134results about How to "Accurate clustering" patented technology

The invention provides a video resource clustering method and a device thereof. The method comprises: a video keywords thesaurus which comprises video keywords used for describing the video resource is established; a video resource database is set up, wherein, the video resource database preserves relevant information of the obtained video resource and extracts the keywords of a video resource title according to the video keywords thesaurus; when the relevant information of the video resource in the video resource database includes the video keywords by judgment, the video resource is clustered into a first-level category which is classified according to the video keywords.

The invention relates to an obstacle clustering method and an obstacle clustering device. The method comprises steps: three-dimensional point cloud data are acquired, and coordinates of a mapping point on a vehicle body coordinate system are determined; the mapping point is projected to a grid map; an obstacle point is recognized according to the mapping point coordinates and the grid map; the obstacle point is clustered to acquire K obstacle clustering clusters and K clustering centers; the similarly between the obstacle point and each clustering center is calculated, and the obstacle point is divided to an obstacle clustering cluster corresponding to a clustering center with the highest similarity; the clustering center is updated; whether each clustering center meets a convergence condition is judged; and when a clustering center not meeting the convergence condition exists, the step of calculating the similarly between the obstacle point and each clustering center and dividing the obstacle point to an obstacle clustering cluster corresponding to a clustering center with the highest similarity is returned until all clustering centers meet the convergence condition. Thus, obstacle clustering can be realized accurately and reliably, and the obstacle recognition rate can be improved.

Disclosed is a radar signal clustering method using frequency modulation characteristics and combination characteristics of signals including: a first step of assigning pulses of received radar signals to cells consisting of parameters including radio frequency (RF) and angle of arrival (AOA) of the pulses; a second step of calculating a pulse density distribution of each cell using a kernel density estimator; a third step of extracting a corresponding cell as a frequency fixed cluster if the calculated pulse density distribution is greater than a threshold of the frequency fixed cluster; a fourth step of making cell groups by merging remaining cells that are not extracted as the frequency fixed clusters; a fifth step of calculating a pulse density distribution of each cell group by using the kernel density estimator for each cell group; and a sixth step of comparing the calculated pulse density distribution for each cell group with each threshold according to a signal combination type of frequency agile clusters, thus to classify and extract each cell group according to the signal combination type.

The invention discloses a remote sensing image change detection method based on fusion and PCA kernel fuzzy clustering. The remote sensing image change detection method mainly solves the problems that in the prior art, the detection effect is not ideal, the accuracy of single-type difference image detection is low, and the application range is narrow. The method comprises the steps: (1) inputting two time phase remote sensing images X1 and X2 and conducting median filtering; (2) calculating a differential image, a logarithmic specific value image and a mean value ratio image of the two images after the filtering; (3) conducting fusion on the three images to obtain an image Xd after the fusion; (4) using a PCA method for conducting feature extraction on the images after the fusion, and obtaining a feature vector of each pixel to form a feature space matrix; (5) using a kernel-based fuzzy C mean value method for clustering the feature space matrix into two classes; (6) obtaining a final change detection result image according to the clustering result. The remote sensing change detection method has the better anti-noise performance and detection accuracy, the better effects of remote sensing images of different types can be obtained, and the remote sensing image change detection method can be applied to the field of environment monitoring and disaster evaluation.

The invention discloses an original article influence analysis system based on collection of media information. The system comprises a media article data acquisition module, an update module for article page views, comments and likes, an original article clustering analysis module and an original article influence calculation module. The media article data acquisition module is used for acquiringarticle information issued by a media platform in the internet, extracting content text from article information and storing the content text. The update module for article page views, comments and likes is used for obtaining dissemination feedback data of the article information and storing the dissemination feedback data of the article information. The original article clustering analysis moduleis used for clustering all content texts stored in a text database in order to obtain original articles. The original article influence calculation module is used for calculating influence of the original articles in the media platform and used for calculating influence of the original articles in all media platforms. The invention further discloses an original article influence analysis method based on collection of media information. By quantitative analysis of original article influence, analysis efficiency is high and accuracy of analysis is great.

The invention belongs to the technical field of computer-aided process planning, the invention discloses a three-dimensional process planning method and a platform for typical automobile machined parts based on MBD, based on 3D CAD software, with the MBD design model as the sole data input, the process MBD model is used as the data output, and the design process includes the establishment of MBD related standards, the creation of MBD design model, feature classification and feature database, feature recognition and information extraction, the generation of machining elements, the generation ofmachining elements clustering process, process sequencing, manufacturing features, the creation of process model and so on. The final application example of the present invention is in the case whereNX is used as a carrier, C + + and NXopen language, the invention can quickly generate a process MBD model which integrates a process model and a manufacturing feature body, can realize the visualization of the process design flow, improve the process design efficiency, and lays a foundation for the integration of CAD / CAPP / CAM. The invention can be used in the three-dimensional CAPP system developed by C + + and NXopen language.

The invention discloses a text clustering method. The method comprises the steps of receiving a text clustering instruction sent by a user; pre-training a pre-determined initial language model by utilizing the to-be-clustered corpus to obtain a target language model; sequentially inputting each text in the to-be-clustered corpus into the target language model for feature extraction, obtaining a sentence vector of each text in the to-be-clustered corpus according to a model output result, and generating a to-be-clustered sentence vector set; and, by utilizing a preset clustering algorithm, clustering the to-be-clustered corpora based on the to-be-clustered sentence vector set to obtain sentence vectors corresponding to each category, and determining a clustering result of the to-be-clustered corpora. The invention further discloses an electronic device and a computer storage medium. By utilizing the method and the device, the text clustering accuracy and efficiency can be improved.

The invention provides a text clustering method, a question-answering system applying the same and a search engine applying the same. The method comprises the following steps of 1) clustering texts in various languages; 2) drawing character word vectors of clustered texts in the various languages; and 3) calculating similarity of character word vectors of the texts in different languages, and clustering all the texts. By using the method, the question-answering system and the search engine, the texts in the various languages can be clustered correctly.

User activity streams are used to automatically learn associations between logical objects and form logical groups. Search results are sorted based upon their relevance to current user activity. Combined with a graphical user interface component and object database, the invention automatically retrieves and display groups and objects related to the active object.

The invention provides short text clustering equipment which comprises a subject analysis unit, a vector generating unit and a clustering unit, wherein the subject analysis unit is used for conducting subject analysis on each text in an auxiliary text collection and a short text collection, thereby obtaining the possibilities that each short text in the short text collection is corresponding to a subject of the auxiliary text collection and the subject of the short text collection; the vector generating unit is used for conducting normalization on the possibilities that each short text is corresponding to the subject of the auxiliary text collection and the subject of the short text collection so as to generate a vector; and the clustering unit is used for clustering the short texts in the short text collection based on the generated vector. Meanwhile, the invention further provides a short text clustering method. According to the short text clustering equipment and the short text clustering method, the independent finding of the auxiliary text subject and the short text subject can be realized, thereby clustering the short texts more accurately.

The invention discloses an iteration text clustering method based on self-adaptation subspace study. The method includes the following steps: (1) initiation: text linguistic data is expressed as a text vector space, initial K clusters are generated through an affine propagation clustering method, and all text clustering categories are expressed as an initial category affiliation indication matrix; and (2) iteration between the subspace projection and the clusters: the initial category affiliation indication matrix is used as prior knowledge, a maximum average neighborhood edge is used as a target to solve a subspace projection matrix, the text vector space is projected to a subspace, K clusters are generated through the affine propagation clustering method in the subspace, and a category affiliation indication matrix is updated; and a convergent function is calculated based on the subspace projection matrix and the category affiliation indication matrix till the function is converged, iteration exits, and text clustering is finished. The iteration text clustering method does not limit the capacity and distribution of text data, subspace solution and clusters are fused under a uniform frame, and an overall optimal clustering result is obtained through an iteration strategy.

The invention relates to a deep migration embedded clustering machine learning method based on local structure preservation, The method is based on the local structure of the data generation distribution saved by the incomplete automatic encoder, fuses the clustering loss and the reconstruction loss to establish a clustering optimization model, and solves the clustering optimization model througha small batch of stochastic gradient descent and back propagation algorithm to achieve clustering. Compared with the prior art, the present invention solves the problem that the prior DEC method cannot save the local structure of data, and has the advantages of simple method and high clustering accuracy.

The invention discloses a text abstract generation method based on a K-means model and a neural network model, and the method comprises the steps of preprocessing an original text, obtaining single sentences and words through segmentation, inputting the sentences and words into a doc2vec model, and carrying out the training, so as to obtain sentence vectors; determining the number of clustering centers of the original text, inputting the sentence vector into an unsupervised K-means model, and training to obtain a clustering center vector; calculating the Euclidean distance between the clustering center vector and the sentence vector, and extracting a sentence corresponding to the sentence vector closest to the clustering center to serve as a reference abstract; and inputting the original text, the reference abstract and the words into a generative neural network model to generate a text abstract. The method has the beneficial effects that the unsupervised model and the supervised neural network model are combined, so that the generated text abstract can be semantically coherent and is convenient for a user to understand.

The invention provides a GM-HMM (Gaussian Mixture-Hidden Markov Model) driving behavior prediction method based on visual characteristics. A six-degree-of-freedom driving simulator and an eye tracker system are selected to carry out a simulation experiment, a fuzzy K-means dynamic clustering algorithm is adopted to divide the distribution of a driver interest visual field area, and the Pauta criterion of an improved Bessel formula is adopted again to remove abnormal line of sight points without a definite boundary among divided driver interest visual field areas; then, from an entry point of the visual representation parameter of a driver, a container type graph analysis method in mathematical statistics is adopted to verify the difference of the parameter through a NEMENYI rank sum test in SPSS (Statistic Package for Social Science) software, and the visual representation parameter sequence of the driver is determined through R type index clustering; and finally, a GM-HMM driving behavior prediction model is established, and the reliability of the GM-HMM driving behavior prediction method is analyzed.

A method and system for parallel optimal configuration of distributed photovoltaic power supply of a distribution network are disclosed. The method comprise that following steps: acquiring real-time operation state parameter of distributed photovoltaic power source; Substituting the operating state parameters into a pre-established distributed photovoltaic power source optimal configuration model;Multi-scene analysis and multi-objective molecular differential evolution algorithm based on parallel computing are used to solve the optimal placement model. According to the solution results, the optimal configuration of distributed photovoltaic power is carried out. The system includes background server, photovoltaic grid-connected inverter, interruptible load controller, on-load tap changer regulator, data acquisition and monitoring device. The invention takes the active management measures as the decision variables, and introduces the active management measures into the optimal configuration of the distributed power generation to obtain the optimal installation position and capacity of the distributed power generation and the optimal implementation scheme of the active management measures, which can effectively increase the consumption of the distributed power generation and improve the voltage quality.

The invention relates to a large-scale multi-view data self-dimension-reduction K-means algorithm and a system, and belongs to the technical field of information processing, and the method comprises the steps: 1, carrying out the normalization of data with different features, and enabling all data to be in a range of [-1, 1]; 2, initializing; 3, optimizing the algorithm; and 4, using a data set tooptimize the algorithm according to the algorithm until the algorithm is finally converged to obtain a final clustering result, measuring the clustering effect by using the interaction information entropy and the purity, and repeating the step 3 by selecting different initial values, and removing the average value of the result to complete the experiment. The relationship between the features andthe clustering targets is fully considered; information complementation among different views is utilized, self-dimension reduction of high-dimensional data is achieved by searching for an optimal subspace on a single view, a loss function is reconstructed through non-negative matrix factorization (NMF), the different views share the same clustering indication matrix, therefore, multi-view information complementation is achieved, and clustering of large-scale multi-view data is completed.

The invention relates to the technical field of computers, in particular to a data classification method, a data classification device and electronic equipment. The method comprises the steps of obtaining at least two attribute values of a target attribute, and selecting one attribute value from the at least two attribute values as an initial clustering center; wherein the to-be-processed data comprises a plurality of samples; according to the distance between each attribute value and the initial clustering center, calculating a probability value that each attribute value can be used as the clustering center so as to determine the clustering center according to the probability value; clustering the attribute values of the target attributes based on the clustering centers, and dividing interval boundaries according to clustering results; and classifying the attribute values corresponding to the target attributes of the samples in the to-be-processed data according to an interval division result. According to the method, discrete processing can be carried out on continuous values, normal values and abnormal values are stored, and the normal values and the abnormal values can be classified into different categories.

The invention provides a document clustering method and device. The method includes the steps of A, vectorizing each document to allow each vectorized document to correspond to a document coordinate in a multi-dimensional space; B, clustering the documents into two clusters and acquiring geometric center of each cluster in the multi-dimensional space; C, calculating average radius of each cluster, clustering documents corresponding to the document coordinates in the two clusters into a inseparable category if the average radius satisfies a preset condition, and corresponding the two clusters into two separable categories if the average radius does not satisfy the preset condition; D, executing step B and C in each separable category; E, terminating clustering when each document belongs to the inseparable category; wherein the average radius the average value of the distance from all document coordinates to the geometric centers. By the method, document clustering accuracy and intelligence are increased.

The invention relates to a text label construction method and device, computer equipment and a storage medium. The method comprises the steps of obtaining to-be-processed text data, performing word segmentation processing on the to-be-processed text data to obtain a word segmentation set, training the word segmentation set through word2vec to obtain similarity among words in the word segmentationset, performing word clustering based on the similarity among the words, and constructing a text label according to a word clustering result. Whole process, the similarity between words in the text data is accurately obtained through word2vec training, clustering is carried out based on the similarity between the words, accurate clustering can be achieved in an iterative clustering mode in the clustering process, and labels of the text data can be reasonably and accurately constructed based on the clustering result of accurate clustering.

The invention relates to a method and a device for color editing of a natural image with repetitive scene elements. The method comprises: preprocessing a to-be-edited natural image by a super pixel segmentation method, so as to divide the natural image into a predetermined number of image sub-block areas; extracting color features and texture features of the natural image, and constructing a feature space vector according to the color features and the texture features; clustering the predetermined number of image sub-block areas according to the feature space vector, so as to locally assign colors to the natural image on the basis of a clustering result; by a global optimization algorithm, carrying out color transfer of the natural image which is already subjected to local color assignment, in order to obtain an image which is a result of image editing. The method and the device of the invention can improve the speed of color editing, effectively keep the consistency of colors of repetitive scene elements, realize editing of different colors of the image sub-block areas with different marks, and obtain rich color editing results.

The invention relates to a user data processing method and device, computer equipment and a storage medium. The method comprises the steps of obtaining multiple pieces of user data, wherein each pieceof user data comprises feature data of multiple dimensions; screening a plurality of target user data from the user data; according to the feature data of the target user data in each dimension and the weight corresponding to each dimension, determining the similarity degree between the target user data; clustering the target user data based on the similarity between the target user data to obtain the category to which each target user data belongs; training a classification model according to the target user data and the category to which the target user data belongs; classifying the residual user data after screening through the classification model obtained by training to obtain the category to which each piece of residual user data after screening belongs; and pushing information according to the category to which each piece of user data belongs. By adopting the method, the accuracy of user data classification can be improved.

The invention discloses an unmanned mine truck obstacle detection method, which comprises the following steps: converting obstacle data obtained by a laser radar and a millimeter wave radar into corresponding vehicle body coordinate systems respectively; drawing a 0, 1 binary image of a ground-high elevation point by adopting a grating map height difference in combination with field difference ground detection; clustering the high elevation points by adopting a multi-parameter model; judging whether the clustering result is an obstacle influencing normal running of the vehicle according to themotion trail of the vehicle; detecting whether the vehicle is in a drivable area; and matching the obstacle data acquired by the millimeter wave radar with the obstacle data acquired by the laser radar, and outputting a final result. According to the invention, based on the actual application environment of the mining dump truck, obstacles in a road are effectively detected, missing detection isprevented, and clustering is accurate; and the method is good in robustness, and reduces the false detection rate through the matching of the detection result of the laser radar and the detection result of the millimeter-wave radar through employing a scheme of the fusion of a plurality of radars.

The invention relates to an image clustering method based on self-representation and graph constraint non-negative matrix factorization, and is particularly suitable for clustering of complex categories in a data set. According to the method, dimensionality reduction is carried out on high-dimensional data based on an image clustering method of non-negative matrix factorization of self-representation and graph constraint, and particularly for the situation that an abnormal value exists in an image, the abnormal value being recorded as LRE-GNMF. A target function is solved by using an alternating iteration method to obtain a low-dimensional representation coefficient matrix; and the images are clustered by using the low-dimensional representation coefficient matrix. According to the method,a non-negative data matrix is used as input, and low-rank embedding (LRE) is adopted, so that data close to each other in a high-dimensional space can still be kept close to each other in a learned low-dimensional space, and therefore, the local structure of the data can be kept. The method can be widely applied to the field of image recognition.

The invention provides a UUV (unmanned underwater vehicle) coast patrol contour construction method based on wavelet cluster, and discloses an improved contour construction method based on wavelet cluster to solve environmental contour construction problems in unknown environment detection and synchronous coast patrol control of a UUV. Original barrier point data are subjected to local wavelet cluster, an isolated point alternative set R and a set Ci of all categories of the UUV in a vision field at a current position are given, isolated points and category data are processed by a series of cluster strategies made by global cluster decision, the isolated points are removed to obtain a barrier point set of a coast contour, and a specific contour is obtained by the barrier point set and an Alpha-Shapes method. Barrier points are acquired and noise data are added in the UUV coast patrol simulation process to serve as original data for simulation, the coast contour is rapidly and accurately given by an algorithm, and the effectiveness and feasibility of the algorithm are proved.

The invention provides an adaptive fuzzy C-means image segmentation method based on a potential function and mainly aims at solving the problems that classifying number needs to be previously set and the segmentation efficiency is low in the existing fuzzy C-means segmentation method. The method is realized by the following steps: (1) inputting a to-be-segmented image; (2) obtaining the histogram potential function and the maximum potential remnant height of the to-be-segmented image; (3) obtaining the histogram c-factorial remnant potential function of the to-be-segmented image; (4) combining pseudo potentials; (5) obtaining the clustering center and the classifying number of the to-be-segmented image; (6) carrying out fuzzy classification on the pixel dots of the to-be-segmented image; and (7) outputting an image segmentation result. The adaptive fuzzy C-means image segmentation method has the advantages of capability of obtaining optimal image classifying number in an adaptive way, high segmentation efficiency, strong region consistency in the segmentation result, smooth margin and the like. The adaptive fuzzy C-means image segmentation method can effectively segment natural gray images and infrared images and can be used for target recognition and tracking.

The invention relates to a video abstract generation method and device, electronic equipment and a storage medium, and belongs to the technical field of the Internet. The method comprises the steps ofobtaining multiple frames of images in a target video, inputting the multiple frames of images into an image clustering model, dividing the multiple frames of images into multiple categories based onthe image clustering model, selecting the target image from each category, and splicing the selected multiple target images to obtain a video abstract of the target video. The method is based on theimage clustering model, can quickly cluster multiple frames of images, does not need to analyze the motion trail of the object in each frame of image, shortens the processing time, and improves the efficiency of generating the video abstract. Moreover, the target image is selected from each category, the target image can represent the content of the multiple frames of images in the category, the content of the target video can be accurately summarized according to the video abstract generated by the multiple target images, and the accuracy of the generated video abstract is improved.

The invention provides a topic clustering method and device, electronic equipment and a storage medium. According to the method, regression analysis can be carried out on the text data set based on aBERT model to obtain a base dataset, paragraph information of each text can be better expressed, the accuracy of text representation is improved, a configuration quantity of data are selected from thebasic data set for labeling, a small amount of annotation information is used for assisting overall unsupervised clustering, by adopting an Agglomerative Clustering model, clustering is carried out by combining a first inter-class distance and a second inter-class distance, and a similarity model trained based on a BERT algorithm is further adopted to obtain the target clustering result, so thatthe clustering result under the large inter-class distance is adopted as guidance, the clustering results under the small inter-class distance are combined, meanwhile, the recall rate and accuracy areensured, the dependence on the clustering distance and category is reduced, and the clustering effect is improved. The invention further relates to a block chain technology. The BERT model, the Agglomerative Clustering model and the similarity model can be stored on the block chain.

The invention discloses a blended data clustering method based on density searching and rapid partitioning. The blended data clustering method is characterized by comprising the following steps of determining a domination type of blended data in a blended attribute dataset; calculating the distance between any two blended data in the blended dataset according to the domination type of the blended data; optimizing the clustering radius within the preset clustering radius value range on the basis of a density searching algorithm according to the distance between the any two blended data, and using a corresponding clustering result corresponding to the optimal clustering radius as the final clustering result. According to the method, the domination analyzing method is executed on the blended data to determine the special type of the blended data, different distance calculation methods are adopted for different blended data, the importance of data dimension information with the domain attribute in overall data information can be effectively brought into play, and the data distance can be accurately calculated; the data clustering algorithm based on density searching and rapid partitioning is adopted, speed is high, and accuracy is high.

The invention discloses an information entropy additive fuzzy defect feature analysis reconstruction method based on infrared thermal imaging, and the method comprises the steps: carrying out the extraction and feature analysis of defects of different spatial indegrees through a reconstruction model and an optimized fuzzy algorithm, and enabling the defect features of different spatial indegrees to be accurately divided; meanwhile, in the optimized fuzzy algorithm, constructing novel target functions. One part comprises the sum of the sponsousness and the hesitance; feature information of elements is enriched; the other part comprises fuzzy entropy; the uncertainty of the feature information is described; the method provides effective help for the distinguishing of defects, such design andconstruction have good stability and high efficiency, have outstanding effects on the description of defect characteristic textures and the representation of characteristic chromatic aberration, andcan reasonably and accurately evaluate and analyze different degrees of defects in different spaces.