45results about How to "Clustering fast" patented technology

The invention discloses a system and a method for parallel video copy detection. The method includes the steps of 1, selecting key-frames of a query video and a reference video by a parallel method and extracting MIFT features of the key-frames; 2, clustering extracted feature data of the reference video by a parallel hierarchical clustering method; 3, quantizing the features of the query video and the reference video by a quantitative method according to clustered results; 4, establishing indexes of quantized data of the reference video; and 5, retrieving by the parallel method, preliminarily searching in the indexes to obtain an alternative video by utilizing quantized data of the query video and then computing space consistency and time consistency to finally confirm a copy video. According to the system for parallel video copy detection, a parallel mechanism is adopted on the basis of fast retrieval of visual vocabulary bag model BOF (beginning of file), and accordingly detection efficiency of the system is improved greatly.

The invention relates to a method for optimizing the proportion of high-proportion renewable energy under the cooperative access of transmission and distribution network, which comprises the followingsteps: 1) acquiring basic data; 2) clustering based on that basic data to obtain a typical plan scene; 3) for that typical plan scenario, establishing a collaborative planning model of transmission and distribution network, The objective of the collaborative planning model is to minimize the cost. The constraints include the independent constraints of the transmission and distribution network andthe collaborative constraints of the transmission and distribution network. The decision variables include the investment capacity of renewable energy units in the transmission and distribution network, the power generated by conventional units, the discarded energy power of renewable energy sources and the load shedding power of renewable energy units. 4) solving that transmission and distribution network collaborative planning model to obtain the optimal renewable energy investment capacity scheme; 5) obtaining the optimal distribution ratio of the high-proportion renewable energy in the transmission and distribution network based on the optimal renewable energy investment capacity scheme. Compared with the prior art, the invention has the advantage of high accuracy, strong practicability and high efficiency.

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 discloses an optimized arrangement method for monitoring points of an urban water supply pipe network. The invention improves the traditional sensitivity analysis and cluster analysis methods separately and combines the two methods to obtain optimized distribution of monitoring points, so that the arrangement of the monitoring points can have the characteristics of being good in sensitivity, being relatively dispersed and facilitating the delimiting of abnormal regions for accurate detection, and meanwhile, the monitoring points are more widely and uniformly distributed as compared with those in the prior art, and are reasonably arranged at key positions, thereby facilitating model verification.

The invention discloses an active learning-based image detection method for a liver injury type. The method comprises the specific steps of 1, locating and extracting a liver target region from an abdominal medical image set by using a liver active shape model; 2, defining and extracting direction mode eigenvectors and grayscale range change mode eigenvectors of pixel points in the liver target region; 3, performing liver pixel density clustering center definition and clustering division based on a two-dimensional space of local density of the pixel points in the liver target region and a minimum distance from one pixel point to another pixel point with the higher density; and 4, performing image detection of the liver injury type based on an active learning method of liver pixel density clustering edge, clustering center and clustering division.

The invention discloses a traffic jam point discovery method based on Spark cluster parallel computing. The invention discloses a trajectory big data mining method, relates to the technical field of trajectory big data mining, solves the technical problem of how to quickly cluster massive trajectory data and find a traffic congestion source of an urban hot region, and comprises the following steps: (1) preprocessing the massive data, including complementing errors between trajectory points and eliminating trajectory point redundancy in a certain region for a period of time; (2) adopting a grid-means clustering algorithm to carry out clustering to obtain a plurality of target data samples; (3) clustering by adopting a neighborhood maximum density grid clustering algorithm to obtain an urbanhot traffic road network model; and (4) calculating the adjacent grid density difference by adopting a neighborhood density difference algorithm to obtain a traffic jam source, and storing the obtained result in a Spark memory. According to the method, mass data can be quickly clustered to obtain an urban traffic network model, and quick discovery of a traffic jam source region is realized.

The invention relates to the technical field of Internet, and discloses a method, a system and a server for fast clustering advertising orders. The comprises the following steps: extracting scenic global semantic features of advertising images from the advertising images included in the advertising orders; according to the scenic global semantic features of the advertising images, acquiring image fingerprints of the advertising images; according to the acquired image fingerprints of the advertising images included in any two advertising orders, calculating similarity between the any two advertising orders; judging whether the similarity between the any two advertising orders surpasses a preset similarity threshold, and if so, storing the any two advertising orders into an advertising order set corresponding to the same clustering identifier in a database. Through the adoption of the method, the system and the server, fast clustering of different advertising orders including identical or extremely similar advertising images can be achieved, so as to improve the advertising effect of the advertising orders.

The invention discloses a Laplacian centrality-based peak clustering method. The method comprises the steps of firstly converting a to-be-classified data set into a weighted complete graph, taking each data point as a node, and taking a distance between two data points as a weight value of an edge between the corresponding two nodes; calculating and evaluating local importance of each node of a network by using Laplacian centrality; obtaining cluster centers, wherein the cluster centers are nodes with local density centers with higher Laplacian centrality in comparison with neighbor nodes around, and relatively long distances from nodes with higher Laplacian centrality; and applying class cluster tags to all the class cluster centers, attributing a node free of the class cluster tag to a class cluster which a node with a shortest distance and the class cluster tag belongs to, and obtaining the tag of the class cluster. According to the method, an algorithm for evaluating the importance of the nodes of the network is introduced in the aspect of evaluating the importance of the to-be-classified data points; the algorithm accuracy is high; and parameter-free quick clustering can be realized.

The embodiment of the invention provides a live broadcast room recommendation method and a related device, which are used for quickly recommending a live broadcast room to a user. The method comprisesthe steps: obtaining a user watching sequence of a live broadcast platform in a preset time period; determining a transfer weight between live broadcast rooms in the live broadcast platform accordingto the user watching sequence; determining a first directed graph of the live broadcast platform according to the transfer weight between the live broadcast rooms in the live broadcast platform; cutting the first directed graph based on a preset out-degree and/or a preset in-degree to obtain a second directed graph; calculating the transition probability of each node in the second directed graph;calculating the target arrival probability of each node in the second directed graph according to the transition probability of each node in the second directed graph; determining live broadcast roomclustering of the live broadcast platform in a preset time period based on the target arrival probability of each node in the second directed graph; and recommending a live broadcast room to the target user according to a set of live broadcast rooms watched by the target user in the preset time period and the live broadcast room clustering of the live broadcast platform.

The invention relates to a marine route network generation method and device and a computer storage medium, and the method comprises the following steps: obtaining the ship trajectory of each ship in an observation water area in a certain period of time; extracting feature points in the ship trajectory to generate a feature trajectory; performing density clustering on each feature point to obtain a plurality of clustering clusters and corresponding clustering center points; taking the clustering center point as a marine route point, replacing each feature point in the feature track with the marine route point of the corresponding clustering cluster, and generating a marine route track; and sequentially connecting the marine route points in the marine route track to generate themarine route network. The marine route network is constructed based on the ship trajectory, and the construction precision is high.

According to the DBSCAN algorithm shore bridge state classification method based on principal component analysis, sensors are installed at all positions of a shore bridge, and data transmitted by thesensors are extracted at set time intervals; Through principal component analysis, after data centralization, a characteristic covariance matrix and characteristic values and characteristic vectors thereof are obtained, the contribution degree of each component is calculated, and a plurality of previous-order principal components are taken to carry out matrix transformation; And initializing a density threshold value and a density radius for data point clustering, realizing shore bridge state classification according to a clustering result, and realizing shore bridge state monitoring. Accurateand rapid clustering of the quay crane state is achieved, non-circular domain distribution data can be clustered, a good effect is achieved, compared with a common DBSCAN algorithm, time complexity is reduced, clustering efficiency and accuracy are improved, and abnormal data can be well recognized.

The invention discloses a face image processing method and system, a server and a readable storage medium. The method comprises the steps of receiving the image information uploaded by a first terminal, and preprocessing the image information; extracting a face image from the image information according to the preprocessed image information; and clustering the face images, and storing a clusteringresult in a database. By preprocessing the image information, extracting the face image of the users, and clustering the face images matched with the users correspondingly, obtaining the clustering result and storing in the database, the similar or same face images can be quickly clustered through the clustering processing, and if the users need to download, the corresponding face images can be quickly obtained through the clustering result, so that the image information processing efficiency of a server is improved.

An image processing clustering method provided by the invention comprises the following steps: reducing the size of a mask image to obtain data to be clustered, and generating a first clustering dataset; performing DBSCAN clustering processing on the first clustering data set to obtain a cluster region, deleting semantic annotations in the first mask image based on the cluster region, and re-extracting data needing to be clustered of the first mask image; and combining the clustering results of the two parts to obtain a final clustering result. According to the method, rapid clustering of large-scale samples can be realized, and features of part of small-area features covered by large-range features can be re-extracted, so that the accuracy of a clustering result can be ensured. Comparedwith an existing clustering processing method, the problem that a small number of features in a small partial region range are omitted in the clustering processing process can be avoided. The invention further provides an image processing clustering system and an electronic device, and the image processing clustering system and the electronic device have the same beneficial effects as the method.

The invention relates to a data quality governance system based on a message bus. The system performs the steps of receiving power monitoring data sent by a plurality of data sending ends through themessage bus and stored in a classified mode; determining a power monitoring data quality evaluation index, and constructing a data quality rule verification library; clustering the power monitoring data by adopting a k-means clustering method optimized on the basis of a grey wolf longhorn beard search algorithm, and dividing the power monitoring data into a plurality of data subsets; for each datasubset, carrying out accuracy evaluation through an LOF abnormal value detection algorithm, and carrying out integrity, consistency, availability and real-time evaluation through data quality rule verification; and calculating the weight of each evaluation index through an improved analytic hierarchy process, and calculating a comprehensive evaluation result of the power monitoring data quality.According to the system, the multi-dimensional quality evaluation of the power monitoring data is realized, and the reliability of the quality management of the power monitoring data is improved.

The invention relates to the field of classification, in particular to a short message clustering method based on a text structure, the method comprising the following steps: S1, preprocessing word segmentation of short messages in a short message set; S2, calculating similarity degree based on short message structure for any two short messages pretreated by word segmentation, and constructing similarity degree matrix; S3, carrying out a clustering process, comprising: comparing each value in the similarity matrix with a preset similarity threshold value respectively; when the similarity valueis greater than the similarity threshold value, classifying two short messages corresponding to the similarity value into the same category, and finally obtaining a short message clustering result.

The invention provides a data stream clustering method and device based on a density peak value, and relates to the technical field of data processing. According to the method, a Jaccard similarity distance and a Gaussian kernel function are combined to calculate the local density of an evolved data stream, a heuristic strategy for kernel density estimation is introduced, and a local clustering center point is selected from points with large local density. And finally, local clustering center points are recorded through a variable-scale bucket sequence, and when a clustering request arrives, the local clustering centers are merged to obtain a global clustering center. According to the method, the concept drift in the data stream can be identified, the data stream is quickly clustered, and good clustering performance is achieved.