the method comprises the steps: obtaining a video frame of a to-be-processed video, and obtaining to-be-clustered pictures; determining a to-be-clustered picture from the to-be-clustered pictures as a first reference picture, taking other to-be-clustered pictures as first non-reference pictures, generating a clustering label as a current clustering label, and marking the current clustering label for the first reference picture; calculating a first similarity betweeneach first non-reference picture and the first reference picture to obtain a first similarity value of each first non-reference picture, obtaining the first non-reference picture of which the first similarity value is greater than a first similarity threshold, obtaining a to-be-labeled picture, and labeling the current clustering label for the to-be-labeled picture; and obtaining the first reference picture and the to-be-annotated picture to obtain an annotated clustering label picture. By adopting the scheme, a user can quickly obtain the content information required by the user.